Saliency and Similarity Learning via Ranking-SVM Based Hand Gesture Recognition

摘要

Visual hand gesture recognition is an important area of research in computer vision and has broad applications in sign language recognition, human-computer interaction, human-robot interaction, smart surveillance and virtual reality, etc. However, existing hand gesture recognition is not real-time and can't deal with illumination changes, which makes practical use impossible. In this paper, saliency-based model of visual attention is employed to detect hands in different illuminations. For hand gesture recognition, we propose a novel recognition method using saliency and similarity learning via Ranking-SVM. In our method, a metric-based spatially weighted similarity is proposed, which fuses the similarity of saliency features and spatial coordinates to jointly determine the similarity of two images. In order to learn the parameters, we employ Ranking-SVM to learn similarity. Finally, two experiments are conducted on Cambridge Hand Gesture Data Set. The first experiment results show that saliency-based hand detection is real-time and can deal with illumination changes. The second experiment shows that our proposed hand gesture recognition is promising on large data set with different illuminations.