A single chip vector quantization processor for real-time speech coding

Vector quantization (VQ) is discussed as a viable technique for low-bit rate coding of speech and image data. Special-purpose processors are required for real-time implementations of a VQ algorithm due to its high computational complexity. A single chip vector quantizer for real-time speech coding application in contrast to earlier off-the-shelf and multichip implementations is described. The architectural considerations and the implementation details of the modules with their estimated performance are presented. The real-time VQ speech encoder can use a codebook up to a size of 400 with a vector dimension of 4. The codebook size can be further increased with a hierarchical VQ configuration. Larger codebooks are also possible with an increase in vector dimension.<>