A NOVEL FEATURE EXTRACTION SCHEME FOR HUMAN GAIT RECOGNITION

With the increasing demand of visual surveillance systems, human recognition at a distance has gained extensive research interest. Gait is a potential behavioral feature to identify humans based on their motion. This paper describes a new scheme for extracting and selecting features from the gait of a human for recognition. The scheme combines both Key Fourier Descriptors (KFDs) and principal component analysis (PCA) techniques. This leads to a strength in reducing feature space by KFD, and increasing accuracy by PCA. Also, it is shown that the proposed scheme leads to a higher correct classification rate than schemes that depend on KFD alone or PCA alone.