Prosodic Model under the Effects of Variability in Speaking Rate

[[abstract]]語音速率是自發性語音系統裡最重要的因素之一，因為與朗讀式的語音相比，自發性語音的說話速率是經常變化的。研究顯示對應音段的聲音特性會受說話速率變化的影響，例如，頻譜與音長都可能改變。一些研究也發現說話速率的改變也會影響基頻的整體基準和範圍。當改變說話速率時，也會引起韻律片語的變化，例如韻律邊界的改變。要將這些變化納入定量模型中有其困難。因此，致力於研究語音速率對聲調、音長、句調及韻律結構的細微作用的韻律模組仍然比較少。而這些是我們在這項三年計畫中想要努力的主題。在第一年，我們將建立不同語音速率的語料庫。然後，在第二年，我們將焦點集中在速度對音段音長結構的影響。在最後一年，我們將建立考慮語音速率變化的基頻模型。成果將有助於增進語音合成的品質和語音識別的辨識率。[[abstract]]Speaking rate is one of the most important factors in spontaneous speech systems, because variability in speaking rate may be often observed in spontaneous speech than in read speech. Studies have shown that the acoustic properties corresponding to phonetic segments of speech are influenced by variability of speaking rate, for example, the spectral patterns and the durations may be changed. Some research also found that changes in speech rate have effects on overall level and range of fundamental frequency (F0). While changing speaking rates, it also causes variations in prosodic phrasing, like prosodic boundaries. It is difficult to incorporate the findings into a quantitative model. Therefore, there has been rather few prosodic model devoted to investigating detailed effects of speech rate modification on the realization of individual pitch accents, duration, intonation, and prosodic structures. This is the topic of this three-year project. In the first year, a speech corpus with different speaking rate is built. Then, in the second year, the focus is on the effect of speaking rate on segmental duration structure. In the last year, fundamental frequency model considering the change of speech rate is built. The achievement is useful for improving the quality of speech synthesis and the recognition rate of speech recognition.[[note]]NSC99-2221-E327-04