Recognition of Hand Gestures using Visible Light and a Probabilistic-based Neural Network

Abstract

Gestures by hand are an effective and efficient means to interact with machines. We present a method for recognizing gestures via the analysis of interrupted light patterns using visible light. Utilizing low-cost and easily accessible components, the system can exploit existing light communications infrastructures. A probabilistic type neural network is applied to learn the sequence of obfuscated light patterns during movement of the hands. A novel preprocessing of the received light is introduced enabling the use of an accurate but low-complexity class of neural network.