A Logical Model for Multiple People Activity Recognition Using Non-intrusive Sensors for Geriatric Care

Most of the work on activity monitoring for geriatric care assumes the presence of a single resident within a smart apartment. But in general, especially in a country like India, an aged person is residing with her housemate/spouse. This class of activity monitoring within a smart home is termed as the “Multiple People Activity Recognition” problem. As per the present state of the art, the challenge of identifying the person with the concerned activity is addressed through the tracking of a resident through wearable sensors. The experience shows that the aged persons (maybe the patient with dementia) are feeling annoyed or may forget to wear these devices. Thus, it becomes a crucial challenge for the researcher to handle the “Multiple People Activity Recognition” considering a smart home equipped with only ambient (non-intrusive) sensors. In this paper, we have attempted to solve the problem using the Logical Factorial Hidden Markov Model for modeling separate chains for different residents. The logic base, in the form of rules, is created from the dataset as prepossessing and subsequently fused within the model for better recognition. The existing works in this domain are limited; we have also compared our proposal with the existing notable solutions. The experimentation on benchmark datasets shows the supremacy of the proposed model.