A Novel Group-Sparsity-Optimization-Based Feature Selection Model For Complex Interaction Recognition

Interaction recognition is an important part of action recognition and has various applications such as surveillance systems, human computer interface, and machine intelligence. In this paper, we propose a novel group-sparsity-optimization-based feature selection model for complex interaction recognition. Firstly multiple local and global features are concatenated into a feature pool, and then based on the group sparsity optimization, different feature types are automatically selected to fit specific interaction categorization. We test our method on the benchmark dataset: the UT-interaction dataset. Experimental results substantiate the effectiveness of the proposed method on complex interaction recognition tasks as compared with current state-of-the-art methods.