Adaptive noise feedback coding of speech in systems with nonadaptive predictors
Abstract
Two solutions for the adaptive feedback filter evolve from minimizing a frequency weighted output noise-to-signal ratio, one quite general and the other from assuming white quantizing noise. The general solution suggests a sequential adaptation, and subsequently a sequential lattice filter is used in simulations because it guarantees stability. The white noise solution is used to predict the theoretical variation of the signal to noise ratio with the amount of noise shaping. Computer simulations show good agreement with the theoretical model. Most importantly, both solutions exhibit subjective improvements over (DPCM) coders, which are special cases of adaptive noise feedback coders. Hence DPCM coders can be improved, without increasing the bit rate, by using adaptive feedback filters. The simulations also show these improvements depend on the number of quantizer levels.
- Publication:
-
Ph.D. Thesis
- Pub Date:
- March 1980
- Bibcode:
- 1980PhDT........72B
- Keywords:
-
- Adaptive Filters;
- Coders;
- Noise Reduction;
- Speech;
- Computerized Simulation;
- Noise Spectra;
- Signal Processing;
- Signal To Noise Ratios;
- Communications and Radar