A new subband perceptual audio coder using CELP

This paper presents an audio coding system which uses filter banks to decompose, in the frequency domain, the audio signal into constant width subbands. A specific compression is applied in each subband. This compression is achieved by means of CELP coders. In order to obtain a high audio quality, psychoacoustic models allocate dynamically the number of bits needed in each subband. A particular care has been taken for the elaboration of the filter banks in order to limit the delay and the computational cost of the system. We have implemented several filter banks and tested their influence on the perceptual quality of the output audio signal. Finally, we show that our proposed coder is capable of delivering excellent audio signal quality at bit rates of 50-60kbit/s.