A K-NN and Sparse Representation Based Method for Gesture Recognition

Sparse representation classification is widely used in pattern recognition. In this paper, a k-nn and sparse represent based method for gesture recognition (KSRC) is proposed. In order to reduce the computational complexity problem in sparse representation, KSRC exploits K nearest neighbors of the testing sample from all training samples and represents the test sample as a linear combination of the K nearest neighbors, then solving the l1-norm constrained least square problem. Taking linear interpolation method to force all data to be in the same space. Experiments show almost perfect user-independent recognition, and user-mixed recognition and user-dependent recognition, and speeds up 4 times than SRC.