Improving single view gait recognition using sparse representation based classification

• Published in: TechSym 2016 : 2016 IEEE Students' Technology Symposium : 30 September-2 October 2016, IIT Kharagpur pp. 317 - 321
• Main Authors: Das, Sonia, Sahoo, Upanedra Kumar, Meher, Sukadev
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2016
• Subjects: Covariance matrices; Dictionaries; dictionary; Feature extraction; Gait; Gait recognition; Histograms; l 1 -minimization; Legged locomotion; sparse representation; Training
• DOI: 10.1109/TechSym.2016.7872703

This paper explores a better way of recognition using sparse based representation, by taking into account a number of covariates that affect single view based gait. Nevertheless, the conventional methods couldn't handle covariates effectively. Our propose framework comprises a dictionary, which describes five segments of a subject over a gait period. The feature vectors are educed from ellipse based parameters from each segments and fused to form a covariance matrix. Each matrix is used as dictionary atom and solved using - l 1 - minimization. The linear representations of sparse codes of different atoms are used for recognition. The proposed method is compared with that of state-of-the-art methods.