Subband audio coding using a perceptually hybrid vector-scalar quantization

This paper presents a novel perceptually hybrid vector-scalar quantization scheme for high quality subband audio coding at low bit rates. A QMF filter bank is used to decompose, in the frequency domain, the input audio signal into constant width subbands. Each subband is then quantized using either a multi-stage vector quantizer or a scalar quantizer. In order to obtain a high quality audio, a dynamic bit-allocation algorithm is used to select the best quantizer for each subband based on the outputs of a psychoacoustic model and the total quantization bits available. The performance of the proposed coder is assessed from segmental signal-to-noise ratios (SEGSNR) and informal listening tests, which show that the proposed audio coder with the perceptually hybrid vector-scalar quantization scheme is superior in both objective and subjective performances compared with a subband audio coder with full scalar quantization