Classification of audio codecs based on coding characteristics

We propose a fast, simple and new method for identification of audio codecs that does not require decoding of coded audio data. The method uses chaotic and randomness features of coded audio to build models associated with different codecs. The most important features of the method are operating on just a few kilobytes of data sampled randomly from audio file and assuming no knowledge on encoding structure of a codec. These two features make it fast and simple. Experiments are performed on both singly coded and transcoded audio samples to measure accuracy of the technique.