Performance evaluation of adaptive quantizers for a 16-kbits/s sub-band coder

Objective and subjective measures were obtained for a 4-band, 16-kbit/s sub-band coder utilizing adaptive quantizers with fixed and dynamic bit allocations. Objective measures indicated that block quantization (block companded PCM with a Gaussian probability density function, BCPCM-G) with dynamic bit allocation is slightly better than an adaptive differential PCM with a variable predictor (ADPCM-V) and fixed bit allocation, but outperforms ADPCM with a fixed predictor and fixed bit allocation (ADPCM-F) by 4-5 dB. Subjective measures, however, indicated that ADPCM-V has about 6-dB advantage over BCPCM-G. A comparison between ADPCM-V and ADPCM-F indicated that adaptive prediction provides a 4-5 dB objective gain and 16-dB subjective gain over fixed prediction. Further, all listeners preferred ADPCM-V over BCPCM-G and ADPCM-F, and 85% preferred BCPCM-G over ADPCM-F.