A Study on Voice Conversion System in Multiple Languages to Produce Affective Speech

点击翻译按钮获取中文摘要

研究の目的:Recently, the text to speech (TTS) synthesis system have already been utilized in many applications such as navigation systems and robotic assistants. However, the synthesized speech from conventional TTS system have suffered the unnaturalness problem. In order to improve the naturalness of synthesized speech from TTS system, this research focus on affective voice conversion研究実施計画:The emotional voice conversion system with degree controllable in dimensional space was proposed using rule-based technique. The prosody related acoustic features such as fundamental frequency, power envelope and duration were controlled by Fujisaki model, target prediction model and interpolation. Perception evaluations show that sad and joy emotion can be converted well. Results were summerized in Elsevier's journal Speech Communication.