Transparent bitrate estimation for perceptual audio coding

In this paper, the relationship between the features of the input audio and its corresponding transparent bitrate is studied. This relationship can be used to pre-estimate the transparent bitrate required for an input audio file, so that this bitrate can be used for encoding and the transparent quality can be achieved with the minimum compressed file size. Two features are selected and tested. Firstly, the perceptual feature is extracted from the test sequences using MPEG-4 advanced audio coding. The results of the signal-to-mask ratio and the transparent bitrate are compared. Secondly, the relationship between the music genre of the audio and its transparent bitrate is explored.