Hand gestures recognition with improved skin color segmentation in human-computer interaction applications

Hand gesture has significant roles in human’s interaction and the hand gesture recognition itself nowadays becomes an active research area in human-computer interaction. Previous researches on hand gesture recognition used various techniques and tools such as Kinect and data glove. Hand gesture recognition area has many challenges, such as variation of illumination conditions, rotation problem, background problem, scale problem, and classification or translation problem. This research uses computer vision techniques to recognize hand gesture in human-computer interaction to control various apps, such as slideshow presentation, music player, video player, and PDF reader app for people with bare hand and in complex background of the image via web camera. Thus, a method is required to cope with background and skin detection problem. The proposed method combines two color spaces into HS-CbCr format for skin detection and uses averaging background for solving the background problem. The experimental results show that the proposed method is able to recognize hand gesture and reach up to 96.87% of correct results in good lighting condition. The accuracy of hand gesture recognition is influenced by lighting condition. The lower changing illumination on video occurs, the higher accuracy of hand gesture recognition is generated.