| 摘要 |
PURPOSE:To efficiently predict a pitch signal which varies gradually by the voice encoding device which utilizes the pitch structure of a voice. CONSTITUTION:The spectrum prediction parameter of an input voice signal is calculated by a linear prediction unit 102, and a spectrum prediction unit 104 and an auditory weighting filter are formed. A pitch prediction unit 104 for a voice outputs an estimation range of pitch length and pitch predicted waveforms as to all estimated values of a variation quantity. A code book outputs all code words and a spectrum prediction unit inputs said two composite waveforms and outputs a reproduction signal. The difference signal between the input voice signal and reproduction signal is inputted to the auditory weighting filter and the energy of the difference signal after auditory weighting is calculated to set the parameters of the pitch prediction unit 104 and code book to the best parameter by an analyzing method using composition so that said energy becomes minimum. |
