Robust Hand Gesture Detection Based on Feature Classifier

Abstract

In this article, a novel automatic hand gesture detection approach based on the boosted classifiers is proposed, which analyze the hand gesture features. At first, hand gesture images are processed with local binary pattern (LBP) operator which has the powerful capability of texture feature description. And then these features are presented with LBP descriptor of hand gesture image which is divided into several blocks, because of too much dimension of feature vector, using Principal Component Analysis (PCA) to reduce dimension and compression. Finally, Kalman predictor is adopted to detect hand gesture, which is able to make the detection result more stable and accurate. Experiments have been carried out and the target regions have been detected and took it as the gesture truth to compare with the detection results. The proposed algorithm provides better detection result in comparison with various hand gesture detection algorithms. The proposed approach is robust to background, and it can detect the hand gesture quickly and effectively.