Hardware Implementation of a 16 KBPS Subband Coder Using Vector Quantization

New Image

Although vector quantization (VQ) is a promising method of reducing speech coder bit rates, the degree of computation required typically makes VQ too expensive for real-time applications. However, new signal processing devices, notably the floating point AT&T dsp32 [1] make real-time VQ coders practical. In addition, many coders incorporating VQ have large coding delays due to the size (in msec) of the speech frame being processed. In contrast, the VQ algorithm implemented here operates inside a subband analysis/synthesis framework and uses short data frames, q q q which results in a very short (12.5 msec) communication delay. Such a short delay is essential if the coder is to support a natural mode of conversation.