Remora Optimization Based Sample Weighted Random SVM For Human Gait Authentication

Abstract

In this paper, we present a novel ESVM-SWRF method for authenticating human using a gait cycle. The different covariates related to walking are analyzed and investigated. The walking speed of people may change due to the individual body structure, gender, and age thereby creating a complex situation. Based on the studies over past decades, different perspectives with cross-speed gait authentication were suggested. The factors influencing the identification of gait are some of the covariate factors namely walking speed, injuries, walking surface, viewpoint, and clothing. Our proposed work uses an effective dataset CASIA-C. Most of the existing techniques achieved a nearly 100% authentication accuracy rate for normal walking conditions but their performance is not optimal when applied under different covariate conditions. Our proposed work proves a high accuracy rate of 89% for different covariate conditions compared to other existing methods.