Gait-Based Human Gender Classification Using Lifting 5/3 Wavelet and Principal Component Analysis

This study describes a representation of gait appearance for the purpose of person identification and classification. The gait representation is based on wavelet 5/3 lifting scheme simple features such as features extracted from video silhouettes of human walking motion. Regardless of its effortlessness, this may lead us to say that, the resulting feature vector contains enough information to perform well on human identification and gender classification tasks. We found out the recognition behaviors of different methods to total features over time functions under different recognition tasks. In addition to that, we provide results of gender classification based on our gait appearance features using a (C4.5 algorithm). So, the result of classification rate for CASIA-B gait databases is 97.98% and the result of recognition rate for OU-ISIR gait Database Large Population Dataset is 97.5%, these results have been obtained from gender classification data. Gait database demonstrates that the proposed method achieves better recognition performance than the most existing methods in the literature, and particularly under certain walking variations.