Human Identification using Gait

Keeping in view, the increasing importance of biometrics in modern security and surveillance systems, the human identification using Gait provides a non-invasive method for reliable identification and is a promising area of interest for research. Due to higher execution time and the complexity for gait recognition, it is difficult for the existing algorithms to meet the real time requirement. To overcome all these problems this paper proposes a technique which uses sparse representation for human identification. The input images used in this work are mainly binary silhouettes. The performance of the algorithm mainly depends on the quality of the binary silhouettes. Here the Background subtraction method is fuzzy correlogram based which is computationally less intensive. Even the background subtraction are robust and some techniques for eliminating shadows can result the erroneous images where some important portions of individual are missing. This will lead to poor performance of recognition. Frame Difference Energy Image (FDEI) is preprocessing algorithm has been used to eliminate these disadvantages. The technique starts with calculation of Gait period which is then represented spatio-temporally by Gait Energy Image. Gabor wavelet and Local Binary Pattern are applied to Gait Energy Image for feature extraction. Finally classification is done by Sparse representation. Experimental results show that proposed algorithm outperforms the conventional methods effectively and it is an efficient and effective representation of gait for human identification. Gait Database is used in the experiment is CASIA B with 90 degrees view