Dynamic gesture recognition algorithm in human computer interaction

摘要

Human computer interaction (HCI) is a classic topic, and hand gesture is one of the most natural languages, so the key point in human computer interaction is hand gesture recognition. A skin color model based on Hue is utilized to detect skin color regions like hands. To distinguish the face and hand, a simple and fast hand segmentation algorithm is developed. Moreover, we use the least square method to fit the trajectory of hand gravity motion and the angle and direction of the movement is got by the inverse trigonometric function of the slope of the fitted line. Ten people doing 80 hand gestures in low and high illumination environment are tested. Experimental results show that the accuracy is 94.375% in average and the processing time is 34ms per frame. These demonstrated the good real-time and robustness of the proposed approach.