Recognition of falls using dense sensing in an ambient assisted living environment

Long-lie situations following a fall is detrimental, particularly for older people as they are not only affected physically but also psychologically. In this paper, we describe a dense sensing approach for falls detection in an ambient assisted living environment such as a room, hall or a walkway. We utilize a smart carpet consisting of an array of Radio Frequency Identification (RFID) tags arranged in a 2-dimensional grid to create an unobtrusive monitoring area and to detect falls among other activities. In particular, we propose an algorithm based on a heuristic and machine learning to detect 'long-lie' situations. The proposed algorithm minimizes the effects of noise present in the RFID information by relying on eight features extracted using only binary tag observation information from a possible location of a fall on the smart carpet. By evaluating the proposed approach with broadly scripted activities, which included a complex set of walking patterns, we show that the proposed algorithm depicts a good overall performance of 93% F-score.