Dr Daniel Roggen, Director of the Sensor Technology Research Centre
22nd Jun 2017
Towards unbounded activity&context awareness in wearables and ubicomp
Human Activity Recognition (HAR) is a fundamental technology in wearable computing and more generally in ubiquitous computing which enables applications in health/wellbeing, sports, industrial assistance and other domains.
Daniel will describe the work he has done towards more robust, adaptive and scalable HAR. Daniel will show the work on an opportunistic recognition pipeline which dynamically adapts to the sensors available in the user's surrounding. The highlights include the automatic translation of recognition models from one sensor modality to another one, or the brain-guided adaptation of recognition models.
Daniel will show the benefits of Deep Learning for HAR where we showed that wearable datasets are large enough for deep network training and we demonstrated significant performance improvements on benchmark datasets. Daniel will show that transfer of layers in a deep network can be used to reduce training time at equivalent performance. This indicates a path towards a universal and reusable "feature basis" for HAR.
In contrast to high-performance recognition, some of our work also focuses on low-power algorithms which may eventually be part of a sensor front-end in an ASIC, in particular a template matching algorithm running at than 120uW.
Lastly Daniel will touch on our recent work towards "open-ended" activity recognition: systems able to discover and classify activities beyond a pre-defined set, which may play a fundamental role in scenarios of high societal value, such as memory assistants for people with dementia.
This work has been supported by EPSRC, Google FRA, NVidia, Huawei, EU FP7 and others.
Dr Alex Casson, Lecturer in Sensing, Imaging and Signal Processing
15th Jun 2017
This talk will overview the work of Dr Casson’s group at the University of Manchester, creating next generation ultra low power human body sensors, and how these relate to the SPHERE project. Future ‘wearables’ will be flexible and conformal to the body, and the talk will demonstrate temporary tattoo sensing patches that attach directly to the skin and intrinsically obtain very high quality signal connections. Next generation sensors will also be integrated with electronic health records to enable dynamic personalised care planning, and the talk will overview a recently started EPSRC funded project working to enable this, without compromising battery life. Further, the talk will discuss our work on sensors which go beyond ‘just collecting data’. Wearable sensors are currently seen as ‘one-way streets’ only for data generation. There is substantial potential for sensor systems which analyse the collected data ‘on-electrode’ and allow data driven personalisation of outputs. For example, monitoring physiological parameters and performing highly targeted treatment release within a few milliseconds of a trigger event, where there is not time to send the signals to a data centre for conventional analysis. An example will be given demonstrating targeted sound stimulation played during sleep as a non-pharmacological intervention for slowing the trajectory to mild cognitive impairment, as first case study to using wearables to automatically deliver targeted data responsive treatments. Finally, the talk will overview our work with the SPHERE partnership fun
Seminar: Dr Nicolas Tsiftes, Networked Embedded Systems Group, Computer Systems Laboratory at RISE SICS 1st Jun 2017
E-care@home: Leveraging the Internet of Things to support independent living of the elderly.
E-care@home is a Swedish 5-year project that strives to create a comprehensive e-health system for smart homes to support independent living of elderly. One of the main visions of E-care@home is to help address the demand for elderly care services over the next decades, as the proportion of elderly in the population of developed countries continues to increase. In the first part of this talk, I will give an overview of the E-care@home project, and introduce its distributed, interdisciplinary research environment. This part will also cover the work carried out by the various partners in E-care@home on topics such as artificial intelligence, semantic web, and health sensors. In the second part of the talk, I will focus on our work at SICS, which primarily involves the IoT architecture in the E-care@home system. This architecture comprises standard IoT protocols such as LWM2M, RPL, and TSCH. Additionally, I will present our work on software safety and security for IoT applications, including software virtualization, formal verification, and runtime assurance.the cardiovascular domain. Finally,
Professor Alejandro Frangi, Director - Centre for Computational Imaging & Simulation Technologies in Biomedicine (CISTIB) 25th May 2017
Image-based Cerebrovascular Modeling for Advanced Diagnosis and Interventional Planning
Current technological progress in multidimensional and multimodal acquisition of biomedical data enables detailed investigation of the individual health status that should underpin improved patient diagnosis and treatment outcome. However, the abundance of biomedical information has not always been translated directly in improved healthcare. It rather increases the current information deluge and desperately calls for more holistic ways to analyse and assimilate patient data in an effective manner. The Virtual Physiological Human aims at developing the framework and tools that would ultimately enable such integrated investigation of the human body and rendering methods for personalized and predictive medicine. This lecture will focus on and illustrate two specific aspects: a) how the integration of biomedical imaging and sensing, signal and image computing and computational physiology are essential components in addressing this personalized, predictive and integrative healthcare challenge, and b) how such principles could be put at work to address specific clinical questions in the cardiovascular domain. Finally, this lecture will also underline the important role of model validation as a key to translational success and how such validations span from technical validation of specific modeling components to clinical assessment of the effectiveness of the proposed tools. To conclude, the talk will outline some of the areas where current research efforts fall short in the VPH domain and that will possibly receive further investigation in the upcoming years.
Professor Yvonne Rogers, Professor and Director of UCL Interactive Centre, University College London
18th May 2017
The Joy of Data: in the Home and Beyond
Much has been written about the fear of data – how people are increasingly worried about what is being collected about them, how it is being used and where it ends up. And rightly so. In contrast, in my talk I will examine the joy of data, exploring its potential for enabling people to understand, increase their awareness and act upon it. In particular, I will describe some of the research we have been conducting as part of our Intel-funded, Interdisciplinary Collaboration Research Institute (ICRI) on the urban Internet of Things. A focus of our work is on how various kinds of data can be openly and collectively sensed in people’s homes and outdoors with the aim of empowering communities. As part of our approach we have been designing novel physical interfaces, intended to entice and encourage the general public to use and be able to readily infer what is behind the data visualisations they help to create. I end, by asking is our approach desirable, ethical or helpful to society?
Professor Antonis Bikakis, Senior Lecturer in Department of Information Studies, University College London
30th March 2017
Rule-based reasoning in Ambient Intelligence: Theory and Applications
Context, Context Representation, and Contextual Reasoning are central notions in the Ambient Intelligence vision to transform our living and working environments into "intelligent spaces". Ontology-based models have been argued to satisfy well the requirements of context representation. Rule-based reasoning has already been successfully integrated into ontology-based applications for domains with similar requirements (e.g. the Web), while it addresses well many of the challenges of reasoning in Ambient Intelligence systems. This talk discusses the challenges of contextual reasoning in Ambient Intelligence systems, argues about the suitability of rule-based approaches, and describes the deployment of rule-based reasoning in two different settings: a centralised, semantics-based context management framework for Ambient Intelligence, and a totally distributed system of logic-based agents.
Dr Iain Bate, Senior Lecturer in Real,Time Systems Group, University of York
16th March 2017
Bringing dependability to networked sensing systems
Networked sensors have been used in a wide range of applications including healthcare systems. A number of researchers and companies have identified dependability as a key concern as either the systems have to be certified by a regulator or they have experience of failed deployments. This talk will discuss some of the problems that might be faced, and describe how our research and experience from years of working in the safety-critical industries (e.g. avionics and automotive) has been and will be applied in future.
Professor Yannis Velegrakis, Director of the Data Management Group, University of Trento
9th February 2017
Improving User Satisfaction during Querying Formulation and Query Answering
In modern business environments analysts and regular users alike need to run queries that help them understand what the data can offer, use the information more effectively, and take better decisions.
To form these queries users need to know the query language which is not always easy for novice users. Furthermore, users often may have an idea of what they are looking but not specific enough to actually express it in a query.
Even when the users are able to express in details what they are looking for, the desire to test very specific situations or the excitement of getting the ideal elements of interest, often lead to over-specified queries that naturally return no answers. Queries with no answers may be frustrating and occur more often in big data volumes where knowledge about what contained in the data is limited.
We present a framework that allows users to query collections using examples and to interact with the system to eliminate empty answers. For the former, it uses a knowledge base to find associations between situations that look highly disconnected at first sight, yet, are of the same nature. For the latter, it uses a probabilistic optimisation framework that is based on the probability that the user will like an answer and that the answer will exist in the database.
Dr Giles Birchley, Senior Research Associate in Healthcare Ethics, University of Bristol
2nd February 2017
Smart homes, private homes? An empirical study of technology researchers' perceptions of ethical issues in developing smart-home health technologies
Studies of the ethics of smart-homes raise concerns about privacy, consent, social isolation and equity of access, but few studies have investigated the ethical perspectives of smart-home engineers themselves. Our study addressed this gap by conducting face to face interviews with research engineers in a large smart-home project, during which they described their experiences and reflect more broadly about ethical considerations important to smart-home design. The interviews were analysed using a thematic approach and two overarching themes emerged: ‘Privacy’: interview participants paid close attention to negative consequences of breaching information privacy, but gave concerns about physical privacy less attention. ‘Choice’: Interview participants often saw provision of choice to end-users as a solution to ethical dilemmas. While participants indicated that choices of end-users may need to be restricted for technological reasons, overtly restricting choice was viewed as problematic. This presentation considers the ethical validity of these positions and some implications for future smart home research.
Dr Kristrun Gunnarsdottir, CRESS Research Fellow and Dr Jie Jiang, University of Surrey
1st December 2016
Making sense of the household
HomeSense is trialling data collection and visualization techniques in UK households using fixed and mobile sensors. The project is cataloguing technical reliability and security issues, the methodological and ethical issues in preparing and carrying out this work both on and off site. It will demonstrate what sensors can add of value into the mix of established methods in household research, and build guidelines for good practice. In this paper, we outline our mixed-methods approach. Walking interviews and records of time use provide valuable insights into household practices such as cooking, mealtimes, sleep, entertainment, home working and house chores. They implicate all sorts of powered devices involved in these practices, but much of household activity can be detected by sensors as fluctuations in electricity use, temperature, humidity, particulates, objects/bodies in range, noise levels, and physical activity. We take examples of how to 'see' household practices in sensor-generated data, what the indications look like in algorithmic signatures, and how other data support what is observed. Finally, we will discuss some of the potentials and limitations in using sensors to attend to specialised research interests in household practice.
Aisling Ann O'Kane- Research Associate, Human-Computer Interaction and Human Factors Engineer, University College London
How Mobile Healthcare Technologies are Actually Used (and Hacked) in the Wild
Adults make choices regarding the technology they use to self-manage their health and wellbeing, and these technologies are often adopted, used and abused in ways that researchers, manufacturers, and clinicians have not accounted for. This talk will give an overview of human-computer interaction research on the situated use of mobile technologies in people’s everyday (and sometimes messy) lives. Accounting for individual choices adults make will be discussed in relation to understanding the use of these devices in the context in which they are actually used, supporting choices through end user customisation, and the emerging trend towards Do-It-Yourself open-source technology innovation that is outpacing healthcare researchers and manufacturers. Examples will be given from research on Type 1 diabetes devices and apps, hearing loss technologies, baby monitoring technologies, and wearables and apps for fitness.
Dr Simon Duquennoy- Research Scientist, SICS, Inria
10th November 2016
On Dependability in Low-power IPv6
Low-power Internet-enabled devices are a key component of tomorrow's Internet of Things (IoT). In this context, the low-power IPv6 stack acts as a hub to heterogeneous communication technologies and applications. A major challenge when going from ad-hoc sensor network solutions to Internet-enabled devices is to achieve the desired level of dependability, in terms of reliability and security. After introducing the low-power IPv6 stack, we review our latest work on IoT dependability. We will discuss security aspects (privacy-preservation and software security), interference mitigation, and reliable medium access and routing. The work we show is grounded in reality, with results that are implemented (based on standards and released as open source) and validated experimentally, at scale (hundreds of nodes).
Dr Helge Wurdemann- Lecturer in Medical Devices, University College London
3rd November 2016
Soft robots are too floppy!
The field of soft robotics has made inroads into a number of applications that have previously been dominated by rigid robots such as assistive technologies, surgical robotics and haptics. The core challenge when employing soft robots is however how to exert effective forces against the environment and how to achieve on-demand stiffness.
This talk will concentrate on a number of stiffness mechanisms for soft robotic systems. These type of mechanisms have been applied to soft robotic manipulators for minimally invasive surgery, to robotic solutions to work closely together with the human in industrial settings, and to feedback actuators for soft tissue palpation tasks.
The talk will explore the synergies that will arise from bringing together soft robotics and stiffness mechanisms and identify the advantages and areas of applications that these new capabilities bring.
Roger Whitaker, Professor of Mobile and Biosocial Computing, Cardiff University
20th October 2016
Smartphones and research "in-the-wild"
In this talk we present a high level overview of research concerning “in-the-wild” observation and participation through smartphones. The relationship between smartphones and their users is now very strong indeed, proving opportunity for implicit and explicit observation of human behaviour and feedback. We present philosophical perspectives on the role and potential of in-the-wild studies. We also present case studies on recent "in-the-wild" research concerning interruption, mobility, personality, and well-being.
Talk by Professor Ian Craddock about the SPHERE project
20th October 2016
With a growing elderly population and an increase in people affected by long-term health conditions, healthcare will need to change. SPHERE is an engineering research project developing sensor systems to provide accurate information about health-related behaviours. Future applications include detection of early stage dementia and enabling self-management of health. Hear how technology can help keep cities healthy at a talk by SPHERE director, Professor Ian Craddock. Free to attend, advance booking required.
Dr Thomas Watteyne- Research Scientist & Innovator, Inria
6th October 2016
From Smart Dust to 6TiSCH: Academic and Commercial Background on TSCH Technology
TSCH technology is nothing new. It was invented as part of the 1997 "Smart Dust" project, lead by UC Berkeley Professor Kris Pister. Because of its unmatched performance, it was quickly adopted as the de-facto technique for industrial process monitoring, and standardized in WirelessHART, ISA100.11a and IEEE802.15.4-2015. Tens of thousands of TSCH networks operate today in applications ranging from industrial process monitoring and urban parking management to building automation and environmental monitoring. Commercial products such as Linear Technology's SmartMesh IP offer over 99.999% end-to-end reliability and over a decade of battery lifetime. The IETF 6TiSCH working group is currently standardizing how to use IPv6 on top of TSCH. Through the OpenWSN open-source reference implementation, TSCH has been adopted by all major open-source projects and is widely regarded as the future of low-power wireless. This presentation serves as an overview of where TSCH technology comes from, what the TSCH ecosystem looks like today, and what future directions and challenges are. We will look at TSCH both from the academic and commercial sides. We will in particular stress that proven off-the-shelf products have been around since 2006, and offer performance light-years ahead of any academic work. Not "reinventing the wheel" is the message. The presentation will be punctuated with real-world customer examples, live demos and detailed performance numbers. Examples will be drawn from Berkeley OpenWSN, the reference open-source implementation and Linear Technology's SmartMesh IP product line, the market leader.
Dr Hristijan Gjoreski- Research Scientist, Department of Intelligent Systems, Jozef Stefan Institute
29th September 2016
Activity recognition with wearable sensors
Human activity recognition is a basic building block in numerous applications, such as: smart homes, robotics, health care and similar. Automatic recognition of daily activities and estimation of energy expenditure may assist with proper management of pathologies such as obesity, diabetes and cardiovascular diseases. In recent years, wearable accelerometers have proven successful for recognizing activities, and are probably the most mature technology for this purpose. The reason for this is that they are capable of measuring human motion (mainly by measuring the linear 3D accelerations) and estimating body postures (mainly by measuring the orientation with respect to the Earth’s gravity). In this presentation various approaches for activity recognition will be presented: machine learning, expert rules, and deep learning. Additionally, a discussion will be provided about the various accelerometer placements on the body and the influence on the accuracy.Finally, three applications that are built on top of the activity recognition module will be discussed, in particular: fall detection, energy expenditure estimation, and stress detection.
Dr Val Mitchell, Programme Director for MA Interaction Design/User Experience Design and Dr Vicky Haines, Head of User Centred Design Research Group, Loughborough University
22nd September 2016
Applying user centred design in domestic energy demand projects: Experiences from Loughborough Design School
This presentation, by Dr Vicky Haines and Dr Val Mitchell from Loughborough University’s Design School, will introduce a selection of recent domestic energy demand research projects in which they have taken a user centred design approach. A range of novel methods used in the research will be described, together with experiences of deploying prototypes and monitoring equipment in occupied homes, at small and large scale. The talk is intended to give an introduction to the range of research being undertaken by Dr Haines and Dr Mitchell, as well as providing reflections on the complexities of energy demand research in real homes.
Kristina Yordanova, Senior researcher, University of Rostock; Research Associate, University of Bristol
4th August 2016
Computational State Space Models for Activity Recognition
Assistive systems support the daily activities and allow even people with impairments or medical conditions to continue their independent life. Such systems have to recognise the user actions and intentions, track the user interactions with a variety of objects, detect errors or changes in the user behaviour, and find the best way of assisting them. One approach for modelling and recognising daily activities in home settings is to describe the human behaviour in the form of rules. These rules are then used to generate probabilistic models with which the system can infer the user actions, goals and context. Such types of models are also known as computational state space models (CSSMs). In this talk I will present the concept of CSSMs for activity recognition. Furthermore, I will discuss their practical applicability for activity recognition problems in home settings.
One drawback of CSSMs is that they have to be manually developed. This is often a time-consuming and error-prone process. To address this problem, I will also present an approach that aims at automatically learning the model structure from textual instructions. The approach reduces the time and effort needed to develop CSSMs. It also potentially allows for even inexperienced model designers to use CSSMs for activity recognition problems.
Chris Efstratiou- Lecturer in Ubiquitous Computing, University of Kent
28th July 2016
Sensing people: applications of mobile sensing in the workplace and in healthcare
Mobile sensing technologies have the potential to capture a range of human activities in a continuous and unobtrusive manner. Smartphones and wearable sensing devices can allow the detection of physical activities and social interactions of people. Such technologies can help us develop systems to get a better understanding of the lives of individuals or groups of people. In this talk I will present our efforts in using people-centric sensing technologies in two application domains: "Quantified workplace" considers the use of sensing technologies to get a better understanding of how organisations operate. I will present our results from a number of deployments of sensing technologies to capture social interactions in the workplace. In healthcare wearable sensing technologies have the potential to change the way we support people with long term health conditions. I will be presenting our ongoing work in using sensing technologies to help patients with epilepsy, and dysphagia.
Seminar- Kevin Wells, Reader in Medical Imaging, University of Surrey
30th June 2016
Motion sensing in healthcare
Surrey’s Centre for Vision, Speech & Signal Processing has developed a unique capability in non-contact measurement of people, and data analysis methods for understanding people: i.e. human shape, motion behaviour, principally built for the creative/entertainment industries. Much of this work previously relied on marker-based measurements, but is increasingly being replaced with marker-less low-cost consumer grade technologies (e.g. Kinect). Parallel pilot work in the Medical Imaging group within CVSSP is now seeing the development vision technologies for measuring people being developed into the healthcare area, in conjunction with the Royal Marsden Hospital (in cancer), Broadmoor NHS Trust (in psychiatric care) and Surrey Sleep Research Centre (sleep apnoea) based around low-cost 3D camera technology. The talk will describe a range of applications and then focus on measurements of respiratory motion using marker-based and marker-less methods. This approach has applications in patient positioning, respiratory motion management in imaging and radiotherapy and respiratory medicine.
Seminar- Prof Mark Gillott, Co-Director of the Institute of Sustainable Energy Technology & Dr Lucelia Rodrigues, Associate Professor, Architecture, Energy and Environment from the University of Nottingham
23rd June 2016
The Sense and Sensibility of Smarter Sustainable Homes
The Architecture, Energy and Environment Research Group aims to inform the sustainable practice of architecture and engineering, in order to enhance the quality of the built environment through world leading fundamental and applied collaborative research. The work undertaken by the group is predominantly related to mitigating the impacts of, and adapting to, climate change, reducing energy use in the built environment, and enhancing comfort, productivity and wellbeing of building users.
The Creative Energy Homes is a multi-award winning research and educational project of innovative energy efficient solutions for sustainable homes and communities. The project provides a living lab for leading firms, including E.ON, David Wilson Homes, BASF, Roger Bullivant, Tarmac, Saint-Gobain and Igloo Blueprint to work with the University of Nottingham to investigate energy efficient solutions for buildings and neighbourhoods.
The seven smart houses built under the project title, at Green Close on the University Park campus, serve as live laboratories to investigate the inevitable relationship between occupants and energy performance. The project also explores the feasibility of smart-grid energy technology to meet the next generation housing demand.
Sensing is a very important part of the research undertaken at the project. The houses are fit with (mostly) unobtrusive domestic occupancy measurement technologies and actuators for domestic automated demand response. Several types of sensors have been developed and employed by researchers, aiming at recording and managing energy use, and understanding the impact of human behaviour.
The talk will briefly describe the work of the research group, followed by a more in-depth description of the Creative Energy Homes focused around sensing technology for smart sustainable homes.
Finn Kensing, Director at Center for IT Innovation, University of Copenhagen
16th June 2016
Designing for self-monitoring patients and better-informed clinicians
Abstract: The talk is based on two research and development projects one of which is still running. The project partners include cardiologists, nurses, lab technicians, patients, computer scientist, social scientists, a Danish software house, and an American supplier of cardiac devises. The project is funded by the partneres and a Danish research council.
The purpose of the projects is to enhance the communication and collaboration between heart failure patients, who are remotely monitored via implantable cardiac devices, and the diverse set of clinicians that takes care of them.
I’ll present the research questions, the research setting, the methodological approach, and the current results.
Seminar- Emma Stack, Principal Research Fellow (SPHERE), University of Southampton
9th June 2016
Do Different SPHERE Movement Sensors Agree with the Gold Standard? Can Sensors Identify when Someone Is At High Risk of an Imminent Fall?
Our first collaborative study within SPHERE revealed that people with Parkinson’s at very high risk of falling are highly sedentary, perhaps because their balance is threatened whenever they stand up, walk, turn, tackle steps or other tasks in standing or when they sit down again. Most, if not all, of these actions are necessary when following the habitual path from favourite chair to the bottom of the stairs, so we suggest that researchers focus observation on this route to provide useful data about mobility without compromising people’s privacy by covering their entire homes with movement sensors. That study revealed something else equally important about evaluating mobility and balance using sensors in the home: whatever their advantages over a human observer (and they certainly have some), sensors are useless if their ‘view’ of the target is obscured, if they were in the wrong place to begin with, or if they are not recording when they are supposed to be recording. Therefore, in our second and third collaborative studies, we have recruited further participants at high risk of falling and invited them to help answer the questions posed in the title. To establish the extent to which data from SPHERE-type sensors (Kinect and wearable) agree with the Gold Standards (CODA 3-D motion analysis technology and experienced/expert clinicians/observers), we set out to record participants performing 12 mobility and balance tests while the sensors recorded under ‘optimal’ conditions (i.e. operated by a specialist, while participants performed isolated movements in unobstructed view). Part of the work was conducted in a movement laboratory and part in the much more challenging residential environment. Whilst to validate agreement the scientists evaluating different sensors must remain blind to each other’s evaluations, eventually data from the Gold Standards (CODA and the experts) will allow the development and refinement of video and accelerometer-based algorithms that will facilitate ‘automatic’ recognition of movement type, quality and stability. To test the possibility that wearable sensors can identify ‘near-misses’, ‘wobbles’ and other signs of imminent fall-risk at least as well as an observer, we have recorded other participants walking repeatedly between their chairs and stairs at home while wearing five sensors over the lumbar spine and extremities. It is possible that the accelerometry will allow researchers to identify aspects of impending instability that the participants reported or perceived but that would not be observable by anyone else, such as ‘dizziness’ or ‘fatigue’. This would mark a potentially significant improvement in healthcare: instead of relying on history, or trying to replicate symptoms after the event, SPHERE technology could record instability in real-time, aiding both communication and timely intervention. The analysis of both studies is ongoing but we will present preliminary findings.
Seminar-Joshua Vest, Adjunct Assistant Professor of Healthcare Policy and Research, Weill Cornell Medical Univesity
26th May 2016
My patient is where? Automatically notifying providers of patients’ hospitalizations and emergency department visits: fit within the organization and impact.
Event notifications (also called alerts or subscription services) use the hospital's ADT system to alert providers about subscribed patients' readmissions or emergency department visits. This seminar will present findings from an evaluation of an event notification system in a large urban center. Specifically, the evaluation studied the effects of organizational capability and related social/organizational issues upon users’ perceptions of the impact of event notifications on quality, efficiency, and satisfaction, as well as quantifying the impact of event notification on readmission rates.
Children of the Nineties
26th May 2016
The SPHERE project is looking forward to talking to participants of the 'Children of the Nineties' study about the project.
If you have been involved in the Children of the Nineties study, we are running an event to talk about SPHERE and to demonstrate our technology and would love to talk to you about it.
The event will held at Colston Hall on Thursday 26th May, and everyone who has been invovled in the Children of the Nineties Study is warmly invited to attend.
For more information, please see: http://www.irc-sphere.ac.uk/friends-of-sphere-alspac
Seminar- Carl Henrik Ek, Lecturer in Computer Science, University of Bristol
12th May 2016
Explaining away with factorisation
The beauty and the bane of models is the concept of explaining away, it is meant to allow us to focus our modelling effort on what we are really interested in by factorising the "model" into a noise and a signal component. However, there are often a significant portion, in many application the dominant part, of the signal that is irrelevant for the inference task that we wish to solve. The effect of this is that the model spends most of its effort on modelling, for the task, "irrelevant" variance. To circumvent this a common approach is to make a trade with the devil and directly construct the posterior not deriving it from a likelihood and prior. This leads to a loss of interpretability and regularisation due to lack of uncertainty propagation making us use data in a inefficient manner.
Seminar- James Hensman
5th May 2016
Variational Inference in Gaussian Process Models
Gaussian process models are widely used in statistics and machine learning. There are three key challenges to inference that might be tackled using variational methods: inference over the latent function values when the likelihood is non-Gaussian; scaling the computation to large datasets; inference over the kernel-parameters. I’ll show how the variational framework can be used to tackle all of these. In particular, I’ll share recent insights which allow us to interpret the approximation in an elegant and straightforward way, using variational Bayes over stochastic processes. Finally, I’ll outline how this technology can be used to help tackle contemporary problems in biostatistics.
Seminar- Aftab Khan, Toshiba Research Europe
28th April 2016
Beyond activity recognition: automated skill assessment from accelerometer data
The next generation of human activity recognition applications in ubiquitous computing scenarios focuses on assessing the quality of activities, which goes beyond mere identification of activities of interest. Objective quality assessments are often difficult to achieve, hard to quantify, and typically require domain specific background information that bias the overall judgement and limit generalisation. In this seminar, I will present a framework for skill assessment in activity recognition that enables automatic quality analysis of human activities. Our approach is based on a hierarchical rule induction technique that effectively abstracts from noise-prone activity data and assesses activity data at different temporal contexts. Our approach requires minimal domain specific knowledge about the activities of interest, which makes it largely generalisable. By means of an extensive case study we demonstrate the effectiveness of the proposed framework in the context of dexterity training of 15 medical students engaging in 50 attempts of surgical activities.
Seminar- Leandro Pecchia, Assistant Professor at the University of Warwick
14th April 2016
Actigraphy and physiological monitoring for sleep patterns and accidental falls prediction
Biomedical signals reflecting autonomous nervous system (ANS) and cardiovascular system (CVS) could be promising means to investigate sleeping patterns, assess the risk of falling and to predict accidental falls in the short term. Several studies demonstrate how those signals can be used to predict cardiovascular and cerebrovascular events , to assess their severity , also in remote monitoring applications . More recently, the association between Heart Rate Variability (HRV) and risk of falling was investigated, with a retrospective study enrolling 168 subjects over 65 years of age . Forty-seven of them experienced fall events within 3 months since registration. Using data-mining methods, abnormal HRV patterns were discovered and the odds ratio (OR) of falling in subjects exposed to these patterns was analyzed. The OR of falling in subjects presenting these depressed HRV pattern was significant (OR 4.32, CI 95% 1.61-11.56, p<0.01), suggesting that a depressed HRV was associated with an increased risk of falling . A Multinomial Naïve Bayes classifier achieved satisfactory results through a rigorous validation procedure, enabling to predict fallers within three months from the baseline recordings with a sensitivity rate of 72% and a specificity rate of 61% . The proposed method was based on 1 hour ECG recording and short term HRV analysis, which are already in use in many outpatient clinics and could be used widely in outpatient settings to identify high-risk patients who need further assessment and could benefit from fall prevention programs.
Starting from these results, we focused on the prediction of specific risk factors that are clearly dependent form ANS/CVS states, which can be reliable detected using HRV. In a recent study, we developed, validated and patented a mathematical model to predict orthostatic hypotension by analyzing HRV features recorded in the few minutes (from 2 to 5min) before rising from a bed or chair, which account for the 30% of indoor falls. This model predicted the drop down of blood pressure with an error below 4mmHg, which is the measurement error .In the past few months, we have been investigating sleeping patterns, using wearable sensors (i.e. ECG, 3d accelerometers) in relation to balance problems. This seminar will give an overview of those results, the methods used and their potential future applications and limits for remote home monitoring.
1. P. Melillo, R. Izzo, A. Orrico, P. Scala, M. Attanasio, M. Mirra, N. De Luca, L. Pecchia (2015) Automatic prediction of
cardiovascular and cerebrovascular events using Heart Rate Variability analysis, PloS one, 10(3): e0118504.
2. L Pecchia, P Melillo, M Sansone, M Bracale (2011) “Discrimination power of short-term heart rate variability measures for CHF
assessment”. IEEE Transactions on Information Technology in Biomedicine 15 (1):40-46. doi:10.1109/TITB.2010.2091647
3. L Pecchia, P Melillo, M Bracale (2011) “Remote Health Monitoring of Heart Failure With Data Mining via CART Method on
HRV Features”. IEEE Transactions on Biomedical Engineering, 58 (3):800-804. DOI 10.1109/Tbme.2010.2092776
4. Melillo, P., Orrico, A., Scala, P., Crispino, F., and Pecchia, L. (2015). Cloud-Based Smart Health Monitoring System for
Automatic Cardiovascular and Fall Risk Assessment in Hypertensive Patients. Journal of medical systems, 39(10), 1-7.
5. Melillo P, Jovic A, De Luca N and Pecchia L (2015) An automatic classifier based on heart rate variability to identify fallers
among hypertensive subjects, IET Healthcare Technology Letters, 2 (4): p 89-94, DOI: 10.1049/htl.2015.0012
6. R. Castaldo, P. Melillo, R. Izzo, N. De Luca, L. Pecchia, Fall prediction in hypertensive patients via short-term HRV Analysis,
IEEE Journal of Biomedical and Health Informatics, (Accepted 11th of March 2016).
7. Sannino G, Melillo P, De Pietro G, Stranges S, Pecchia L, (2015) “Short term heart rate variability to predict blood pressure
drops due to standing: a pilot study”, BMC Medical Informatics and Decision Making, 2015, 15(S3):S2
Seminar- Julie Barnett, Professor of Health Psychology, University of Bath Connecting data for policy: learning in partnership
17th March 2016
The drive to derive and use knowledge from connecting data for policy innovation is being shaped at the intersection of a number of agendas: requirements for evidence, for transparency and accountability, spending restrictions and demand for services. Requirements for sharing the data to be connected need to be negotiated across and within organisational boundaries. Information governance protocols and procedures are derived or used whilst attending to the anticipated and imagined - if not the actual - sensibilities of the citizenry. These agendas do not always sit in easy alignment with each other, and tension and disjunctions between them are realised acutely where new uses of data in the policy making process are practiced at the local level. Since April 2014, Bath and North East Somerset (B&NES) Council, NHS B&NES Clinical Commissioning Group (CCG) and an interdisciplinary team at the University of Bath Institute for Policy Research (IPR) have been involved in a co-produced research project to explore the potential for connected data to inform citizen-focused local policy and practice. In this presentation I will highlight and discuss some of the issues we have identified and consider the implications of these for developing social science agendas around data and policy as well as for policy development practices.
Seminar- Giles Birchley - Senior Research Associate in Healthcare Ethics, University of Bristol
10th March 2016
The Centre for Ethics in Medicine is the hub for bioethics research at the University of Bristol. But what is bioethics, and what methods are utilised in bioethics research? This seminar discusses the work of the Centre for Ethics in medicine, it's various collaborations in the region and its relevance to the SPHERE project.
Seminar- Ruth White - PhD student, University of Reading
3rd March 2016
Investigating Eating Behaviours using Topic Models
Chronic conditions, such as diabetes and obesity are related to quality of diet. However, current research findings are conflicting with regards to the impact of snacking on diet quality. One reason for this is the lack of a clear definition of a snack or a meal. This talk will present a novel approach to understanding how foods are grouped together in eating events using a machine learning algorithm, topic models. Different approaches for applying topic models to a nutrition application will be discussed. The results and analysis of implementing a topic model for the UK National Diet and Nutrition Survey Rolling Programme and National Diet and Nutrition 2000 Adult Survey datasets will be shown. The results demonstrate that the topics found are representative of typical eating events in terms of food group content and associated time of day. There is a strong potential for topic models to reveal useful patterns in food diary data that have not previously been considered. Planned future work using the Irish National Adult Nutrition Survey dataset will also be considered.
Seminar- Gustaf Hendeby - Associate Professor and Docent in Automatic Control, Linkoping University
18th February 2016
Extracting Information from Inertial Measurements
In recent years technological advances have made sensors smaller and more affordable. Today, especially "simple" sensors, such as inertial measurement units (IMUs), can now be found almost everywhere and are very affordable. These sensors can be a great source of information, but it is often required to combine several of them to provide the type of answers sought. Sensor fusion provides a powerful framework to combine information from several sources. IMU based results from human motion capture and simultaneous localization and mapping (SLAM) will be used to exemplify this. Methods for motion capture have many usages ranging from animations in movie making to safe home based rehabilitation. One method to do this without the need for external and limiting infrastructure is to put IMUs on the limbs that should be tracked, and then from the measurements derive the pose. However, treating each limb (IMU) separately is insufficient, as the result will be inherently poor. It will be shown that a solution is to treat the whole pose simultaneous and this way introducing domain knowledge, resulting in state-of-the-art performance. Using the sensor fusion framework furthermore allows for easy inclusion of additional sensors when needed. Another problem where IMUs are commonly used is SLAM, e.g., to capture quick movements in vision based SLAM solutions. Pure IMU solutions for navigation suffer from drift, which is considerable when using consumer grade IMUs. It will be shown how posing the problem in a sensor fusion framework and introducing domain knowledge about the problem makes using IMUs alone a viable solution for short term localization. The framework then allows for straightforward inclusion of additional measurements, e.g., ultra wide band (UWB) measurements of opportunistic received signal strength (RSS) measurements, that can provide surprisingly accurate results without any pre-installed infrastructure.
Seminar- Antonio Skarmeta - Information Systems, Computer Security and Reliability, Computer Communications, University of Murcia
11th February 2016
Security and Privacy Challenges in IoT and Smart Cities
The extension of the Internet to smart things is estimated to reach 50 to 100 billion devices by 2020, defining the so-called Internet of Things (IoT). However, the large-scale use of IoT creates the need to address trust and privacy functions adequately. In that sense, the vision of sensors must evolve into an integrated view of smart objects forming part of our personal space, and as being shared, borrowed and, in general, having temporal associations with the users and their personal identities, with these aspects addressed while considering security and privacy rules.There are several research efforts to be undertaken in coming years in order to realize this vision, a selection of which will be discussed in this talk like Operational Aspects of Authentication, Authorization, Access control and Privacy management, taking into account approaches like Distributed Capability-Based Access Control (DCapBAC) and its integration with authentication protocols
An Iinvitation to Help Shape SPHERE’s Future
30th January- @bristol
You are warmly invited to At-Bristol's education space on Saturday 30th January 2016 to find out more about SPHERE's sensors for home healthcare and how you can get involved in helping develop them.
The event will be suitable for both adults and children, with plenty of activities to explore how the technology works. The afternoon will include the chance to see a FREE show at
the planetarium. If you are unable to attend on this date, but are still keen to get involved, we will be hosting addition events later in 2016. Please note that the event will be held at At-Bristol but does not include entry to the exhibits.
What is SPHERE? • SPHERE is an exciting five-year research project using sensor technology in people’s homes to help solve health problems • This information could be used to tackle lifestyle-related conditions such as asthma, obesity and diabetes • It could also help spot medical issues such as strokes or predict falls How does it work? The sensors are small and will be placed in rooms around the house. They will gather information about the environment – such as air quality, humidity and energy use. Video and wearable sensors will measure people’s movement around the house.They will also be used to track breathing patterns. This information can be used to help with a range of healthcare and wellbeing issues.
Seminar- Meelis Kull- Senior Research Associate, University of Bristol
28th January 2016
Context aware mining for SPHERE
There are many aspects of our well-being which benefit from the attention of health professionals, but often this attention is not received due to the lack of resources or lack of information about us. It is simply infeasible to get health professionals follow us around the clock to give us better advice. As a solution, equipping homes with sensors can provide valuable data to be used for early diagnostics and monitoring of ongoing health conditions. However, extracting useful information from such data is a challenge, particularly due to the differences in the context of each household, such as the room layout, location of sensors and activity patterns of inhabitants.
This talk will discuss how to make data mining methods context-aware, that is to take all available information about the context into account. In particular, we will concentrate on the machine learning methods which can deal with some changes in the deployment context compared to the training context, as identified in the REFRAME project (http://reframe-d2k.org). Examples include changes in misclassification costs, class distribution and feature distributions. We will discuss which of those methods could be useful for mining home sensor data in the SPHERE project (http://irc-sphere.ac.uk).
Seminar- Michel Valstar- Associate Professor, Faculty of Science, Nottingham University
21st January 2016
The Computational Face - Novel approaches in an age of big data
In this talk I will present recent advances in computer vision and machine learning made by my team at the University of Nottingham. I will let the audience choose which two out of four topics for me to present: 1). Behaviomedics - a novel area of using affective computing and social signal processing to help diagnose, monitor, and treat medical conditions that alter expressive behaviour, including recent work on automatic depression detection. 2). Facial Expression Analysis - our latest facial expression recognition research, including FERA 2015, our ICCV 2015 work on multi-task learning, and dynamic deep learning 3). Visual Object Tracking - TRIC-track: tracking by regression with incrementally learned cascades, as introduced in our ICCV 2015 paper, and tracking with multiple temporal scale motion models. 4). Facial Point Localisation/Face Alignment - discussing our ground-breaking work on direct-diplacement based point detection and recent advances thereof.
Seminar- Yang Hao- Professor in Antennas and Electromagnetics, Queen Mary, University of London
10th December 2015
Antennas and Propagation for Body-Centric, Wireless Communications: Current Status, Applications and Future
Body-centric wireless communications refer to human-self and human-to- human networking with the use of wearable and implantable wireless sensors. It is a subject area combining wireless body-area networks (WBANs), Wireless Sensor Networks (WSNs) and Wireless Personal Area Networks (WPANs). Body-centric wireless communications has abundant applications in personal healthcare, smart home, personal entertainment and identification systems, space exploration and military. This talk presents a review of some current work conducted at Queen Mary University of London, related to antennas and propagation for body-centric wireless communications. Aspects related to measurement setup, numerical modelling, channel characteristics are briefly discussed. Applications and future trend of this research will be also presented.
Seminar - Patrick Kierkegaard, Research Associate in Technology Applied to Health, University of Bristol
3rd December 2015
eHealth implementation and adoption – thinking beyond numbers, letters, commands, images and sounds
Healthcare is becoming more digital where governments around the world are prioritising the rapid implementation and adoption of eHealth technologies such as electronic health records, health information exchange systems, telehealth, etc. eHealth offers enormous benefits that can better support collaborative healthcare delivery services and continuity of care, however, healthcare is a complex concept that involves several complex intertwining factors that must be coordinated in order to successfully design and integrate an eHealth solution that meets the needs of patients and healthcare professionals. This includes physical, emotional, intellectual, social aspects, organisational, legal, behavioural, economic, and political factors. In this presentation, I will provide examples of my contributions to the eHealth research so far, provide an overview of methodology employed, and delineate my current activities and future work areas that involves the intersection between technology and health.
Michela Goffredo - BioLab3, University Roma TRE
2nd December 2015
Markerless Human Movement Analysis and Activity Monitoring @ BioLab3
In this seminar, an alternative approach to lab-based human motion analysis will be described, based on the experience of the Laboratory of Biomedical Engineering – BioLab3 (Department of Engineering of Roma Tre University) in developing methods and technologies for the analysis and processing of data in the context of human movement science.
The talk will introduce systems for the analysis of kinematics without external markers allocated on the subject's body, i.e. Markerless Human Motion Analysis. With specific reference to the clinical and rehabilitative fields, human motion analysis is often implemented by means of high accuracy marker-based stereophotogrammetric systems, which usually suffer from a number of drawbacks that prevent them from being fully disseminated in clinical and rehabilitation contexts: high costs; burden time for marker positioning; limitations on subjects’ movement and its naturalness; difficulties in individuals acceptability (e.g., children, elderly…). For these reasons, Markerless Human Motion Analysis aims at applying techniques derived from computer vision and image processing to track relevant elements from video streams. In this context, methods to analyse common functional tasks performed during daily living activities (e.g. sit-to-stand) and thus assessing the motor status of fragile people (e.g. elderly) with a remote system will be presented. Moreover, the recent launch of low cost RGB-D cameras facilitated the extraction of human kinematics and the development of motion-based applications. To this extent, the seminar will describe a gesture-based serious game for the kinaesthetic practice of phonological skills by preschool children, based on the research activity BioLab3 has done on the implementation of RGB-D-based systems for helping revealing and rehabilitating specific disabilities.
Furthermore, the talk will introduce methods and algorithms for the remote monitoring of activities of daily living in uncontrolled environments, based on wearable inertial sensors. Novel paradigms for the analysis and classification of signals from accelerometers and gyroscopes will be described, and innovative methods to automatically and robustly segment the different phases of an activity cycle will be presented.
Nicola Bellotto- Reader in Computer Science, University of Lincoln
24th November 2015
Making Sense of Human Motion: Estimation and Interpretation of Human Trajectories for Autonomous Systems Recent advances in autonomous systems research are bringing robots and smart environments closer and closer to our daily lives. In order to operate safely and provide reliable services, these systems must identify and understand human behaviours. In this talk I will present past and ongoing research in the area of people tracking for Human-Robot Spatial Interaction (HRSI) and Activity Recognition, discussing some of the tools and techniques used to estimate and interpret human motion trajectories. In particular, I will show some work done by our group on the use of a Qualitative Trajectory Calculus (QTC) to represent HRSIs, and discuss a recent extension of QTC with some preliminary results on human action classification. The presentation will conclude with an insight into current limitations and future research in this area.
Aristodemos Pnevmatikakis- Associate Professor, Athens Information Technology
19th November 2015
Infrastructure and algorithms for understanding user actions in e-caring homes
e-care at home is an alternative to traditional care provision in mild situations. The e-care solution is about having some care-giver tasks performed by systems understanding the actions of the care recipient. To do so, we build home sensing environments, featuring sensing devices that observe certain selection guidelines, and perceptual algorithms that extract metadata from the sensor signals. The collected metadata are processed by reasoning modules to understand the activities of the care recipient. Based on this understanding, notifications and alarms are dispatched to the care recipient and the care providers. The use case of eWALL is utilised to demonstrate the sensing, processing and reasoning modules of such an e-care system.
Seminar- Praminda Caleb-Solly, Associate Professor in Independent Living Systems, UWE
12th November 2015
Assistive service robots to support active ageing and independent living
Our objective in developing assistive robots is to support independent and healthy living, helping people manage and maintain a high quality of life. A robot as the interface has the potential to offer a more social and entertaining interaction experience. However, if assistive robots are going to be a part of potential solution, they need to be acceptable – what does this mean? Praminda will discuss the research being conducted at the Bristol Robotics Lab into people’s perceptions and expectations of assistive robots in a home environment, which is reframing our approach to considering robot embodiments and human-robot interaction.
Seminar- Malcolm Hart and Gemma Hargreaves, Bristol Community Health
29th October 2015
Supported self-care through technology A population based approach to long term condition management and prevention inthe community
Bristol Community Health (BCH) and Philips are working together to develop betterways of managing long term conditions in the community. The aim is to do this at scale, using a range of technologies but remaining people centred by creating a staffed clinical hub that patients and other health care professionals interact with. The approach is called supported self-care because it recognises that many people can self-care, provided they are appropriately supported. This can happen through technology, but always requires a human face. The approach is aligned with the strategic objectives of Bristol Clinical Commissioning Group. To build local experience, capacity and evidence for a supported self-care hub, BCH, Philips and the Lennard Practice in Bristol have come together in a project to implement Supported Self-care at a single practice. In total 93 patients were enrolled at three different levels of service, depending on disease acuity and personal preference. However, not all of them became regular users. Overall, 22 patients in the high risk cohort used intensive monitoring with installed equipment. 35 patients in the middle risk cohort were supported with text messaging service using their own phone and some provided peripherals to collect and reflect vital sign data. Finally, 17 patients in the low risk cohort registered and downloaded a self-monitoring App on their Smart Phone. Results and outcomes in terms of activity, healthcare use, patient reported outcomes, clinician observations and patient opinion are reported. A 50% reduction in admissions for the higher risk patient cohort was seen. Also practice activity (visits, phone calls etc) went down by about a third for all acuity levels. A 38% to 56% increase in patient activation level was seen, and 98% of patientswould recommend this to family and friends. Feedback from clinicians was positivewith all of those involved reporting positive changes in patients. Clinicians would favour a more integrated information flow. Overall, the project has been a success and this report captures many of the success stories. BCH and Philips are now working to identify appropriate funding source to take the Supported Self-care concept to scale in Bristol.
Seminar- Guang-Zhong Yang, Director, The Hamlyn Centre, Imperial College London
22nd October 2015
Body Sensor Networks- From Vital Sign Monitoring to Behaviour Profiling
With demographic changes associated with the aging population and increasing number of people living along, the requirement and practical provision of future healthcare delivery are changing rapidly. Shorter hospital stay and better community care is set to be the future trend of healthcare. Such schemes must be underpinned by an intelligent and pervasive information link between patients and specialist centres through wearable or implanted wireless sensory devices to provide early warnings of adverse events. The provision of “ubiquitous” and “pervasive” monitoring of physical, physiological, and biochemical parameters in any environment and without activity restriction and behaviour modification is the primary motivation of Body Sensor Network (BSN) research. This talk will highlight the key technical challenges, as well as the latest developments and practical examples of BSN, from vital sign monitoring to behaviour profiling, as well as how this may help to reshape the future of healthcare.
Seminar- Sebastian Bader, Senior Researcher, University of Rostock
15th October 2015
From Causal Models to Proactive Assistance
Computational State Space Models (CSSMs) provide a means to formally describe dynamic systems. In particular they can be used to describe and reason about the behaviour of human agents or groups of agents. In my talk, I will outline the idea of CSSMs and our implemented system. In addition, I will present ongoing research projects of the group of „Mobile Multimedia Information Systems“ at the Institute of Computer Science and of the department „Ageing of individuals and society“ within the Interdisciplinary Faculty of the University of Rostock, Germany. The talk will cover topics from proactive assistance in smart environments (like meeting rooms), to automatic recognition of mental disorders, reconstruction of activities of daily living from noisy sensors and assistive systems for people with dementia.
Seminar- Richard Bibb, Professor of Medical Applications of Design, Loughborough University
1st October 2015
Design and Additive Manufacture of wearable medical devices
Additive Manufacturing (AM), also known as 3D Printing offers increasingly affordable opportunities to produce custom fitting wearable medical and rehabilitation devices. The ability to produce accurate and complex forms is a distinct advantage. However, the capture of human anatomy shape and the design of wearable devices are not as easy as it first appears. This presentation will touch on some of the obstacles of designing AM wearable devices and illustrate various cases that have been successful or are close to implementation.
Seminar- Amos Storky- Reader in School of Informatics, University of Edinburgh
24th September 2015
Deep Learning: Applications in Computer Go and in Computing Fair Representations
I will introduce Deep Learning methods, and explore two applications involving novel Deep Learning developments.First we recognise, that mastering the game of Go has remained a long standing challenge to the field of AI. Modern computer Go systems rely on processing millions of possible future positions to play well, but intuitively a stronger and more 'humanlike' way to play the game would be to rely on pattern recognition abilities rather then brute force computation. Following this sentiment, we train deep convolutional neural networks to play Go by training them to predict the moves made by expert Go players. We tell the story of these developments and show how we have been able to beat the top computer Go players using this approach.Second we note that the representations learnt in deep networks are commonly opaque. Yet sometimes we require representations to satisfy certain properties. We show how adversarial deep learning can be used to this end, and demonstrate this on the idea of learning fair representations: representations that do not allow any discrimination with respect to some protected variable.This is joint work with Christopher Clark, Isaac Henrion, Martin Mueller, Andrew Jacobsen (Deep Go) and Harri Edwards (Fair Representations).
Seminar- Veronika Williams- Research Fellow, University of Oxford
17th September 2015
Patient experiences of using digital health application to support self-management in COPD: opportunities, challenges and (possible) solutions.
Seminar- Theo Tryfonas- Senior Lecturer in Systems Engineering, University of Bristol
10th September 2015
Application of a Game Theoretic Approach in Smart Sensor Data Trustworthiness Problems
In this work we present an Intrusion Detection (ID) and an Intrusion Prevention (IP) model for Wireless Sensor Networks (WSNs). The attacker’s goal is to compromise the deployment by causing nodes to report faulty sensory information. The defender, who is the WSN’s operator, aims to detect the presence of faulty sensor measurements (ID) and to subsequently recover compromised nodes (IP). In order to address the conflicting interests involved, we adopt a Game Theoretic approach that takes into consideration the strategies of both players and we attempt to identify the presence of Nash Equilibria in the two games. The results are then verified in two simulation contexts: Firstly, we evaluate the model in a middleware-based WSN which uses clustering over a bespoke network stack. Subsequently, we test the model in a simulated IPv6-based sensor deployment. According to the findings, the results of both simulation models confirm the results of the theoretic one.
Seminar- Francois Bremond- Research Director at INRIA Sophia Antipolis, Nice University
25th June 2015
Scene understanding for Activity Monitoring
Since the population of the older persons grows highly, the improvement of the quality of life of older persons at home is of a great importance. This can be achieved through the development of technologies for monitoring their activities at home. In this context, we propose activity monitoring approaches which aim at analysing older person behaviors by combining heterogeneous sensor data to recognize critical activities at home. In particular, this approach combines data provided by video cameras with data provided by environmental sensors attached to house furnishings. There are 3 categories of critical human activities:
Activities which can be well described or modeled by users
Activities which can be specified by users and that can be illustrated by positive/negative samples representative of the targeted activities
Rare activities which are unknown to the users and which can be defined only with respect to frequent activities requiring large datasetsIn this talk, we will then present several techniques for the detection of people and for the recognition of human activities using in particular 2D or 3D video cameras combined with other sensors. More specifically, there are 3 categories of algorithms to recognize human activities:Recognition engine using hand-crafted ontologies based on a priori knowledge (e.g. rules) predefined by users. This activity recognition engine is easily extendable and allows later integration of additional sensor information when needed [Robert 2012, Sacco 2013].Supervised learning methods based on positive/negative samples representative of the targeted activities which have to be specified by users. These methods are usually based on Bag-of-Words computing a large variety of spatio-temporal descriptors [Bilinski 2012, 2013].Unsupervised (fully automated) learned methods based on clustering of frequent activity patterns on large datasets which can generate/discover new activity models [Pusiol 2012].We will also discuss important issues related to Assisted Living and end-user benefits.We will illustrate the proposed activity monitoring approaches through several home care application datasets
Seminar- Eloisa Vargiu- Manager of the Integrated Continuous Care Line at EURECAT Health R&D&I, University of Cagliari
18th June 2015
Monitoring People that Need Assistance through a Sensor-based System: the BackHome Experience
People that need assistance, as for instance elderly or disabled people, may be affected by a decline in daily functioning that usually involves the reduction and discontinuity in daily routines and a worsening in the overall quality of life. Thus, there is the need to intelligent systems able to monitor indoor and outdoor activities of users to detect emergencies, recognize activities, send notifications, and provide a summary of all the relevant information. In this talk, a sensor-based telemonitoring system that addresses all that issues will be presented. Its goal is twofold: (i) helping and supporting people (e.g., elderly or disabled) at home; and (ii) giving a feedback to therapists, caregivers, and relatives about the evolution of the status, behavior and habits of each monitored user. The proposed system is part of the EU project BackHome and it is currently running in three end-user’s homes in Belfast. The overall experience in applying the system to monitor and assist people with severe disabilities will be illustrated and lessons learnt discussed.
Seminar- Chris Todd- Professor of Primary Care and Community Health, University of Manchester
11th June 2015
Using technology to predict, detect and prevent falls amongst older people: FARSEEING & ProFouND
9th June 2015
SPHERE will be showcasing some if it's technology at Venturefest on 9th June at the Engine Shed, Bristol. For more information see: http://venturefestbristolandbath.com/
3rd June 2015
The Director of SPHERE, Professor Ian Craddock talked about the SPHERE project at the prestigious Cheltenham Science Festival Summary of Talk: Internet of Things technologies — fitness wristbands and smart watches — aremoving health from the hospital to the home. But if your watch, thermostat and games console couldmanage your well-being, how would you feel about being constantlymonitored? Engineer Ian Craddock and social scientist Madeleine Murtagh delve into the technology and the ethics, and ask if this is a future of health we can live with.
Seminar- Nadia Berthouze- Professor in Addictive Computing and Interaction, University College London
28th May 2015
Emo&Pain project: Facilitating physical activity in Chronic Low Back Pain
Can interactive technology support people with chronic pain to remain physically active? In this talk I will first present results from a set of qualitative studies with: (1) people with chronic pain to understand what strategies they use to carry out physical activity in daily life and what factors deter them; and (2) physiotherapists to understand how they support them. I will then present two design studies that build on this understanding. The first study demonstrates how auditory feedback could be used to address some of the psychological barriers and needs identified and lead to an increase in self-efficacy, motivation and confidence in physical activity. The second study features our system to automatically recognize people's pain level and pain-related affective states during physical activity.
Seminar- Graham Stuart- Consultant Cardiologist, Bristol Royal hospital for Children
Exercise and teenagers with congenital heart disease: a gamification approach.
There is increasing evidence that regular exercise is beneficial for teenagers with heart disease. However, most teenagers with congenital heart disease are relatively sedentary and few teenagers receive a formal exercise prescription. I will present a pilot study which uses a gamification approach to exercise prescription. This uses a commercially available wrist worn accelerometer (NIKEBand) which links to a smartphone and social media. I will also present some data on the in-house development of a smartphone based app designed to encourage patient interaction with healthcare professionals and which we hope will eventually link with this gamification approach.
Seminar- Mark Hawley- Professor of Health Services Research, University of Sheffield
30th April 2015
Activity monitoring in the home for health and well-being: the Sheffield experience
This talk will describe the work of the University of Sheffield’s Centre for Assistive Technology and Connected Healthcare in researching,developing and evaluating technology to help people live and age well. – and will illustrate this work with two projects linking activity and health in long-term conditions (LTCs). For many LTCs, greater physical activity is known to be protective against deterioration, and increasing physical activity is an element of rehabilitation. However, following what is usually a short period of rehabilitation, people are left to self-manage their condition and they frequently reduce their physical activity, potentially leading to a downward spiral. The key is to change people’s behaviour over the longer term and a smart-phone app, which aims to encourage people to self-manage their physical activity in the period following pulmonary rehabilitation for Chronic Obstructive Pulmonary Disease (COPD), will be described. The activity that people with LTCs carry out is affected by their health. Activity monitoring systems in the home, often known as lifestyle monitoring systems (LMS), are designed to detect changes in activity that may be indicative of changes in health state. However, most systems simply attempt to detect change in activity and do not make the link with health explicitly – relying on carers to interpret whether changes in activity are significant. The results of a mixed methods study aiming to explore the link between activity in the home, as measured by LMS, and health state, will be reported.
Seminar- Dimitrios Makris- Associate Professor, Kingston University London
23rd April 2015
Action Recognition and Pose Tracking using Non-Linear Dimensionality Reduction Methods
This seminar presents novel dimensionality reduction methods based on Laplacian Eigenmaps and their application on human motion analysis. Firstly, action recognition is tackled by Temporal Laplacian Eigenmaps (TLE) which produces elegant low dimensional manifolds that capture the nature of specific action while at the same suppress stylistic variations. Secondly, human pose tracking for a specific action is achieved by Generalised Laplacian Eigenmaps (TLE) and the Graph-Based Particle Filter (GBPF). GLE produces graph-based manifolds where both the temporal and stylistic variations are coherently modelled. GBPF exploits the structure of GLE to optimise tracking in the low dimensional manifold. State-of-the-Art results are shown to demonstrate the value of the proposed methods in action recognition and pose tracking.
Faculty of Engineering Research Showcase for Industry Partners
Monday 13th April 2015
University of Bristol, Merchant Venturers Building • Sector summaries from leading academics • Sector-based research showcase • Prize-winning student displays • Networking reception
Seminar- Rachel King- Research Assistant, University of Reading
26th March 2015
Observing Frequent Fallers at Home (Part 2): Complex data in complex environments
This seminar is part of a series, presenting the initial findings from a small ethnographic study conducted by a collaboration of SPHERE researchers from the Universities of Southampton, Reading and Bristol. This study focuses on the observation of frequent fallers in the home to broaden our understanding of falls and to guide the development of tools for the management and self-management of fallers using sensor systems. Five people with moderate to severe Parkinson’s Disease were recruited, and over the course of six visits were observed in their own homes. During four of these visits it was aimed to record their activity and functionality using video, Kinect, and body worn sensors. During this seminar the body worn sensors developed for this work will be described and the techniques used to align the three separate sensing modalities outlined, including the annotation of data. A summary of the body worn sensor data recorded will be presented and the challenges involved in collecting sensor data in the ‘real world’ will be discussed. The initial observations from the body worn sensor data will then be presented and how this work may guide further research.
Seminar- Steve Kitson- Folium Optics and Andy Van Heusen Pumpco
19th March 2015
MY HEALTH TAGS: CLOUD CONNECTED TAGS AS MEDICINE REMINDERS
The World Health Organisation estimates that up to half of prescribed and dispensed medicines are not used as recommended. For the NHS that represents a cost of up to £300m pa. Non-adherence and underuse represents a loss to patients, the healthcare system and society at large. To help improve adherence, Folium Optics and Pumpco have been working on an internet of things solution - smart tags that incorporate low power plastic displays and that are connected via the cloud to reminder apps and devices. The tags show patients when they are due to take their medicines, and log when they have been taken. In this seminar we will review the project and describe the technology that makes it possible.
Seminar- Bo Tan and Evgeny Tsimbalo- Research Assistants, University of Bristol
5th March 2015
Energy efficiency is a key attribute of many modern wireless communication systems. For instance, in healthcare applications, it's vital for a wireless sensor to consume as little energy as possible while transmitting data to a central access point to prevent battery drain. Various energy-efficient communication standards have been developed recently, including Bluetooth Low Energy, WiFi Low Energy and ZigBee. Typically, energy-efficiency is achieved by stripping off all unnecessary components, leaving out only basic communication blocks. As a result, the reliability of such systems is limited, and wireless transmission is only possible at very short distances. In the first part of the talk, it will be shown how to increase the reliability without compromising the energy efficiency of the transmitter. The Bluetooth Low Energy standard will be used as an example.
The radio frequency signal can be designed for diverse purposes: communications, timing, target detection, power transfer. Wireless communications signals for example mobile, FM, WiFi, DAB etc, are universally exist and easily accessible RF signal nowadays for information distribution or exchange in outdoor and indoor. In this study these wireless signals are investigated for applications beyond communications, such as target discovering, movement capturing, timing and positioning. The results and examples of using 802.11x signal for indoor scenario will be introduced as examples. The feasibility of using BLE signal for similar purposes will be discussed at the end.
Seminar- Emma Stack- Research Associate- University of Southampton
26th February 2015
Observing Frequent Fallers at Home
This seminar presents initial findings from a small ethnographic study conducted by a collaboration of SPHERE researchers from the Universities of Southampton, Reading and Bristol. Sensors at home could tackle many outstanding issues in a) our current understanding and prevention of falls and b) the management (and self-management) of fallers. Lessons learned from our research could apply to other events that, like falls, happen predominantly unwitnessed at home. We aimed to observe people with marked postural instability functioning in the ‘real world’ setting of home. We gathered insight into how they live that will focus further studies and help ensure we maximise the benefits of sensor technology while minimising risks. In five case studies, we recruited people with moderate or severe Parkinson’s Disease. We visited each one at home on up to six occasions, observing how they functioned and attempting to record their activity using video, Kinect and body-worn sensors. From our observations of spontaneous movement and the activities that people associated with previous falls, near-misses or fear of falling, we drew the following conclusions.
Firstly, although physical obstructions present a challenge to cameras, people’s interaction with obstructions is highly informative.
Secondly, people do not walk in extended straight lines at home: negotiating direction changes, open spaces and changes in height (associated with transfers and steps/stairs) are frequent challenges.
Thirdly, rather than monitor mobility throughout the home, focusing on the habitual route ‘from chair to stair’ would illustrate the activities of greatest interest, yielding parameters that sensors could learn to measure.
Seminar- Guy Lever- Research Associate, University College London
19th February 2015
Modelling transition dynamics in MDPs with reproducing kernel Hilbert space embeddings
I will present a method of model-based reinforcement learning which represents stochastic transition dynamics in Markov decision problems non-parametrically as "embeddings" in a reproducing kernel Hilbert space. The approach combines well with dynamic programming methods such as value iteration: for example, iterating Bellman maps converges and guarantees on the value of the learned policy can be derived. I will discuss some recent extensions of the method to handle online data acquisition and practical speedups using greedy feature selection, and present experiments on simulated control tasks.
Seminar- John Tarlton, Reader in Regenerative Medicine, University of Bristol
12th February 2015
In 2012 laying hen battery cage systems were banned in the EU. Whilst there are undoubtedly clear benefits in terms of hen welfare, the use of more extensive systems come with their own problems. Hens used in battery cages have been bred for many decades to maximise egg production. The ancestral hen, the Red Jungle Foul, produces perhaps 20 eggs per year, and the modern laying hen produces over 300 eggs per year. The massive production of egg-shell has severe skeletal consequences which were not necessarily apparent in the hazard-free cage environment. Transfer of hens to free range and aviary systems has resulted in a huge increase in the prevalence of keel bone fractures, up to 96% in some systems, mainly as a consequence of interacting with hazardous structures within a complex environment. Although we have shown that different systems come with widely divergent risks, and we know what certain elements, such as suspended perches, are hazardous to the keel bone, our lack of knowledge of how the hen moves around its environment, or the nature of its interactions with particular elements, means we are unable to apply evidence based design for hen-friendly systems. We have developed 3D-accelerometers capable of measuring specific behaviours and potentially damaging collisions. However, using current technology such as RFID, coded infra-red LEDs and magneto-inductive tracking, we are, as yet, unable to track hens sufficiently accurately around their complex environment to determine how these are related to defined housing elements. This seminar will discuss the problem and its partial solutions, and raise the question as to how we can progress further.
Seminar- Peter Veltink- Professor of Technology, University of Twente
5th February 2015
Assessing daily-life motor performance using on-body sensing
People with motor impairments due to central neurological disorders are clinically treated and trained with the objective to improve their motor function. It is currently unclear how this impacts daily-life functioning of these people. Quantitative monitoring of body movements during daily-life can contribute to assessing this impact. We investigate the assessment of daily-life motor performance using on-body sensing in two of our current projects: the EU project INTERACTION and the PowerSensor project, funded by the Dutch Technological Science Foundation.
INTERACTION Stroke often results in impaired motor control, affecting functional performance of both upper and lower extremity. During post-stroke rehabilitation, motor functions, including reaching, grasping and mobility, are trained in order to prepare the stroke survivors for their return to daily-life. The objective of training is to optimize daily-life functional motor performance, but only functional motor capacity is evaluated during rehabilitation using clinical tests. In fact, it is unclear how the stroke survivors perform during daily-life and how this is related to the rehabilitation training they received. It is the objective of the EU project INTERACTION (FP7, project number 287351) to monitor motor performance of stroke survivors during daily-life and to make this information accessible to the clinical professional on a distance for guiding individual stroke subjects to optimize their daily-life performance, advise them concerning continued training and to evaluate and optimize the impact of rehabilitation programs. A modular textile-integrated sensing system was developed and performance and capacity measures were proposed and clinically tested in stroke subjects. Telemonitoring facilities were developed and tested. The system is currently tested in stroke subjects during their daily-lives.
Environmental interaction with our hands is important in daily-life and may be affected in people with neurological disorders. Analysis of this interaction requires sensing of
hand and finger movements, interaction forces, and quantification of the dynamic interaction with the environment.
We have developed an inertial and magnetic sensing system for hand and finger kinematics and a 6DoF micromachined fingertip force/moment sensor, and demonstrated methods for assessing hand and finger kinematics and dynamic interaction with the environment. The hand and finger kinematic sensing system is currently applied in assessing hand and finger functions in people with Parkinson’s disease and in the elderly.
Seminar- Mischa Dohler- Chair Professor of wireless Communications, Kings College London
29th January 2015
Will 5G enable the Internet of Things?
Based on my entrepreneurial experience with Worldsensing as well as academic research over past years, I will review the possibility of a cellular system to enable the connectivity needed for the emerging Internet of Things (IoT). I will discuss the tradeoff between different connectivity technologies, such as low-power short-range, low-power wide-area and cellular systems. I will discuss why the low-power approach was wrong all along, and why systems like Zigbee will shortly be discontinued. I will then review the work we do at King’s on machine-to-machine (M2M) in 5G, involving novel approaches to the random access channel as well as the decoupling of up and downlinks (which recently won us the Best Paper Award at Globecom 2014).
Seminar- Grigoris Antoniou- Professor of Semantic & Knowledge Technologies, University of Huddersfield
22nd January 2015
Semantic and Knowledge Technologies for Smart Environments
Emerging complex smart environments are characterized by the presence of a number of sensors and a variety of computing and interactive devices; typically various activities by actors (be it human, digital or robotic) take place in such environments. While the scope for mobile computing/networking and machine learning/mining is clear, in this talk we will argue that semantic and knowledge technologies have also a crucial role to play for smart environments to reach a high level of sophistication. We will review three relevant lines of research. The first is concerned with a rule-based approach to activity recognition. Key features of this approach include its ability to recognize activites which are hierarchically organised, and its adaptability to the desired granularity (level of abstraction). The approach was tested in a real ambient intelligence environment, and was demonstrated to be robust in the presence of noise. Then we will discuss the application of reasoning for processing streams of (sensor) data. In addition we willshow how reasoning can be used to combine sensor data with other information sources; it is this linkability and semantic interoperability that is a key benefit of using semantic technologies. Finally we will present a distributed approach to reasoning about context in smart environments, capable of realing with deficient (conflicting, missing) information. A proof-of-concept implementation was developed for handheld devices and was demonstrated in a real ambient intelligence environment.
Jonathan Chevallier- Oxehealth
15th January 2015
Oxehealth is a spin-out from Professor Lionel Tarassenko’s labs within the Institute of Biomedical Engineering at the University of Oxford. The business is focused on commercialising technology which enables digital video cameras to monitor health. The presentation will explain at a high level the techniques that are being used, provide details of some of the successful studies that have been performed and outline the work currently underway and planned to further develop the technology. The advantages of camera based monitoring when compared to other devices will also be explained.
Seminar- Rod Hose- Professor of Computational Biomechanics, University of Sheffield
4th December 2014
The Virtual Physiological Human initiative is building patient-specific models to simulate physiological processes in health and disease, with the aim of providing diagnostic and decision support tools to the clinical community. The exquisite resolution of modern imaging devices supports the development of excellent and detailed three-dimensional anatomical models. A great challenge for us is to integrate with and to interpret the wealth of clinical data, at both individual and population levels, to inform our models. We have a concept of a patient avatar that develops over the lifetime of the individual, using all of the measurements that are made – whether in the clinic or elsewhere, to become an increasingly accurate digital representation of the individual. This avatar will be subject to an array of computational tools, running continuously or on demand, to provide early and timely diagnosis of changes in physiological state and of the consequences. It will also serve as the basis of poredicitve studies to assess the likely effects of candidate interventions. We have put a lot of effort into developing a computational infrastructure to support this activity. I will discuss a range of clinical applications, one or two in detail, and will provide an overview of the infrastructure developed by VPH-Share to serve the community.
Seminar- Karl Woodbridge- Department of Electronic and Electrical Engineering, UCL
4th December 2014
Human Movement Tracking and Activity
Detection using Passive Wireless UCL has been working for some years on Passive Wireless Movement Detection (PWMD) technology, which harnesses existing wireless network or mobile device signals to detect and classify human movements. PMWD does not require a dedicated transmitter or a tagged or transmitting target and can therefore be used unobtrusively with minimal impact on monitored subjects. Aided by the rapid roll out of wireless communications PWMD technology has numerous applications that include security surveillance, human-machine interaction, indoor positioning and healthcare monitoring.
In this presentation the UCL PWMD prototype system will be introduced and an overview given of the basic principles, software defined radio hardware and signal processing methods developed. The high speed Doppler processing developed enables real time display of a wide range of motions and activities. Results will be presented for in-room and through-wall activity detection ranging from large to very small body movements. Initial results on signs of life detection and indoor positioning and tracking will also be shown.
The system has many synergies with the activities within SPHERE and the potential for integration of PWMD
Seminar- Gregory O'Hare- Director Ucd Earth Institute, University College Dublin
Ubiquitous Intelligent Systems: Scalable and Sustainable Capture, Interpretation and Acting upon Diverse Harvested Data
We reside in a world that is becoming ubiquitously sensed, delivering vast volumes of data, which is often noisy, lossy and contradictory across a myriad of real-time data streams. This talk considers the inherent challenges of capturing, routing, interpreting, making inferences upon and delivering feedback based upon such vast volumes of data. It will consider how to deliver the necessary inherent intelligence and the middleware infrastructure needed to sustain continued operation. This talk will draw upon a number of discrete application areas.
Seminar- Timothy Hospedales- Senior Lecturer- QMUL
19th November 2014
Weakly Supervised Learning with Topic Models for Computer Vision
The performance bottleneck in many computer vision applications today is the volume of strongly annotated training data. Strongly annotated data such as object locations, segmentations or tracks may be prohibitively expensive to obtain at scale. In contrast, coarse annotations such as image or video level tags may be more practical to obtain cheaply, however their relation to media content is ambiguous, making learning more challenging. In this talk I will give an overview of a line of work in our lab addressing such weakly-supervised learning problems using generative topic models. In the process I will demonstrate how topic models can be applied to a variety of computer vision applications including behaviour modelling in surveillance, detection, recognition, segmentation and attribute description.
Seminar- Mo Haghighi talking at Pervasive Media Studio
14th November 2014
Researcher Mo Haghighi from Bristol University will introduce us to multidisciplinary smart house project SPHERE
The UK is the most obese nation in Europe. Our ageing population is especially at risk from isolation, depression, strokes and fractures caused by falls in the home. With a rapidly ageing population - could technology be the answer to some of these problems?
SPHERE is developing a number of different sensors that will combine to build a picture of how we live in our homes. This information can then be used to spot issues that might indicate a medical or well-being issue.
The system will be general-purpose, low-cost and accessible. Sensors will be entirely passive, requiring no action by the user and suitable for all patients, including the most vulnerable. An example of SPHERE’s home sensor system could be to detect an overnight stroke or mini-stroke on waking, by detecting small changes in behaviour, expression and gait. It could also monitor a patient’s compliance with their prescribed drugs.
SPHERE will work with clinicians, engineers, designers and social care professionals as well as members of the public to develop these sensor technologies. ??We want to make sure the technology is:
Acceptable in people's homes
Solves real healthcare problems in a cost effective way
The project generates knowledge that will change clinical practice
SPHERE, with a total of £15m grant, is an interdisciplinary research collaboration (IRC) led by the University of Bristol will work in partnership with Bristol City Council, IBM, Toshiba and Knowle West Media Centre (KWMC).
Seminar - Dr Tom Diethe, Research Assistant, University of Bristol
13th November 2014
Sparsity and Random Projections
Sparsity in its many forms plays an important role in signal processing, computational statistics, and machine learning. Methods based on random projections are particularly relevant in situations where the data is sparse in some domain, and have also found applications in both fields. In this seminar I will give an overview of some of the different types of sparsity with some motivating examples, and I will describe examples of two classes of algorithms - greedy methods and sparse regularisation - that lead to sparse solutions. I will then go on to describe the random projection method, and show how this is linked to sparsity, particularly in the relatively new but now well established field of compressed sensing. During the talk, I will highlight some of my research in this field that takes advantage of these methods.
Designability 46th Annual Lecture Data Scientist – heal thyself 7th November 2014
Internet of Things technologies have emerged in recent years as enabling technologies for life-long healthcare. Increasingly these technologies are seen as moving health from the hospital to the home and at the same time shifting the responsibility for “health” from the Clinician to Computer Scientists and Engineers.
This could be an exciting vision of the ultimate system for personalised medicine with the individual at its focus. However, it could also be seen as undermining the relationship between clinician and patient that has delivered such a transformation in our lives over the past few centuries.
Do Engineers and Scientists have the skills to deliver and maintain these internet-enabled healthcare systems and do we think they can be trusted?
Professor Craddock will present his views on this exciting but unpredictable and rapidly-moving debate.
Seminar- Roisin McNaney and Dr Thomas Ploetz, Open Lab, Newcastle University
4th November 2014
Roisin McNaney Everyday technologies such as gaming systems, online crowdsourcing platforms and smartphones- and even future wearable devices such as Google Glass- are increasingly being used to enable the provision of non stigmatising, health based digital interventions. In this talk I will discuss my ongoing engagement with People with Parkinson’s to explore the design and deployment of a range of digital technologies to aid in symptom self-monitoring and management. I will reflect upon a variety of HCI methods we have been using, and are planning to use, to support concepts of self-care so important for this particular group, for whom feelings of lost independence and confidence are commonplace.
Dr Thomas Ploetz Traditional activity recognition aims at automatically identifying what humans do and when. With the advancement of both sensing and analysis technologies more thorough analysis of activities and, ultimately, behaviours has now become within reach of automatic approaches. In this talk I will motivate automatic analysis of the quality of human activities through a number of application cases and present novel analysis techniques that go beyond traditional activity recognition. These techniques contribute to the emerging field of Computational Behaviour Analysis (CBA). I will reflect on skill assessment in specific CBA domains and discuss future directions of quality and skill assessment based on more generic modelling and analysis approaches.
Seminar- Alberto Martínez Cantera, TECNALIA RESEARCH & INNOVATION
30th October 2014
TECNALIA carries out health-related applied research in order to improve health and quality of life for citizens with an special focus on elderly people and people with disabilities. Our international and multi-disciplinary team is made up by 100 researchers specialized in the fields of health-oriented biotechnology, biomaterials, robotics and ICTs; this allows us to provide solutions to many health challenges from different perspectives such as health-care, food or medical devices, among others.One of our areas of specialization is devoted to Human Behaviour and Activity Monitoring taking profit from Smart-Home environments and Ambient Assisted Living technology. The goal behind is to infer early symptoms of potential injuries and pathologies that might support physicians on diagnostic decisions, prevention and other sorts of smart assistance. It’s done on the basis of ethical criteria, user-centric methodology, minimally invasive sensor networks and behaviour and activity pattern modelling and evolution analysis
Semianr- Dr Maurizio Filippone, University of Glasgow
16th October 2014
Pseudo-Marginal Bayesian Inference for Gaussian Processes
Statistical models where parameters have a hierarchical structure are commonly employed to flexibly model complex phenomena and to gain some insight into the functioning of the system under study. Carrying out exact parameter inference for such models, which is key to achieve a sound quantification of uncertainty in parameter estimates and predictions, usually poses a number of computational challenges. In this talk, I will focus on Markov chain Monte Carlo (MCMC) based inference for hierarchical models involving Gaussian Process (GP) priors and non-Gaussian likelihood functions. After discussing why MCMC is the only way to infer parameters "exactly" in general GP models and pointing out the challenges in doing so, I will present a practical and efficient alternative to popular MCMC reparameterization techniques based on the so called Pseudo-Marginal MCMC approach. In particular, the Pseudo-Marginal MCMC approach yields samples from the exact posterior distribution over GP covariance parameters, but only requires an unbiased estimate of the analytically intractable marginal likelihood. Finally, I will present ways to construct unbiased estimates of the marginal likelihood in GP models, and conclude the talk by presenting results on several benchmark data and on a multi-class multiple-kernel classification problem with neuroimaging data.
Dr Joseph Wherton, ATHENE research Team, Queen Mary University,
London9th October 2014
ATHENE - Assistive Technologies for Healthy Living in Elders: Needs Assessment by Ethnography
An aging population is fuelling interest in assisted living technologies (ALTs), such as ‘telecare’ and ‘telehealth’, to support ‘ageing in place’ – that is, to enable older people to live independently at home, avoid or defer institutional care in later life and remain active participants in society. But while innovation is important, there is a well-documented gap between the development of new technologies and the consistent use of these technologies in practice. Our work in the ATHENE project demonstrates that if ALTs are to be fit-for-purpose, the ways in which they are designed and deployed must be grounded in care recipients’ experiences of ageing. More than that, however, it suggests that the technology industry, health and social care providers need to rethink how ageing in place is technically, organisationally and socially configured to enable the involvement of care recipients and their networks of carers in the ‘co-production’ of ALT solutions over time.
Southampton University, Work Package 4
2nd October 2014
Energy harvesting techniques and devices for wearable sensors
This seminar will provide and update on the progress made at Southampton in work package 4 of the SPHERE project. Researchers will present details of the technical principles, materials and fabrication of the following range of energy harvesting devices:* Inductive wireless power transfer – whilst this is an established technology, implementing it in textiles for wearable application presents challenges for both fabrication and operation.* Harvesting mechanical energy from human motion – flexible active materials that convert motion into electricity.* Photovoltaic textiles – integrating PV technology in or on textile substrates.* Energy harvesting hip implant – a self contained miniature mechanical energy harvester that can be embedded in a replacement hip joint.
Celebrating Age Festival Launch Event at M Shed Saturday 2
7th September 2014
Age UK Bristol are organising the Festival this year which will allow for the sharing of information and opportunities and access a wide range of services that can help make life just that little bit easier. This years festival theme is Bristol Ageing Better and will include the Achievement of Older People Awards and an Art and Craft Makers Competition. The event is free, fun and promises to be a really excellent day for everyone. There will be an opportunity to see some of the SPHERE technology in action and meet some of the team. Come along and say hello!
Bristol Bright Night
Friday 26 September 2014
Bristol’s streets come to life as Bristol Bright Night sweeps through the city for the first time with a programme of free activities, giving public audiences the chance to discover the wealth of scientific research taking place on their doorstep, and to meet the researchers responsible. From early afternoon into the evening Bristol's Harbourside will host interactive activities including exhibitions at At-Bristol, talks in the watershed and one-of-a-kind science-inspire street performances. There will be an opportunity to see some of the SPHERE technology in action and meet some of the team. Come along and say hello!
Dr Hammadi Nait-Charif, Senior lecturer, University of Bournemouth
25th September 2014
Indoor Monitoring and Fall Detection Using Ceiling Mounted Cameras
This talk will discuss indoor monitoring in a supportive home environment for elderly peopleliving alone. A computer vision based solution to detect important and rare events such as falls will be presented. As older peoples’ homes are often highly cluttered with belongings and furniture brought, the position and orientation of the cameras was chosen to minimise occlusion of the occupant. The monitoring setup uses ceiling mounted, standard cameras with vertically oriented optical axes, fitted with wide angle lenses. A combination of background subtraction and particle filters is used to track the occupant for an extended time. Then a context specific spatial model in terms of semantic regions, specifically inactivity zones and entry zones, is learned. Maximium a posteriori estimation of Gaussian mixtures is used in conjunction with minumum description length for selection of the number of mixture components. Learning is performed using expectation maximisation algorithms to maximise penalised likelihood functions that incorporate prior knowledge of the size and shape of the semantic regions. The resulting contextual model enables human readable summaries of activity to be produced and unusual inactivity such falls to be detected. SPHERE House Brainstorming Event SPHERE has acquired a research house in Bristol in which we will be installing the first prototypes of the newly developed SPHERE sensors .The housewill be an experimental facility, fully furnished and ready to serve as a highly-instrumented living lab. We plan to recruit a variety of healthy participants and people with health conditions to live in the housefor periods of time, in order to investigate hypotheses from clinical science and learn how long-term recorded sensor data can be used to monitor quality of life, well-being and various health conditions.We would like to invite health researchers as well as health professionals and other partners to define the experiments that will take place in this unique new facility over the coming year.
Seminar- Dr Vincent Van Hees- Move Lab, University of Newcastle
4th Sept 2014
How can we use wearable accelerometers to say something about a person’s physical activity and sleep behaviour?Wearable accelerometers are increasingly used for the assessment of physical activity and sleep both for research and non-research purposes. Typically, members of the public or specific patient groups are asked to wear the accelerometer for an entire week on their wrist. Over the last six years I have been trying to come up with methods to analyse these data and to help researchers with the implementation and interpretation of those methods. In this talk I will give an overview of the methods I have developed, including: sensor auto-calibration, monitor non-wear detection, energy expenditure estimation, and activity type (sleep) classification. Further, I will show how I am trying to make my methods available to the research community via R-package GGIR.About me:I studied Human kinetic technology in The Hague and Human movement sciences in Amsterdam, after which I moved to Cambridge to do a PhD in Epidemiology focussed on the implementation of raw accelerometry in population research. At the moment I am currently doing a post-doc in Newcastle upon Tyne in which I am continuing this work, but now with a stronger focus on intervention studies and the assessment of physical activity and sleep in patient groups.
Seminar- Dr Leah Aver-, Move Lab, University of Newcastle
4th September 2014
Identification of active ingredients to increase physical activity behaviour of adults with Type 2 Diabetes in Primary Care: Development of a behaviour change intervention.
Laeah is a post-doctoral research associate and registered practitioner health psychologist from Newcastle University. She is interested in the development and evaluation of behaviour change interventions and ways in which these interventions can be successfully integrated within routine clinical settings and the lifestyles of specific population groups. Leah’s Doctoral work involved the development of a multifaceted behaviour change intervention Movement as Medicine for Type 2 Diabetes that targeted consultation behaviour of primary healthcare professionals and physical activity behaviour of adults with Type 2 diabetes. Her talk will describe the development process of this intervention, present some findings from the open pilot phase of the work and highlight some important considerations for the development and optimisation of future behavioural interventions.
SPHERE at Bristol Bright Night
4th Sep 2014
Bristol Bright Night is the regions contribution to European Researchers Night - an event that showcases all the most exciting research taking place around Europe at the moment. There will be an opportunity to see some of the SPHERE technology in action and meet some of the team. Alison Burrows will also be presenting a talk on the project and her role in User Centred Design. Come along and say hello!
Seminar - Ruth White - University of Reading and Rubén Fernández Beltrán - University Jaume I
31st July 2014
Topic Modelling for Video Retrieval and Daily Routine Discovery
Topic models were originally developed by the text processing community as a computational tool for organising and searching large collections of documents. The algorithms enable the themes, or topics, that run through an unstructured collection to be automatically discovered. This seminar will explain what topic models are and how they work. In addition it will explore how they can be used for video retrieval and daily routine discovery, in relation to the SPHERE project.
Seminar: Professor Lynne Baillie, Director of the Interactive and Trustworthy Technologies Group, Glasgow Caledonian University
26th June 2014
Envisaging the future of Rehabilitation Technology
Multimodal Interaction encompasses several areas of research e.g. tangible, mobile and social interactions. Applications and systems in the future will be built mainly for mobile and ubiquitous settings it is therefore imperative that researchers lead the way in developing and investigating: novel interaction modalities, novel genres, and novel interface solutions for particular user groups, such as elderly users. The talk will present two of the work packages from the award winning £1.5 million Envisage project. The two work packages focused on designing, developing and then trialling, in the home, novel wireless motion tracking rehabilitation technologies for patients recovering from a fall or after knee replacement surgery. The project was praised for its clinical utility and value to older adults, its innovative concepts and its impact in the field of health and wellbeing.
Seminar: Professor Lars Sundstrom, Director of Enterprise and Translation for WE-AHSN
19th June 2014
West of England Academic Health Science Network: What is it and who cares? S
PHERE at The Festival of Nature
13-15th June 2014
The Festival of Nature unleashes the city’s wild side once again this summer, taking over the Harbourside on the weekend of 14 - 15 June. Featuring over 100 organisations, the Festival of Nature Wild Weekend gives wildlife-lovers of all ages a unique opportunity to discover and enjoy the natural world in the heart of the city. SPHERE will be at the event in the University of Bristol tent, giving visitors the chance to have a go with some of the SPHERE technology. Come and see us and let us know what you think!
Seminar- Marcus Munafo, Professor of Biological Psychology, University of Bristol
12th Jun 2014
Physical and mental health applications in SPHERE
Health behaviour (e.g., smoking, diet) and mental health are major contributors to the burden of disease in developed countries such as the UK. Interventions to promote healthy behaviours and improve mental health exist, but they are typically delivered only when the individual seeks them out. SPHERE provides a unique opportunity to test the feasibility of monitoring behaviour in the home and delivering interventions proactively. I will discuss three potential applications, relating to smoking, obesity and depression.
Seminar- Dr Andy Skinner- Lecturer- university of Bristol
5th June 2014
Capturing Data for the IEU
The MRC Integrative Epidemiology Unit (IEU) at the University of Bristol comprises six programmes exploring causal analysis methods and health behaviours, supported by a number of cross-cutting activities and scientific themes. The role of the Data Capture theme is to develop innovative tools and techniques for measuring the wide range of variables of interest to the IEU, from molecules to behaviours. In this talk I’ll introduce the Data Capture theme in more detail, and discuss a number of the projects we’re already leading and supporting across the IEU, many of which will be of potential interest to the SPHERE community.