Leth-Gait: A Fusion Formula for Gait-Based Gender Classification Using Depth Images

Gender can be easily identified by humans by looking at a person's appearance. As the human brain is able to identify and classify gender, this study will describe a method for classifying gender using depth images. OU-ISIR Gait Database Large Population Dataset was used to perform this study. This classification process was performed by machines such as computers and software. The proposed method used Gait Energy Image (GEI) to normalize the depth image. After that, they were projected into their PCA plane, gaining a very strong cyclical picture of a change of perspective. Then the feature was extracted by measuring some body parts from the reconstructed image. Fine Gaussian Support Vector Machine (Fine Gaussian SVM), Decision Tree, and Weight K-Nearest Neighbor Classifier (KNN) was used separately as the classification method. The experimental results showed that the proposed method produced significant results, obtaining 96% accuracy using Fine Gaussian SVM, 98% accuracy using Weight KNN, and 78% accuracy using Fine Decision Tree classifiers.