| 摘要 |
PURPOSE: To enhance a tone quality of a sound whose spectral change is violent by windowing a partial section by a small analyzing window, finding an autocorrelation function from its signal, and performing linear prediction analysis on a synthesized autocorrelation function to which this is added. CONSTITUTION: In a partial sectional extraction 12, continuous signals (a partial section) having various lengths are extracted from various places on an input sound signals 11 of an analytical section. In a windowing 13, a window coefficient is multiplied on respective signals. Next, autocorrelation analysis 14 is performed, and an autocorrelation function is found. Processing of this windowing 13 and the autocorrelation analysis 14 is performed on plural partial sections in an analytical section, and respective autocorrelation functions are weighted, and are added together, and a synthesized autocorrelation function is obtained. Next, linear prediction analysis 16 by an autocorrelation method is performed on the synthesized autocorrelation function, and a liner predictive coefficient is found. The liner predictive coefficient on which information of a local part of the analytical section is more reflected is obtained by changing weight. |
