Skip to main content

Research Repository

Advanced Search

Intelligent data processing to support self-management and responsive care

Gupta, Prankit

Intelligent data processing to support self-management and responsive care Thumbnail


Authors

Prankit Gupta



Abstract

This research is situated in the area of ambient intelligent systems for assisted living. The motivation for the research was to understand how ambient intelligent systems could be used to support people with learning disabilities in providing more personalised care, as well as function as an aid to support independent living. In the first phase of the research a series of interviews conducted with formal carers of people with learning disabilities highlighted kitchen activities as a potential area of support. This provided a focal point for the research whereby subsequent research involved the development of sensor data mining techniques and machine learning methods to recognise specific meal-preparation activities with a view to supporting task prompting. The goal of task prompting is to enable automated intervention for service users performing meal-preparation activities by tracking the activity in real-time by analysing ambient sensor data. In the second phase of the research a public smart-home dataset was used to develop a novel methodology which uses "temporal clusters" of sensor events as a pre-processing step for extracting features from the data and creating visualisations. In the third phase of the research, a data set comprising different meal-preparation activities undertaken by three participants in a shared kitchen was collected over a period of 8 weeks. This fully annotated dataset includes a combination of data from a range of ambient smart-home sensors and low-resolution thermal cameras. This dataset was used to experiment with knowledge-driven activity recognition techniques, which were used to develop a novel hybrid offline-online learning methodology for real-time activity recognition and prediction. This methodology is shown to overcome the shortcomings of existing supervised activity recognition methods, which require re-training with new data if the activity changes. The new methodology has been designed to enable learning from the user in order to track meal-preparation activities in real-time, detect deviations from the activity, and adapt to changes in the user's performance without requiring re-training. The research presented in this thesis, together with the meal-preparation dataset, are a crucial stepping-stone for the development of future technologies that offer the potential for real-time task prompting and thus could be useful in supporting people with learning disabilities in performing activities more independently. The approaches developed can also generate information that could help carers better understand how their service users are able to perform these activities and hence personalise and adapt the support they provide.

Citation

Gupta, P. Intelligent data processing to support self-management and responsive care. (Thesis). University of the West of England. Retrieved from https://uwe-repository.worktribe.com/output/7563714

Thesis Type Thesis
Deposit Date Jul 27, 2021
Publicly Available Date Jan 4, 2022
Public URL https://uwe-repository.worktribe.com/output/7563714
Award Date Jan 4, 2022

Files





You might also like



Downloadable Citations