Voice Transformation Using Two-Level Dynamic Warping and Neural Networks

Voice transformation, for example, from a male speaker to female speaker, is achieved here using two-level dynamic warping algorithm in conjunction with an artificial neural network. An outer process which temporally aligns blocks of speech (dynamic time warp, DTW) invokes inner process, spectrally based on magnitude spectra frequency DFW). The mapping function produced by warp used move spectral information source target speaker. Artifacts arising this amplitude are reduced reconstructing phase information. Information obtained train network produce input data. performance the compared Mel-Cepstral Distortion (MCD) previous voice transformation research, and it shown perform better than other methods, their reported MCD scores.