A hybrid approach for the recognition of hand postures using texture and skin saliency map

Abstract

Recognition of Hand Postures using skin, texture and saliency map is proposed in this paper. The first task is to separate the hand from the background in the input image. Hand detection depends on whether the image includes only the hand against a background or the entire person and this is a very challenging task. All the techniques which have been proposed in the past have certain limitations. Their performance is affected by the illumination or lighting conditions and complexity of backgrounds of the hand images. To overcome the above mentioned limitations, a new technique is proposed for the recognition of hand posture which involves using the Skin, Texture and Saliency map for the classification of hand gesture. This technique efficiently makes the use of different color spaces, skin detection, saliency map and texture features of the hand image. To recognize the different hand gesture classes, the features are extracted using the proposed methodology. The classification is done by combining the features from these different techniques and this final feature vector is used for the classification of hand posture. The SVM classifier is used for classifying the hand posture based on the feature. The application of this technique on array of experimental images shows that the proposed model has stable performance for a wide range of applications in different and complex environments.