|
|
| (One intermediate revision by the same user not shown) |
| Line 1: |
Line 1: |
| Project Description:
| | Client: José Alcalá, [[Informetis]] <jose.alcala@informetis.com> |
|
| |
|
| As the elderly population is dramatically increasing in the UK as well as in most other countries,
| | A rapidly aging population means assisted living is fast becoming a major societal challenge. To support “Carers” assisting “Carees”, local technology company Informetis provides smart sensors installed in the fuse-box that determine if household appliances are ON/OFF. As routines are typically closely related to appliance use, this proxies for inhabitants' wellness. Your task is to create an app for carers to support householders when they struggle to perform daily routines. You may choose focus on detection of activities from the raw data, or the interaction design challenges of choosing what should be communicated to the carer and when. |
| assisted living is becoming one of society’s biggest challenges. Furthermore, the lifestyle of elderly | |
| people has changed significantly. Nowadays, elderly people prefer to live an independent life in their
| |
| own home rather than moving to a residential care facility; they are more familiar using smart
| |
| devices than before; and their loved ones usually live some distance away to visit them frequently.
| |
| | |
| Therefore, there is an urgent need on modernising social care services in order to respond to this
| |
| increasing demand and to adapt to the modern lifestyle.
| |
| | |
| Social Care companies are aware of this issue and they are seeking for new innovative solutions
| |
| using cutting-edge technology. The mainstream solution is to use Internet of Things sensors (IoT) to
| |
| capture “Activity of Daily Livings” (ADLs) inside the house. Then, a caree’s routine is inferred using
| |
| Artificial Intelligence and provided to a carer. Using IoT solutions, elderly people can be monitored
| |
| remotely on a daily basis in their own homes. However, deploying sensors inside a house is a very
| |
| intrusive and expensive way of monitoring people, which can cause rejection and therefore delays in
| |
| its adoption.
| |
| | |
| Informetis has devised a completely non-intrusive technique to monitor elderly people. Using a
| |
| single Smart Sensor installed in the fuse box, we are able to tell when the appliances in the house
| |
| are being turned ON and OFF. As our routine is very much related to the use of electrical appliances, | |
| this can be used as a proxy for the wellness of inhabitants, providing invaluable information to | |
| Carers.
| |
| | |
| Your task is to design and implement a prototype of an assisted living application (a web-based or | |
| native application for mobile devices) that can be used for Carers as a toolkit for monitoring Carees’
| |
| daily routine. The application is designed for Carers, not for Carees.
| |
| | |
| We would like you to consider
| |
| three different machine learning parts:
| |
| | |
| • “Activity Recognition (AR)”: This is a well-studied Machine Learning domain whose aim is to
| |
| predict ADLs out of a binary sensor dataset. The major challenges are data sparsity [1] and
| |
| access to annotated data [2]. Given the aforementioned dataset, time series data of ON/OFF
| |
| connection of appliances, we need to infer “activities”, such as: cooking, sleeping,
| |
| entertainment, time off, making hot drinks; as a first step. Both supervised and unsupervised
| |
| methods can be used; although supervised approaches require manual labelling.
| |
| | |
| • “Activity Analysis”: Once the activities are extracted. It is equally important to analyse their
| |
| periodicity and variability. Thus meaningful routines can be extracted and summarised to be
| |
| presented to the Carers [2].
| |
| | |
| • “Insights from Activity Analysis”: What are the valuable applications that can be developed?
| |
| | |
| This is an exploratory stage to come up with new intelligent services for the Carers. For
| |
| instance, based on the Activity Analysis:
| |
| | |
| o Can we group users based on their activity’s similarities?
| |
| | |
| o Can we detect whether a user is struggling to perform a particular ADL?
| |
| | |
| o Can we detect early symptoms/predictors of elderly illnesses such as dementia or
| |
| sleep disorders?
| |
| | |
| The Team should also provide an Application (eg. iOS, Android or Web App). We encourage out of
| |
| box thinking and novel UI/UX design. Questions to be considered:
| |
| How can this information to be presented to Carers so they find this a useful tool for monitoring the
| |
| “day-to-day” activities of their Caree? A Carer is a relative or friend of the Caree. Another type of
| |
| Carer could also be a professional Carer working for a care home provider.
| |
| | |
| Data will be accessed via Informetis Web API. You will have access to data across several homes over
| |
| a period of a few months.
| |
| | |
| Definitions:
| |
| | |
| - Carer: Care Assistant/Helpline Staff or a relative/friend who keeps an eye on the Caree
| |
| | |
| - Caree: Typically, an elderly person (around 75 years plus) living on their own. They might
| |
| also be a person with a degree of dependency which requires some support/help to carry
| |
| out their Activity of Daily Living
| |
| | |
| - Activity of Daily Livings (ADLs): the activities that are performed on a daily basis for the
| |
| majority of Caree’s (eg. Cooking, Cleaning, Drinking etc.)
| |
| | |
| - Activity Recognition (AR): it refers to a specific Machine Learning domain aiming to infer
| |
| ADLs from a binary sensor dataset.
| |
| | |
| - Binary sensor dataset: a dataset mainly compounds of time-series data whose value is either
| |
| 0 or 1. Usually the dataset is skewed (i.e. there are more 0s than 1s).
| |
| | |
| Bibliography:
| |
| | |
| [1] Javier Medina-Quero, Shuai Zhang, Chris Nugent, M. Espinilla, “Ensemble classifier of long shortterm
| |
| memory with fuzzy temporal windows on binary sensors for activity recognition”, Expert
| |
| Systems with Applications, Vol. 114, pp. 441-453, ISSN 0957-4174. 2018.
| |
| | |
| [2] Julie Soulas, Philippe Lenca, André Thépaut, “Unsupervised discovery of activities of daily living
| |
| characterized by their periodicity and variability”, Engineering Applications of Artificial Intelligence,
| |
| Vol. 45, pp. 90-102, ISSN 0952-1976, 2015.
| |
