Gesture recognition system based on Adaptive Resonance Theory

We report on the moving hand gesture recognition technique using Adaptive Resonance Theory (ART). To detect the start and end points of a continuous moving gesture (known as “gesture spotting” problem), we propose the adaptive distributed prediction technique. Our results show that, unlike conventional non-recurrent neural networks, the proposed technique can be utilized usefully in reliable real-time learning (2000 times faster than with alternative methods) and recognition of continuously moving patterns.