Feature selection for gait recognition

This paper focused on unique concept of extracting the gait features of walking human from sequences of silhouette images for recognition purpose. Discrete Cosine Transform (DCT) was evaluated as feature extraction solely followed by combination with Principal Component Analysis (PCA) as feature selection. Then, the entire feature vectors that were extracted are used as input to classify using artificial neural network. All developed method tested extensively with CASIA gait databases of various gait sequences taken that evaluated for both genders, without restriction on clothing and carrying of object(s) by subject. Recognition rate of 80% archived with Principal Component Analysis as dimensional data reduction was promising. Initial findings showed that combination of DCT and PCA is apt for gait recognition.