摘要 |
PURPOSE: To effectively perform quantization by differentiating LSP parameters to a part being auditively important and a part being not important so much and applying processings suitable to respective parts. CONSTITUTION: A voice signal 11 stored in a buffer 12 is divided into two subframes by a subframe divider 13. A first subframe among divided frames is subjected to an LPC analysis by an LPC analyzer 14 to be outputted to a divider 16. In the divider 16. LSP parameters are divided into the part being auditively important and the part being not important so much. Besides, a second subframe among subframes divided by the subframe divider 13 is subjected to an LPC analysis by an LPC analyzer 15 to be outputted to a divider 17. In the divider 17, the same processing as that in the divider 16 is performed. Then, quantization information are made not to be transmitted at the time of renewaling of parameters as to parameters whose importance are auditively low and, on the other hand, correct quantization information are made to be transmitted every renewal time as to parameters whose importance arc auditively high. |