| 摘要 |
PROBLEM TO BE SOLVED: To obtain technology for improved compression (encoding) into a voice signal and an audio signal quantizing a signal of voice information on the basis of a 1st specific signal and generating an encoded signal on the basis of the quantized signal. SOLUTION: A short-period remainder is removed by an LPC predictive error filter 20 from an input voice signal (s). The obtained LPC predictive remainder signal (d) contains some long-period remainder owing to the pitch periodicity in a voiced sound. This long-period remainder is removed by a pitch predicting processor 30. After pitch prediction, a final predictive remainder signal (e) is transformed to a frequency range by a transforming processor 40 which performs fast Fourier transformation. Adaptive bit allocation is performed in order to allocate bits to a predictive remainder FFT coefficient by a remainder quantizer 60 according to perceptual importance determined by an auditory model quantizer control processor 50. |
