Study of Objectionable Sound Recognition Based on Histogram Features and SVM

It is important to protect children from harmful effects of objectionable materials, such as pornography, which are now prevalent on the Internet. In this paper, a new method from the feature porno-sounds recognition point of view is proposed to detect adult video sequences automatically which serves as a complementary approach to the recognition method from image's point of view. To the special of erotic sound, a new set of features based on histograms and contours are introduced for training and classification. Experimental comparisons with popular features for audio classification are presented and discussed. A recognition method is designed based on the standard SVM classification framework due to its extraordinary generalization performance. These efforts result in superior performance that 89.17% recall rate and 10.78% false positive rate are achieved for experiments on real data from the Internet.