摘要 |
PROBLEM TO BE SOLVED: To perform vector quantization with hardly recognizable degradation of sound quality even with a small bit number in a vector quantization method used in a voice encoder or the like. SOLUTION: A target extraction part 12 for calculating a quantization target by using an input vector 11, a quantization part 13 for encoding the quantization target and obtaining a decoding vector, a distortion calculation part 14 for calculating the distortion of an input vector and the decoding vector, and a comparison part 15 for determining a code through comparison of the distortion, transmitting it and updating the contents of a decoding vector storage part 17 are provided. Thus, a weighting search is performed so as not to feel the degradation in terms of auditory sense and the distortion is made hardly audible in terms of the auditory sense even at a low bit rate. |