A fast hand motion detection based on FIFO

Abstract

This paper proposes a fast hand motion detection based on FIFO and path overlap detection. The previous researches in foreground hand detection need to spend too much time to filter out noises and non-hand information. Because a low cost embedded system cannot afford a heavy time-consumed computation, hence, for the foreground detection to be implemented on a low cost embedded system, its time complexity cannot be too complex. In the paper, a new efficient foreground hand detection is proposed. The multiple consecutive difference images are used. Through the FIFO and path overlap, the outcome tracked video of the hand motion detection becomes much smoother. Besides, we combine the motion detection and KCF on HOG to improve the tracking while hand is passing face. The searching area of KCF can be limited in a smaller area to reduce the searching time. The experimental results show that the proposed method has a high probability of locating the raising-hand and can save the time wasted in performing hand recognition in the non-hand foreground area. Especially, when hand passes the front of face, the proposed can track the moving hand correctly and more efficiently.