Transparent Low Bit Rate Below 64 kb/s Audio Compression Using Wavelets
Gwan U Park, Jae Ho Jeong · 1999
In this paper, we propose a low bit rate, high fidelity audio codec using wavelet, which is well-matched to the psychoacoustic model. We decompose the audio signal into 29 mon-uniform subbands approximated to the critical bands using the tree structured wavelet packet decomposition. When this transform is performed, we select a wavelet which can be best applied to the information obtained from the psychoacoustic model. In the quantization procedure of wavelet coefficients, we select the scale factors of each frame from gradually broader range as moving to the higher frequency region, and quantize these scale factors according to their pre-searched dynamic ranges to minimize the number of bits needed in expression of each audio frame under certain distortion level. Using the proposed method, we get the compact disc(CD) quality of audio at 52-64 kbps(kilo bits per second) per channel and in comparison with MPEG-1 LayerⅡ, the coding quality of the proposed codec is greatly superior to that of MPEG-1 LayerⅡ at 48 kbps.