 |
    |
Achieving Shorter Search Times in Voice Conversion Using Interactive Evolution
Yuji Sato
Faculty of Computer and Information Sciences, Hosei University, 3-7-2 Kajino-cho, Koganei-shi, Tokyo 184-8584, Japanyuji@k.hosei.ac.jp
Abstract. We have already proposed using evolutionary computation to adjust the voice quality conversion parameters, and we have reported that this approach produces results that are not only closer to the desired target than the results of parameter adjustment based on designer experience or trial and error, but which also have relatively little sound quality degradation. In this paper we propose improved techniques for the generation of initial entities and genetic manipulation in order to reducing the workload associated with human evaluation in interactive evolution. We perform voice quality conversion experiments both on natural speech recorded with a microphone and on synthetic speech generated from text data. As a result, we confirm that the proposed improvements make it possible to perform voice quality conversion more efficiently than when using the technique proposed earlier.
LNCS 3103, p. 1328 f.
lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2004