Identification of Multi-gait Using Spectral Analysis of Gravity Center Track

Abstract

A new interference factor that may influencing accuracy of gait recognition system seriously is addressed in this paper. The gravity center track (GCT) of multi-gait is supposed to be the linear combination of single GCT of each participant. We decompose GCT of multi-gait by Fourier transformation and take the frequency distribution coefficient as the gait feature. As pedestrians will adjust their speed from walking alone to traveling together, we perform speed variation on GCT of single gait before matching with multi-gait. The experimental results show that GCT achieves good recognition effect on multigait, and has great robustness to dress.