Real-time recognition of activity levels for ambient assisted living

Activity level as a metric to monitor the daily living of the elderly is often carried out using passive sensor networks. With the reduction of camera prices, there is a growing interest of video-based approaches in the domain of assisted living. In this paper the concept of activity level recognition in context of tracking the movement pattern of an individual is explored using a video-based framework at real time. Simple motion features are modelled over time and classified them into different activity levels using a neural network. For the experiments, the Sheffield Activities of Daily Living (SADL) dataset is used where each activity is simulated within a simulated assisted living environment under two different illumination conditions. Our experiments show that the detection rate for each of the activity level is well above 80% and the detection time is approximately 30 seconds.