DefinePK

Abstract

This paper proposes a new methodology to classify the person with normal walk or abnormal walk for surveillance purposes. Recognizing human walk is emerging as a critically important biometrics, challenging computer vision problem. However, the inclusion of abnormal gait dataset with normal gait databases has to be very useful to classify the normal and abnormal walking style of a person. The silhouettes are trained and tested with K nearest neighbor classifier. We introduce a more challenging abnormal walk patterns like Antalgic gait, Charlie chaplin gait, steppage gait, scissor gait, circumduction gait, inclusive with normal gait data base. The database consists of about 5000 frames with 5 different walk styles. Manual selection of persons with different walking styles resulted in high degree of variability in pose and illumination. The method starts with the extraction of human silhouettes from input videos. Initially the continuous input videos are converted into frame-by-frame by means of conversion algorithm. Each frame consists of noises and shadows. Then silhouettes are removed from noises and discontinuities to produce an abnormal gait database. From the gait data base, parameters are measured by segmenting into six portions from head to neck, neck to torso, hip to knee of both right and left leg, knee to toe of both legs, height of the blob and width has also taken as features for training. The same features extracted with test data has to be compared with trained data for classification. The proposed methodology achieves 77% classification rate for abnormal gait.