A tracking method of human body motions and an identification method thereof are disclosed. The tracking method includes the following steps: providing a continuous and multi-frame dynamic image and establish a background based on the dynamic image; simultaneously performing a pretreatment and a human body position process on every frame of the image so as to acquire and output human body position information of the frame. The identification method includes the following steps: arranging a human body identification mode, such as a mouse mapping mode, a state mode, a motion mode and the like; identifying the human body position information of the every frame based on the human body identification mode and outputting the identification information. By using the methods of the invention, a video can be tracked continuously, high-accurately and real-timely for the every frame and the human body position information can be output. Especially, the methods can be used to tracking the motions of small face sport organs, such as eyes, a mouth and face muscles and motions of an upper limb part with a large motion amplitude, a high frequency and complex motion characteristics. Simultaneously, attitude information, such as a mouse position, a mouse movement, single click, double click, hands position, a left hand movement from left to right and the like, can be identified and output.