To solve the problem of Real-time differential in current ongoing human action recognition algorithm, data motion capture method based on DTW integral-histogram is proposed. Firstly, the original action is classified by integral-histogram. Then, spatio-temporal series of poses are clustered to create a statistical representation of actions and the similarity between two histograms is measured using Bhattacharyya distance. Finally, a dynamic programming algorithm, inspired from the Dynamic Time Warping method is proposed to finish recognition. The approach proposed is tested on well-known benchmark datasets and compared to several state-of-the-art methods. The obtained results have clearly shown the success and effectiveness of our solution, even in the presence of noise and/or similar actions in the datasets.