A LOW-DELAY CELP CODER FOR THE CCITT 16 KB S SPEECH CODING STANDARD

This paper presents a low-delay code-excited linear prediction (LD-CELP) speech coder, which is expected to be standardized in 1992 as a CCITT G Series Recommendation for universal applications of speech coding at 16 kb/s. The coder achieves a one-way coding delay of less than 2 ms by making both the LPC predictor and the excitation gain backward-adaptive, and by using a small excitation vector size of five samples. The pitch predictor in conventional CELP coders is not used due to its sensitivity to channel errors, and the resulting performance loss is compensated for by increasing the LPC predictor order from 10 to 50. The excitation gain is updated by a 10th-order adaptive linear predictor based on the logarithmic gains of previously quantized and scaled excitation vectors. The LPC predictor and the gain predictor are updated by performing LPC analysis on previously coded speech and previous log-gain sequence, respectively, with the autocorrelation coefficients calculated by a novel hybrid windowing method. The excitation codebook is closed-loop optimized and its index is Gray-coded for better robustness to channel errors. The perceptual weighting filter is designed to optimize speech quality for three asynchronous tandem encodings without significantly degrading the performance for a single encoding. An adaptive postfilter is used at the decoder to improve coder performance, the tandeming performance in particular. This coder has been implemented in real-time using the AT & T WE(R) DSP32C digital signal processor. The official CCITT laboratory tests revealed that the speech quality of this 16 kb/s LD-CELP coder is either equivalent to or better than that of the CCITT G.721 standard 32 kb/s ADPCM coder for almost all conditions tested. This paper describes the LD-CELP algorithm, its implementation on the DSP32C for CCITT testing, and some performance results from these tests.