Dividing the input speech 5, the spectral envelope information and the sound source to improve the encoding characteristics in coding a sound source in the frame-by-frame to a plurality of sound source position and sound source gain. In the source encoder 11 for encoding the sound source into a plurality of sound source position and sound source gain, and provide a doctor gain calculating section 40 for calculating a pseudo gain assigned for each sound source position candidates, and the pulse position search section 41 is such using a doctor gain to determine a plurality of the sound source position and gain encoder 12, the coding gain of the sound source using the sound source position is determined.
展开▼