FLGBP: Improved Method for Gait Representation and Recognition

Abstract

Gait recognition is one of the lately emerged technologies in the field of biometrics which has various applications in the security and medical domains. This paper presents a novel holistic method for vision-based gait representation and recognition, named FLGBP. This method adopts the gait energy image 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 image using a fuzzy local binary pattern (FLBP) operator. Gabor magnitude is utilized due to its richness of discriminative gait information rather than the phase of Gabor responses. A comparative study is made to measure the degree of improvement of FLGBP over other local Gabor patterns. Experiments are conducted using CASIA B dataset and classification is performed using the support vector machine (SVM). Experimental results show promising performance for the proposed FLGBP descriptor.