抄録
E-026
Noise Robust Voice Conversion using GA-based Informative Feature
Noise Robust Voice Conversion using GA-based Informative Feature
◎澤田耕平・田上陽嗣・田村哲嗣・竹原正矩・速水 悟（岐阜大）
We propose noise robust voice conversion (VC) using GA-based informative feature (GIF). This method is realized by adding an extraction process of GIF to a conventional VC. GIF is proposed as a feature that can be applied for pattern recognition, and on speech recognition, it could improve recognition accuracy in noise environment. We evaluated the performance of the conventional VC and the proposed VC in various noise environments. Experimental results indicate that the performance of the proposed VC was better than that of the conventional VC.