Statistical gabor-based gait recognition using region-level analysis

Recently, great attention has been given to human identification based on gait biometric. Gait recognition is the process of identifying people through their style of walking. Gait can identify subjects from long distance and without inconvenient interruption. However, most of the existing Gabor-based gait recognition approaches suffer from the curse of dimensionality, even after utilizing dimensionality reduction techniques. Consequently, this adds more computational and storage burdens and may make the human identification process more difficult. This paper presents an effective gait recognition method based on the statistical analysis of Gabor patterns. This method adopts the Gait Energy Image (GEI) to capture the spatio-temporal characteristics of the human gait sequence during one motion cycle. Then, it applies Gabor filters and encodes the magnitude of the resulting Gabor responses from different-sized regions using set of statistical measures. Consecutively, classification is performed using Support Vector Machine (SVM) classifier. Finally, intensive experiments are carried out on the two benchmark OU-ISIR A and CASIA B gait datasets to demonstrate the effectiveness of the proposed method. The experimental results have shown that our proposed statistical method can achieve promising performance in term of accuracy and efficiency.