Encoding of PARCOR coefficients using fuzzy rules
Francesco Beritelli, S. Casale, Marco Antonio Russo · 2002
We present a simple and efficient new method for encoding LPC parameters based on soft computing (SC). More specifically the authors, using a novel training technique that is able to extract fuzzy knowledge, propose a method of transformation from reflection coefficients into a new set of parameters related to the fuzzy inferential process. Performance evaluations in terms of spectral distortion and comparisons of perceptive quality in listening tests, show that this new approach is appropriate for low bit-rate speech coding as it achieves a good trade-off between quality and bit rate with a very low computational load.