Human-computer interaction must follow the natural interaction principle of dasiaPeople-orientedpsila. People prefer to do most of the interactions with hands in the daily life. This paper presents a vision-based multi-user gesture recognition human-computer interaction technology which achieves natural interactions through clicking and moving the fingers on the computer screen. The method uses three cameras to capture video images of computer screen, and recognizes the fingers using image processing technique. The input positions are detected with the obtained fingertip positions and cameras parameters. A training process of gesture recognition and an average arithmetic of redundant data are introduced to improve the accuracy and robustness of system. Experimental results demonstrate the usability and adaptability of system.