| 摘要 |
PURPOSE:To obtain a synthesis signal of excellent sound quality at a low bit rate by finding the amplitude and position of a processed signal which is obtained by processing a voice signal through a filter from the variable and constant of an equalizing filter which minimize the error electric power between the synthesis signal and a voice signal. CONSTITUTION:The voice signal x(n) inputted from an input terminal 100 is divided with every frame stored with the sound signal on a buffer memory 110. Here, an LPC analysis part 120 finds a spectrum parameter representing the spectrum envelope of the voice signal of a frame, thereby obtaining, for example, a linear prediction coefficient. Then a multipulse generation part 130 generates a specific number of multipulses and a synthetic filter 150 inputs the linear prediction coefficient from the analysis part 120 to find a synthetic voice. Then the amplitude and phase of the processed signal obtained by processing the voice signal through the filter 180 can be found from the variable and constant of the equalizing filter 180 which minimizes the error electric power between the voice input x(n) and synthetic signal. |
