1. Home
    Item seperator
  3. Item seperator
    A Robust Low-Delay CELP Speech Coder at 16 kb/s

A Robust Low-Delay CELP Speech Coder at 16 kb/s

New Image

Following the standardization of 32 kb/s ADPCM, CCITT is now planning to standardize a 16 kb/s speech coder by 1991. In June 1988, CCITT established a stringent set of performance objectives and requirements and started seeking submission of candidate 16 kb/s coding algorithms. Based on these requirements, we have developed a robust, low-delay, and high-quality 16 kb/s speech coder which has been submitted by AT&T as a candidate algorithm. This coder is basically a code-excited Linear Prediction (CELP) coder [1] with a backward-adaptive predictor and a robust backward gain-adaptive vector quantizer [2] for excitation coding. Gray-code VQ index assignment was used to enhance the coder's robustness against channel errors [3], [4]. This low-Delay CELP (LD-CELP) coder uses a block size (or vector dimension) of only 5 samples so as to meet what is perhaps the most demanding objective of the CCITT 16 kb/s standard - a one-way coding delay of less than 2 ms.