A new effective part selection approach for part-based gait recognition

In part-based human gait recognition, choosing the appropriate body parts is the most challenging problem. The various cofactors such as carrying conditions (backpack or side bag or hand bag), cloths (long coat or jacket or gown) affect various body parts. Here, we proposed a method for detecting the various cofactors in early stage. We have taken some pixel points which are marked as a boundary for each cofactor. The weight of these points is calculated and used to detect the particular cofactors in early stage. We divide the human body into seven body parts based on anatomical studies of gait. Discarding the body parts where the cofactor is present, the remaining parts are used in classification. We have achieved better result compared with other classical methods.