Mp3 bit rate quality detection through frequency spectrum analysis

Abstract

The proliferation of the lossy MP3 format as the standard for audio transferred over the internet is of great concern to audiophiles, those who deeply care about good audio quality. Typically, the bit rate of an MP3 file is used as a relative measure of audio quality, however, this check fails if the audio has been transcoded from a lower bit rate to a higher bit rate. In this paper, we propose a method of detecting the original lower bit rate of a given audio file by analyzing its high frequency spectrum. Using a Support Vector Machine classifier and five classes of bit rates (CBR 128 kbps, 192 kbps, 256 kbps, 320 kbps, and VBR-0), our algorithm returned an average success rate of 97% in correctly detecting the original compressed bit rate of an audio file in the absence of any coding format knowledge other than the audio signal itself. Furthermore, our algorithm also detected the original lower bit rates of 320 kbps MP3s transcoded from 128 kbps and 192 kbps sources with a success rate of 99%.