A Human Behavior Recognition Method for Home Service Robot

In view of the limited computing power of the processors available for service robots and the complex illumination changes and shadows in the home environment, we propose a method of human behavior recognition based on skeleton model and depth image in this paper. We define several key actions for the sequence of actions, then optimizes the actions of each kind of actions, and finally completes the behavior recognition by the similarity between the measured and the template. This experiment is carried out on a general configuration processor to simulate a service robot processor. The experimental results show that the proposed method achieves a high recognition rate for eight kinds of behavior, such as brushing teeth, standing, etc. and a stable recognition effect can be obtained under the condition of changing illumination.