Human gait classification using combined HMM & SVM hybrid classifier

Abstract

The paper describes the work on human gait recognition using Hidden Markov Model (HMM), Support Vector Machine (SVM) and Hybridized classifiers (developed using both HMM and SVM). Human gait data obtained from CASIA gait database were segmented to locate major human body part and generate corresponding stick view in order to extract gait features. A total of 25 features were obtained using the length of body parts and major joint angles along with other features and classified using HMM, SVM and Hybridized classifiers. The Hybridized classifier outperforms individual classifiers by 11.25% and 18.14% during training and testing respectively.