CONTINUOUS PROBABILISTIC TRANSFORM FOR VOICE CONVERSION

Voice conversion, as considered in this paper, is defined as modifying the speech signal of one speaker (source speaker) so that it sounds a s if it had been pronounced by a different speaker (target speaker), O ur contribution includes the design of a new methodology for represent ing the relationship between two sets of spectral envelopes, The propo sed method is based on the use of a Gaussian mixture model of the sour ce speaker spectral envelopes, The conversion itself is represented by a continuous parametric function which takes into account the probabi listic classification provided by the mixture model. The parameters of the conversion function are estimated by least squares optimization o n the training data, This conversion method is implemented in the cont ext of the HNM (harmonic + noise model) system, which allows high-qual ity modifications of speech signals, Compared to earlier methods based on vector quantization, the proposed conversion scheme results in a m uch better match between the converted envelopes and the target envelo pes, Evaluation by objective tests and formal listening tests shows th at the proposed transform greatly improves the quality and naturalness of the converted speech signals compared with previous proposed conve rsion methods.