| 摘要 |
PURPOSE:To obtain a natural speech when multiplexed data generated by encoding fluctuations of the higher harmonic frequency of a speech are sent out and decoded. CONSTITUTION:A fundamental frequency extractor 11 extracts the fundamental frequency of an input speech signal and a higher harmonic frequency detector 12 detects higher harmonic frequencies which are an integral multiple and an odd-numbered multiple of the fundamental frequency included in the input speech signal. A spectrum amplitude estimating unit 13 estimates spectrum amplitudes at the fundamental frequency of the input speech signal and the higher harmonic frequencies from the higher harmonic frequency detector 12 that are the integral multiple and odd-numbered multiple. The parameters of the fundamental frequency extracted by the fundamental frequency extractor 11, the integral multiple and odd multiple higher harmonic frequencies from the higher harmonic frequency detector 12, and the spectrum amplitude from the spectrum amplitude estimating unit 13 are quantized by a multiplexer 14 and the multiplexed data (sent code) are sent out. |
