Speech Steganalysis Based on Multi-classifier Combination

Abstract

Aiming at the problems of low detection rate and high feature dimension in the current steganalysis method based on support vector machine (SVM), a speech steganalysis scheme based on multi-classifier combination is proposed. The steganalysis features are first input into the classifier set, and the first type of prediction results are obtained. Then, the output of the first set of classifiers is taken as the input of the other set of classifiers to get another result.Finally, the two results are processed by the fusion algorithm to obtain a detection result.Based on the feature of pitch delay second-order difference, the proposed steganalysis scheme is evaluated comprehensively, and compared with the existing steganalysis method based on SVM.Experimental results show that this method has better performance than existing steganalysis methods based on SVM.