Boosting real-time recognition of hand posture and gesture for virtual mouse operations with segmentation

The design and implementation of polylogarithmically or polynomially bounded algorithms on faster processors has gained popularity and attracted the attention of both researchers and practitioners. The evolution in the computer hardware technology has boosted the development of real-time applications which are expected to respond within a strict time frame. One attractive sophisticated application, which requires real time response, is image capturing and recognition for effective human computer interaction. It is gaining popularity, especially after the development of hand held devices and touch screens. Real-time video processing response time is expressed by means of frame sequences; device dependent capability (20 frame/sec) designates real-time restrictions (a frame is needed to be processed within 50 ms). Video processing of virtual mouse operations requires real-time recognition, i.e., no delay in response can be tolerated. There are indeed several attempts to recognize hand gestures for different purposes. Sign language recognition stands out as the most popular one. However, virtual mouse operations may also be used in general by the majority of people in parallel for the proliferation of different applications on a variety of platforms such as tablet PCs, embedded devices, etc. One significant advantage of such systems fulfills the need for extra hardware system. To this end, we have developed a novel real-time virtual mouse application. Our system architecture recognizes defined postures and gestures. We have implemented, tested, and compared the performance of four methods, namely Chai (static), face (dynamic), regional (dynamic), and Duan. Further, various conditions, such as lighting, distinguishing skin color, and complex background have been considered and discussed.