HAND POSTURE RECOGNITION USING SKIN SALIENCY MAP AND TEXTURAL EVIDENCE

Abstract

The aim of this thesis is to present a hand gesture recognition approach in static hand posture images. The major steps of the work includes (a) segmentation of hand region from rest of the image, (b) formation of saliency image (c) feature extraction using Gabor filter and Pyramid histogram of oriented gradients (PHOG). The YCbCr color model is used to detect the skin region of the hand, whereas the saliency map assigns a higher rank to the visually prominent area along with edges of the hand region. Gabor filter is used to extract texture feature at various orientations and scales while PHOG extract the shape of hand by computing the spatial distribution of skin saliency image. Finally, the extracted features are used to classify through Support Vector Machine (SVM). The performance of the proposed algorithm is demonstrated on publicly available datasets, and the recognition accuracy achieved on these datasets are compared with similar state-of-the-art, which shows superior performance.