Voice conversion using subspace method and Gaussian mixture model

摘要

In voice conversion, if phonetic information and speaker information included in speech data are separated, similar voice to a target speaker will be obtained by exchanging speaker information. In this paper, we propose a voice conversion method from an original speaker to a target speaker by a subspace method. At first, a speaker space and a phonetic space are separately constructed for each speaker and the speaker spaces are exchanged between the original speaker and the target speaker. The proposed method was experimentally compared with the conversion method based on GMM. As a result, the proposed method was shown to be superior to GMM in terms of subjective evaluation.