The continual increase in the population of older adults in the next 50 years envisages an increase of dependants on the family and the Government. Assisted Living technologies are information and communication technologies to assist, improve and monitor the daily living of the old and vulnerable population by promoting greater independence and providing a safe and secure environment at a reduced cost. Most of the assisted living technologies are passive sensor-based solutions where a number of embedded or body-worn sensors are employed or connected over a network to recognize activities. Often the sensors are obtrusive and are extremely sensitive to the performance of the sensors. Visual data is contextually richer than sensor triggered firings. Visual data along with being contextual is also extremely sensitive.
Since visual data is intrusive, a qualitative study among older adults within the community was carried out to get a context of the privacy concerns of having a camera within an assisted living environment. Building on the outcomes of the focus group discussions, a novel monitoring framework is proposed. Following the framework, Activity Level, as an effective metric to measure the amount of activity undertaken by an individual is proposed. Activity Level is estimated by extracting and classifying pixel-based and phase-based motion features. Experiments reveal that phase-based features perform better than pixel-based features. Experiments are carried out using the novel Sheffield Activities of Daily Living Dataset, which has been developed and made available for further computer vision research for assisted living.