Activity Recognition via Feature Decomposition

This paper presents a new recognition method for human motion, which is represented by Haar wavelet transform and recognized by Coupled Hidden Markov Model. We tackle the challenge of detecting the feature points by Haar wavelet transform to improve the accuracy. We extract binary silhouette after creating the background model. Then the low-level features are detected by Haar wavelet and principal vectors in two subspaces are obtained. We utilize Coupled Hidden Markov Models to model and recognize them, and demonstrate their usability. Compared with others, our approach is simple and effective in feature detection, strength in robustness. Therefore, the video surveillance based on our method is practicable in (but not limited to) many scenarios where the background is known.