Intra-lingual and cross-lingual voice conversion using harmonic plus stochastic models