Voice conversion based on Non-negative Matrix Factorization in noisy environments

This paper presents a voice conversion (VC) technique for noisy environments. We prepared parallel exemplars (dictionary) that consist of the source and target exemplars, which have the same texts uttered by the source and target speakers. The input source signal is decomposed into the source exemplars, noise exemplars obtained from the input signal, and their weights (activities). Then, the converted signal is obtained by calculating the linear combination of the target exemplars and the weights which are calculated using the source exemplars. In the proposed method, a Gaussian Mixture Model (GMM) -based conversion method is also applied to the feature vectors generated by the sparse coding in order to compensate a mismatch between the weights of source and target exemplars. The effectiveness of this method was confirmed by comparing its effectiveness with that of a conventional method.