Hand Gesture Recognition in Complex Background Based on YOLOv4

Traditional methods often use a monochrome background such as green as the background so that the recognition process is not affected by the background, but this operation makes recognition limited to the use of the background plate, greatly reducing flexibility and losing the original gesture recognition to improve the level of human-computer interaction. In this paper, we propose the use of background subtraction to first eliminate the background so that recognition is not affected by the background, solving the problem of complex backgrounds affecting gesture recognition; then use YOLOv4 for gesture recognition to recognize Chinese hand gestures. After using background removal, the gesture recognition accuracy of YOLOv4 can be as high as 80%.