摘要 |
PURPOSE: A text-to-speech system and a method thereof are provided to obtain a parameter through analysis of recoded sound and to prevent the degradation of sound quality. CONSTITUTION: A database(120) stores a sound source about a vocal tract area, an LSP(Line Spectrum Pairs) vector, and a sound source vector index. A sound source information obtaining unit(130) obtains sound source information by using the sound source vector index. A decoding unit(140) extracts a sound source signal. A synthesis filter unit(150) synthesizes the sound source into one sound. |