| 摘要 |
PROBLEM TO BE SOLVED: To change a speaking speed in every frame unit, and to precisely find a pitch cycle from a voice wave form with a small amount of processing to change the speaking speed. SOLUTION: A voice wave form is divided into frames (step s1), a vocal interval within the frame is extracted (step s2), the vocal interval is determined as a processing object voice interval, the voice wave form in an interval prefixed with a starting point of the processing object voice interval on a time base as an origin is set as a copy interval (step s3), and the voice wave form of the copy interval is parallelly displaced on the time base by a prescribed sampling point at a time (step s4). A correlation between the voice wave form of the copy interval in each parallel displacement point and the voice wave form of the processing object voice interval is found in every parallel displacement, and the parallel displacement point having the maximum correlation is found for the next dividing point (step s5-s8). Duplication-processing or thinning-processing is conducted among the respective found dividing points. A fragmentary part is omitted from a speaking speed changing processing object. The processed frames are connected to be brought into the initial order. |
