An 800 bps adaptive vector quantization vocoder using a perceptual distance measure

This paper presents an adaptive vector quantization algorithm using a perceptually motivated distance measure. The adaptation is accomplished by continuous alteration of the the template set to match the current speaker and environment. The distance measure approximates auditory perception to optimize the template usage. Adaptation protocols requiring no additional channel bandwidth have been devised. The system achieves an intelligibility score (DRT) of 86.8 percent on an 800 bps channel. Combinations of this system with spectral frame-fill produce systems with data rates as low as 500 bps at somewhat reduced intelligibility.