Real-Time Dynamic and Multi-View Gait-Based Gender Classification Using Lower-Body Joints

Gender classification based on gait is a challenging problem because humans may walk in different directions at speeds and with varying patterns. The majority of investigations the literature relied gender-specific joints, whereas comparison lower-body joints received little attention. When considering it important to identify gender person his or her walking style using Kinect Sensor. In this paper, logistic-regression-based model for proposed. proposed approach divided into several parts, including feature extraction, selection, human classification. Different joints’ (3-dimensional) features were extracted To select significant joint, variety statistical techniques used, Cronbach’s alpha, correlation, T-test, ANOVA techniques. average result from Coronbach’s alpha was 99.74%, which shows reliability Similarly, correlation data show difference between males females during gait. As p-value each zero less than 1%, T-test demonstrated that all nine are statistically Finally, binary logistic regression implemented classify selected features. experiments real situation involved one hundred twenty (120) individuals. suggested method correctly classified 3D captured real-time Sensor 98.3% accuracy. outperformed existing image-based systems.