A ANN Based High Quality Method for Voice Conversion

In this paper, we describe a novel conversion method for voice conversion (VC). Artificial Neural Network (ANN) model is employed for performing joint spectrum and pitch conversion between speakers. The conventional method converts spectral parameters and pitch independently. Those separate transformations lead to an unsatisfactory speech quality. The main reason maybe that F₀ sequences are usually converted by a simply linear function. To overcome this problem, we apply joint parameters for train and conversion. A comparative study of voice conversion with ANN and Gaussian Mixture Model (GMM) is conducted. Experimental results indicate that the performance of VC can be dramatically improved by the proposed method in view of both subjective evaluation and objective measurement.