A Behavioral Recognition Method with Data Fusion and Supervised Learning for Tele-Assistance Systems

The paper presents a learning-based method for human behavioral pattern/activity recognition with design for tele-assistance systems. The method includes a behavioral classification component aiming to make reliable predictions or to early detect abnormal cases potentially revealing the health degradation for the individual. The functional architecture contains Feature Engineering (FE) techniques besides the typical processing steps. Adding a FE step helps to enrich the original feature space with new informative variables in order to discover hidden patterns within data, with significant added value for the human behavioral pattern recognition.