Sparse representation classification for battlefield textual information

Sparse representation based classification has recently been shown to provide excellent results in many object recognition and classification tasks. The high cost of computing, however, is a major obstacle that limits the applicability of these methods in battlefield information process scenarios where process time and computational power is restricted. In this paper, we study a fast and computationally efficient sparse representation classification scheme for battlefield textual information in which the block sparsity of sparse coefficients is exploited. A novel sparse approximation algorithm tailored for this low complexity classification method is proposed. Experiment results show that our classification algorithm that leverages the sparse structure of the textual information outperforms plain sparse representation classification procedures in both classification accuracy and computationally efficiency.