Dithered subband coding with spectral subtraction

In this paper, we present a combination technique of a subtractive dithered subband coding and spectral subtraction for improving the perceptual quality of coded audio at low bit rates. The quantization noise is rendered into the signal-independent white noise by dithering quantization process. This noise is estimated and removed by spectral subtraction. Moreover, the residual noise reduction technique is presented and applied to the system. The results show the effective improvement of the proposed method over the conventional one in terms of SNR and human listening tests. The proposed system can be combined with other existing or future coding methods such as perceptual coding to improve their performance at low bit rates.