Subband coding with vector quantization

Vector quantization (VQ) has previously [1] been examined as a technique to enhance performance in subband coding of speech at 9.6 kbit/s. The model in [1] serves as the basis for the new model presented here. Investigations of the statistics of the subband gains and normalized subband signals, and VQ of these, are presented. The investigations revealed some new aspects of using VQ in subband coding. Based on these observations a new model was developed and simulated. Compared to a scalar quantization (SQ) model, our VQ model proved to offer a gain of 2 dB in segmental signal-to-noise ratio (SNRseg) and a distinct improvement in perceptual speech quality.