| 摘要 |
PURPOSE: To quickly execute the highly accurate identification of a signal sort by finding out the number of zero crosses in a talking signal, analyzing the number of zero crosses and identifying the sort of the talking signal based upon the analytical result. CONSTITUTION: A number-of-zero crosses analyzing part 120 finds out the number Z of zero crosses of a talking signal (a) in each fixed frame period. An identification condition decision part 130 receives maximum power Wmax , minimum power Wmin , maximum power frequency F, band width B, and the time variation quantity D of a frequency spectrum from a spectrum analyzing part 100, receives a 0-order self- correlation coefficient R0 , primary self-correlation coefficient R1 and secondary self- correlation coefficient R2 which are total power and the time variation quantity H of these self-correlation coefficients from a self-correlation analyzing part 110 and receives the number Z of zero crosses in a signal from the analyzing part 120. The decision part 130 executes the detection of a voiced section, spectrum analysis variation analysis, the analysis of the number of zero crosses, and the judgement of continuity by using these identification parameters to identify the sort of the talking signal (a). Consequently the sort of the talking signal (a) can be highly accurately identified and its identification result can quickly be obtained. |
