Speech recognition based on fuzzy vector quantization and fuzzy logic
Liusheng Liu, Zhijian Li, Bingxue Shi · 2002
This paper investigates the use of fuzzy segment matrix vector quantization (FSMVQ) and fuzzy logic recognizer (FLR) for speech recognition. Unlike the standard VQ, the fuzzy vector quantization (FVQ) makes a soft decision, generates a output vector whose components represent the degree to which each codeword matches the input vector. Because of this soft decision, quantization error can be reduced and some of recognition error can be remedied. The FSMVQ and FLR based recognition system requires a lesser amount of training data and have good generalization for untrained data.