| 摘要 |
The present invention relates to a method for extracting scale and duration from a human voice. The method for extracting the scale and duration according to the present invention comprises; a step of sensing the human voice for a period of time; a step of collecting a pitch group of the human voice which is sensed according to a preset section for a predetermined period of time; a step of calculating an utterance property coefficient which shows whether pitch values are higher than a natural scale or not by comparing the pitch values of the collected pitch groups with pitch values of the natural scale based on a twelve-tone scale; a step of firstly correcting the pitch value of the pitch groups as the same off-net using the utterance property coefficient; a step of selecting a representative pitch value using multiple voting methods in the pitch groups which are corrected firstly; and a step of secondly correcting the pitch values of the pitch groups which are corrected firstly based on a difference between the representative pitch value and an average value of the pitch values of the pitch groups which hare corrected firstly. |
