Gait Recognition: Template-Classifier Performance Analysis

Gait recognition is identifying the individual person based on their walking pattern. This paper investigates the effectiveness of different gait templates and classifiers to accurately identify individuals based on their unique walking patterns. Three gait templates—Gait Energy Image (GEI), Gait Moment Image (GMI), and Motion Flow Energy Image (MFEI) utilized to extract spatial and temporal characteristics of gait sequences. histogram of oriented gradients (HOG) was employed for feature extraction and directional information, while linear discriminant analysis (LDA) was applied for dimensionality reduction to enhance computational efficiency and class separability. We evaluated the performance using three classifiers—Support Vector Machine (SVM), k-Nearest Neighbors (KNN), and Nearest Centroid (NC)—under varying view angles, clothing, and carrying conditions. Among these, the combination of the GMI template and NC classifier achieved the highest accuracy of 83.85%, demonstrating robustness to intra-class variability and covariate conditions. The GMI template combined with the NC classifier is effective in recognizing individuals even when factors such as different view angles, clothing, and carrying conditions are present. This robustness to variability and covariate conditions makes gait recognition technology a promising tool for enhancing security and surveillance systems.