Feature fusion and prototype learning algorithm based smoking shot recognition

This paper present a novel human smoking shot recognition method based on feature fusion and the prototype learning algorithm. We first extract the color feature and the texture of the skin patch and the cigarette patch. Then we use the subspace learning algorithm to fuse the features. The prototype learning algorithm based on metric learning is used to classify the fusion features. The experimental results show the effectiveness of our method. © 2011 Springer-Verlag.