Weighted Group Sparse Bayesian Learning for Human Activity Classification

Recently, many researchers have focused on the human behavior recognition based on micro-Doppler signal. In this paper, we propose a sparse representation classification approach based on weighted group sparse Bayesian learning (SRC-WGSBL) for human activity classification, which introduces the property of group sparsity to distinguish the sparse coefficients between different classes. In addition, the use of Bayesian model for sparse coding is helpful to have robust classification performance in practice. Extensive experiments on a public database have been carried out to compare the performance of the proposed approach with support vector machine (SVM) and sparse representation classification based on orthogonal matching pursuit (SRC-OMP). Experimental results demonstrate that the proposed approach is effective and has better performance.