Abstract: Smart home aims to offer ambient assisted living environment to its occupants through the design of activity recognition system. Most of the existing works in smart home design is carried on single occupant monitoring and recognition. In reality, more than one occupant resides in a home environment and thus it essential to extend the activity recognition to multi occupancy. Sensor data association and occupant enumeration are the major concerns in the design of multi occupancy activity recognition system for smart home. Hence, the Multiple Occupants Activity Recognition and Tracking (MOART) framework is proposed for data association and enumeration using statistical and artificial intelligence techniques. Hidden Markov Model and Joint Probability Data Association approaches are used for activity recognition and data association respectively and is further enhanced using contextual approaches. Experiments are carried with the real time smart home dataset and the study shows that the proposed approach outperforms existing approaches.

Keywords: Smart home, Multioccupancy, Activity recognition, Occupant count.