Identification method using partitioning of subject silhouettes

A method of identifying a human subject by gait recognition, the method comprising:

• obtaining a number of image frames of the human subject showing the human subject's walking style and storing the image frames in a memory,
• receiving by circuitry a gait sequence, the gait sequence including a predetermined number of the image frames of the human subject;
• processing by circuitry, the received gait sequence to generate a gait-energy-image (GEI), wherein the GEI is a sum of walking silhouette images divided by the number of images;
• computing by circuitry, a plurality of Gabor filter responses by convoluting the generated GEI with a bank of Gabor filters, the filter bank including a first predetermined number of unique scales, and a second predetermined number of unique orientations,
• partitioning by circuitry, each Gabor filter response of the computed plurality of Gabor filter responses, into a predetermined number of overlapping regions, wherein the partitioning is performed as a fraction of the height of the human subject and width of the human subject, and is denoted by horizontal and vertical lines;
• extracting by circuitry, a predetermined number of statistical features only from the overlapping regions, the extracted statistical features corresponding to texture content of the subject;
• and
• identifying by circuitry, the human subject based on a classification of the extracted statistical features.