Genetic feature selection for gait recognition

Faezeh Tafazzoli; George Bebis; Sushil Louis; Muhammad Hussain

doi:10.1117/1.JEI.24.1.013036

25 February 2015 Genetic feature selection for gait recognition

Faezeh Tafazzoli, George Bebis, Sushil Louis, Muhammad Hussain

Author Affiliations +

Journal of Electronic Imaging, Vol. 24, Issue 1, 013036 (February 2015). https://doi.org/10.1117/1.JEI.24.1.013036

Abstract

Many research studies have demonstrated that gait can serve as a useful biometric modality for human identification at a distance. Traditional gait recognition systems, however, have mostly been evaluated without explicitly considering the most relevant gait features, which might have compromised performance. We investigate the problem of selecting a subset of the most relevant gait features for improving gait recognition performance. This is achieved by discarding redundant and irrelevant gait features while preserving the most informative ones. Motivated by our previous work on feature subset selection using genetic algorithms (GAs), we propose using GAs to select an optimal subset of gait features. First, features are extracted using kernel principal component analysis (KPCA) on spatiotemporal projections of gait silhouettes. Then, GA is applied to select a subset of eigenvectors in KPCA space that best represents a subject’s identity. Each gait pattern is then represented by projecting it only on the eigenvectors selected by the GA. To evaluate the effectiveness of the selected features, we have experimented with two different classifiers: k nearest-neighbor and Naïve Bayes classifier. We report considerable gait recognition performance improvements on the Georgia Tech and CASIA databases.

Citation Download Citation

Faezeh Tafazzoli, George Bebis, Sushil Louis, and Muhammad Hussain "Genetic feature selection for gait recognition," Journal of Electronic Imaging 24(1), 013036 (25 February 2015). https://doi.org/10.1117/1.JEI.24.1.013036

Published: 25 February 2015

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available