Human gait recognition using relevance vector machine classifier

Gait refers to the walking style of every individual person. Recent days gait is emerged as a supporting biometric using machine vision technique. This Work aims to develop a system capable of Human gait recognition by using model free approach. The gait database consists of silhouettes ie. outer frame of the human body. These silhouettes are affected by noises and discontinuities. So preprocessing of silhouettes is required. The silhouettes after pre processing are taken for training and testing. In order to analyse the gait, temporal and spatial metrics are extracted from the model, such as variation in angles of the limb and height and width of the walking person can also be used to create the gait features. These feature vectors are used for training and testing by using Relevance Vector Machine Classifier. RVM classifier efficiently identifies the person. The silhouettes are taken from Institute of Automation Chinese Academy of Science (CASIA), China. By using robust algorithm, the quality in human silhouettes is initially improved in order to increase the gait recognition rate of a person. The proposed method produced 90% of recognition rate.