Real time hand shape recognition for man-machine interfaces

The authors describe a hand shape recognition method for man-machine interfaces using a pipeline image processor. The basic strategy is to use simple but effective real-time dynamical image processing to enhance the reliability of hand shape recognition. Low-level image processing reduces the data required for recognition of a hand shape. An algorithm for a noncontact-type hand shape recognizer has been developed for incorporation into a virtual reality environment. To realize this recognizer, the most important considerations are processing speed, the pointer's positional accuracy, and command recognition rate. The usefulness of the proposed method is verified through experimental results for these three points.<
>