摘要 |
An input unit (10) receives an audio signal. A marking unit (100) adds, to the audio signal, a pitch mark, which indicates a representative point of a primitive period, in every primitive period. An extracting unit (210) windows a part of the audio signal and extracts a partial waveform that is the audio waveform of the windowed part. A calculating unit (220) frequency analyzes the partial waveform to calculate a frequency spectrum. An estimating unit (230) generates an artificial waveform, which is a waveform in accordance with the intervals between the pitch marks, for every harmonic component, and estimates, from the generated artificial waveforms, harmonic spectrum characteristics indicating the characteristics of the frequency spectrums of the harmonic components. A separating unit (240) uses the harmonic spectrum characteristics and the frequency spectrum of the partial waveform to separate the partial waveform into periodic components the audio sources of which are periodic vocal-cord vibrations and into non-periodic components generated by non-periodic audio sources other than the vocal-cord vibrations. |
申请人 |
KABUSHIKI KAISHA TOSHIBA;MORITA, MASAHIRO;LATORRE, JAVIER;KAGOSHIMA, TAKEHIKO |
发明人 |
MORITA, MASAHIRO;LATORRE, JAVIER;KAGOSHIMA, TAKEHIKO |