Human activity analysis in visual surveillance and healthcare

An automatic recognition of human activities enables their use in several interesting applications of daily life. This dissertation emphases on the analysis of human activities in a visual surveillance scenario and the classification of physical activities in the therapeutic procedure using visual data. The first part of the dissertation proposes a robust gait representation to recognise the identity of a person using his/her walking style, dealing with its several real world challenges as well as taking into consideration the effects of cross-view recognition. In the second part, a complete framework is proposed to capture and analyse the movement of different body parts in human which is useful in the clinical assessment to detect any movement disorders and the assessment of the desired therapeutic program.