Motion Posture Recognition Method Based on Support Vector Machine

To solve the problem of differential human motion recognition of different users in human-computer interaction, SVM is used to classify and recognize the motion posture. We use Kinect sensor to capture human motion, generate depth image, and establish three-dimensional human model after processing. Then the target behavior recognition adopts a two-level SVM classifier to map various action signals to the feature space to form a feature vector with a certain dimension. During the process of outputting recognition results, corresponding confidence is output and the motion posture is determined by the change classification of motion features. The experimental results show that this research method uses the invariance and orthogonality of support vector machine to improve the recognition rate of vector optimization to more than 95%. Its the action recognition effect is good, which has better robustness with similar algorithms.