Efficient Encoding and Decoding of Speech.
Final rept. 1 Nov 79-31Dec 80,
BOLT BERANEK AND NEWMAN INC CAMBRIDGE MA
Pagination or Media Count:
This report concludes our work for the past year on efficient coding and decoding of speech. During the past year, we have developed and optimized an adaptive predictive coding APC system for high quality speech encoding at 16 kbs. In our research, we have examined many issues which we described in this report. The system uses the noise-feedback APC configuration, chosen because of the computational efficiencies in implementing spectral noise shaping. The many-level, non-uniform quantizer is derived as a piece-wise linearization of a mu-law non-linear quantizer. The encoding scheme is an entropy coder using a self-synchronizing variable-length code which is matched to the quantization. A simple pole-zero spectral noise shaping is adapted as a function of the input speech spectrum. System parameters are perceptually optimized via listening experiments. As a conclusion to the algorithm development, different implementation strategies were studied as well as methods for decreasing computational complexity. The computational requirements of the algorithm were related to the features of several technologies. Author
- Human Factors Engineering and Man Machine Systems