Speed Invariant Gait Identification using Frequency Domain Feature

Abstract

Changes in walking speeds is one of the most challenging covariates in gait identification. Walking speed change is accountable to significantly alter walking patterns of an individual. This covariate is also responsible for the intra-class variations; and for that, the training dataset fails to cope with. This covariate may affect some parts of human gait and keep other parts untouched. To address this issue, parts definition and selection is carried out based on GEI feature to show which parts are contributing more and which parts are contributing less in gait identification under different speed changes. After having a suitable combination of different parts of gait, frequency domain feature is extracted from that part and used in gait identification. We evaluate and analyze the performance of our proposed method and other existing methods on a publicly available OU-ISIR Treadmill Dataset A. In this study, we focus on the walking speed and deal with cross speed gait identification and achieve better identification rate.