Unobtrusive Human Activity Recognition

With recent developments in cheap sensor and networking technologies, it has become possible to develop a wide range of valuable applications such as the remote health monitoring and intervention depicted above. Theseapplications offer the potential to enhance the quality of life for the elderly, afford them a greater sense of security,and facilitate independent living. For example, by monitoring the daily routines of a person with dementia, an elder assistant service can track how completely and consistently the daily routines are performed,and determine when the resident needs assistance.
Central to realising these applications is automatic activity recognition, which is emerging as an important area of
research in recent years. Although many approaches have been proposed, activity recognition techniques still do not currently perform well outside of the laboratory. First of all, existing approaches typically exploit supervised learning algorithms such as conditional random fields (CRF), Markovmodels, and dynamic Bayesian networks. These approaches require a significant amount of  labeled sensor data for training. Unfortunately, manually collecting and labeling human activity data is extremely time consuming, laborious, error-prone, and sometimes even impossible (e.g., older people with dementia). In addition,they can only recognise pre-selected activities and do not deal with behavior changes (e.g., even same individual might perform a same activity in different ways). Finally, the success of the existing approaches relies heavily on people’s involvement (e.g., wearing battery-powered sensors), which might not be very practical in real-world situations (e.g., people may forget to wear the sensors).
The aim of this project is to develop a novel system for automated human activity discovery and monitoring. We will develop innovative techniques that automatically discover, identify, and track an individual’s frequent routine activities in real-time without any manual annotation of activity data. We will also explore low-cost, unobtrusive radio-frequency identification (RFID) sensor network technology for human activity recognition.


Mr. Asanga Wickramasinghe

Dr. Qinfeng (Javen) Shi

Dr. Damith Ranasinghe



Young Volunteer

Old Volunteer: Dataset1 Dataset2



Sequence Learning with Passive RFID Sensors for Real Time Bed-egress Recognition in Older People