Gait identification based on FIR system characterizing motion of leg

Abstract

This paper presents gait identification method based on Finite Impulse Response (FIR) system characterizing motion of leg. First, the four gait features, area of footstep, angle of footstep, height and width of the human silhouette image, are calculated from the human silhouette image. Then they are expanded into Fourier series to reduce the fluctuation of human body motion. The motion of leg can be characterized by two FIR gait identification systems. For the first FIR system, the Fourier coefficients of the width of human silhouette image and the area of footstep are used as input and output of the system, respectively. For the second FIR system, the Fourier coefficients of the height of the human silhouette image and the angle of footstep are used as input and output of the system, respectively. The obtained impulse responses of the two FIR systems are used as the individual feature for gait identification. The gait identification experiments were performed on CASIA GAIT Dataset B [6], which contains 8,184 gait data for 11 view angles from 124 persons. The average of error rates obtained from 90° view angle was 3.48%.