Dynamic Hand Gesture Recognition System for Robot Control with Minimal Number of Combined Feature Extraction Using Support Vector Machines

In this paper we proposed a vision based hand gesture recognition system that can recognize the hand gestures of a user in a real time environment. The recognized hand gestures are used as commands to activate robot motion. Proposed approach follows three main stages hand detection, hand posture recognition and gesture classification. The predefined gesture commands are recognized and classified in a real time environment. In real time situation, recognizing the gestures with good accuracy and speed is a great challenge. Our approach stands stable against multiple issues like occlusion, affine, illumination and cluttered background. Real time data is pre-processed with image enhancement techniques. Efficient hand detection is achieved through SURF features with Adaboost classifier. Hand postures are recognized using moment invariant and four morphological features with Support Vector Machine (SVM) classifier. Hidden Markov Model (HMM) with four states is used as a dynamic gesture classifier. Experimental results of the proposed approach shows good performance when compared with other existing methods.