A Framework to Recognise Daily Life Activities with Wireless Proximity and Object Usage Data

Human behaviours are complex and challenging task to learn from daily life activities. To recognise daily life activities, smart environments are becoming very popular as platforms that can be used to obtain the context sensitive information for human activity and behaviour recognition. The motivation here is to investigate a mechanism that can recognise both indoor and outdoor tasks and activities of low entropy people such as elderly people and dementia patients, by using wireless proximity data and objects usage data. Wireless proximity data is used to recognise the outdoor tasks and activities, whereas, object usage data is generated by objects that are used when performing every day activities inside the home/building and is collected through Radio Frequency Identification (RFID) sensors. The approach is divided into two levels, i.e. lower tier and higher tier. The lower tier is responsible for the recognition of tasks from the raw sensor data in the form of a list of tasks. This list of tasks is further utilized by higher tier to recognise the high level activities performed by the target users. Number of different scenarios and experiments are performed to test the functionality of the said approach in different circumstances. Keywords-Tiered Approach; Human Behaviour; daily Life Activities; Low Entropy; Elderly care