Author : Riant Kaur and Heena

Every individual has its own style of walking. It varies from person to person. A person can be identified by its walking style which is termed as Gait. This is a better way of identification as it is associated just with the individual and not with the surpassing information from one place to another in the background. In this dissertation, we work on live home videos in MATLAB which are converted into frames and each frame is being worked upon. Hanavan’s model has been introduced for feature extraction, extracting six anthropometric parameters giving clear discrimination among various individuals. Feature detection and feature matching is executed by implementing SURF detector which augments the accuracy. SVM classifier is used for separability of extracted features. It is enhanced by implementing K-means clustering analysis algorithm and MDA (multi-linear discriminant analysis framework) in recognition phase producing CCR (Correct Classification Rate), matching time, MSE (Mean Square error) and PSNR (Peak Signal to Noise ratio) comparisons. Therefore, these techniques give faster results and better accuracy.

Keywords: Feature extraction, Gait Pal and Pal Entropy (GPPE), Hanavan’s model, Speed up Robust Feature detector (SURF), Support Vector Machine (SVM), K-Means Clustering iterative algorithm, Multi-linear Discriminant Analysis (MDA).

Article published in International Journal of Current Engineering and Technology, Vol.6, No.1 (Feb-2016)