Adaptive voice-quality control based on one-to-many eigenvoice conversion