Static Hand Gesture Recognition Using SOM-Hebb Classifier

Abstract: Gestures are physical actions made by human to convey feelings or expressions to others. Hand gestures are basically classified as static and dynamic. It is used as sign language for deaf and dumb, controller less video gaming, Smart TV, video surveillance, human robot interaction, biometrics, virtual and augmented real time applications. This proposed work focuses on static hand gesture recognition system using hybrid network consisting of SOM (Self-organizing map) and Hebb classifier. SOM classifier is a single layer feedforward neural network. It uses Hebbian learning algorithm to in its association to help in identifying categories. Mapping is done by SOM method and network training is obtained by Hebbian learning algorithm. The performance of this proposed work is simulated using MATLAB and evaluated in terms of recognition accuracy. The obtained accuracy of the system is 91.11%.

Keywords: Euclidean distance, Fourier descriptor, Hebb, Self-organizing maps (SOM), Skin color Segmentation.