Gait recognition basded on Contourlet Transform and Collaborative Representation

Gait recognition is an important task for video surveillance systems. In this paper, we propose a novel gait recognition algorithm based on Contourlet Transform and Collaborative Representation. Contourlet transform is adopted to extract multi-scale and multi-direction features of Gait Energy Image (GEI). Then, different from recent works, we use Collaborative Representation based Classification (CRC) instead of Sparse Representation based Classification (SRC) to classify the gait sequences. CRC uses l₂-norm instead of the expensive l₁-norm for coefficients regularization, which makes it much more efficient. The experimental results on the benchmark CASIA B gait database demonstrate that our method can get state-of-the-art performance.