An Improved DTW Method for Human Behavior Recognition
Download as PDF
Shengnan Li, Yan Ma, Hui Huang, and Shunbao Li
Human behavior recognition has been widely used in computer vision. However, the accuracy and stability of the recognition algorithm will be affected by the different data acquisition objects and object behavior. To solve this problem, an improved human behavior similarity measure method based on improved dynamic time warping (DTW) is proposed. First, the data of human joints is acquired by Kinect. Then, the path is corrected with DTW algorithm according to the motion distance between the motion to be tested and template. Next, the corrected path is further adjusted via curve fitting. Finally, the behavior similarity between them is obtained by the adjusted path. The experimental results show that the performance of the proposed method is more stable and accurate.
Human Behavior Recognition, Kinect, DTW, Curve Fitting, Behavior Similarity