| 摘要 |
PROBLEM TO BE SOLVED: To provide a method and a device for speech speed conversion in which high degree of comprehension is obtained and real time speech conversion is realized employing a simple process. SOLUTION: A measurement is made for a length Tn of the noise segment between a phrase that is a voice segment and a next phrase. When the length Tn exceeds a time threshold Tth, the phrase following the above phrase is discriminated to be a beginning of sentence phrase. When the length Tn is shorter than the value Tth, the phrase that follow are discriminated to be middle of sentence phrases. Then, for the beginning of sentence phrase, a two-steps A pattern expansion is conducted from the leading phrase. For the middle of the sentences, a three-steps B pattern expansion is conducted. |
