Human Posture Recognition Method Based on Skeleton Vector with Depth Sensor

Abstract

This paper proposes a human body gesture recognition method. The method collects bone information of the human body through the Kinect depth sensor, and then uses the direction cosine method for feature extraction.Finally, the feature vector is sent to the BP neural network for training and recognition. In the case of less sample training, the system can still accurately identify the five postures of standing, sitting, leaning forward, leaning backward and underarm. The input is composed of low-dimensional bone information, so the speed and recognition response speed during network training is better than the conventional RGB image recognition method. Experiments show that the recognition rate and real-time performance of the system can meet the needs of practical applications.