Feature-based Classication of Compressed Audio Stream Encoder
AN Kai-shen · Radio and communications technology · 2014
Today's digital audio coding algorithms use sophisticated models to maximize the encoding rate while minimize the audible distortion. As a result of this complexity,different implementations of one encoding standard tend to produce varying output streams for the same uncompressed input data. The paper presents a method to distinguish between the encoding schemes used to compress MPEG1 Layer-3(MP3) audio files on the basis of the statistical features that can be extracted from the compressed streams.The method adopts a SVM machine learning classier to determine the most likely encoder from a vector of 13 features. The experimental results show that the accuracy can be up to 95%,and the accuracy of the identification results can be further improved by adding new characteristics from the code flow. So the method can be considered as a generic tool to increase the overall reliability of steganalysis of MP3 les.