Reduced Search Space Frame Alignment Based on Kullback-Leibler Divergence for Voice Conversion
- Jan 2013
- NOLISP
- Mons, Belgium
- Link to paper
Authors
A. S. Shahrebabaki, J. Amini, H. Sheikhzadeh, M. Ghorbandoost, N. Faraji
Abstract
A new text independent voice conversion based on Kullback-Leibler divergence (KLD) is proposed. This method only uses acoustic information and does not require any linguistic or phonetic information. The KLD is used to find reliable correspondence between the source and target GMM clusters and to reduce the search space for alignment of source and target frames. Subjective evaluation results show that the proposed method can achieve the same performance as parallel voice conversion methods.