| 摘要 |
PURPOSE:To obtain a synthetic voice waveform being similar to an original voice waveform by adopting a multi-pulse driving type voice encoding method by which a section partitioned by a zero intersection for shifting a low frequency component signal of an input sound signal from a negative level to a positive level or from a positive level to a negative level becomes a frame. CONSTITUTION:A low-pass filter 4 extracts a signal of a low frequency part in a sound signal and sends it out to a fixed level counter 5. The fixed level counter 5 extracts a zero intersection for shifting a level of the low frequency signal from negative to positive or from positive to negative, and sends out section information delimited by the zero intersection to a representative sound source extracting part 11. The representative sound source extracting part 11 takes a correlation by an input sound signal, and spectrum envelope information from a linear prediction analyzer 2, and generates a pulse signal by a position and amplitude calculated by an algorithm of a multi-pulse driving type encoding method at every section determined by section information from the fixed level counter 5. In such a way, a synthetic voice waveform being similar to an original voice waveform is obtained. |
