DEVICE AND METHOD FOR CELP TYPE VOICE DECODING

摘要

PROBLEM TO BE SOLVED: To obtain a synthesized voice of higher quality by providing a diffusion pattern storage a selection part which selects one of stored diffusion patterns and a pulse vector diffusion part which convolutes the diffusion pattern selected by the selection part as well as a pulse vector, etc. SOLUTION: A diffusion pattern storage and a selection part of a noise code book 304 reads diffusion patterns corresponding to combined numbers of diffusion pulses, channel by channel, and outputs them to a pulse vector diffusion part 313. A pulse vector generation part 312 generates the pulse vectors corresponding to the combined numbers of pulse vectors as many as the channels and outputs them to a pulse vector diffusion part 313. Then the pulse vector diffusion part 313 generates a diffusion vector from the diffusion patterns received from the selection part 311 and the pulse vectors received from the pulse vector generation part 312 through a specific convolution operation and outputs the diffusion vector to a diffusion vector addition part 314.