摘要 |
PURPOSE: To prevent offensive sound to the ear from being generated by correcting multiple pitch errors in the case where three parameters of pitch frequency, voiced-unvoiced decision of respective harmonics of voice signal and spectral amplitude information are coded. CONSTITUTION: Using a pitch frequency estimation section 101, an improved cepstrum factor calculating section 103 and a judging section on each harmonic wave about the voiced-unvoiced decision, estimated pitch frequency of input voice signals, improved cepstrum factors and the judgment of each harmonic wave about it is voiced or not are respectively calculated. Further when estimation error of pitch frequency occurs, the pitch frequency correcting section 102 whose inputs are input voice signals, estimated pitch frequency, improved cepstrum factors detects and corrects multiple pitch errors by utilizing the fact that the spectrum of the synthesis signals calculated in the pitch frequency correcting section 102 differs from that of the input signals. |