F0 Modeling In Hmm-Based Speech Synthesis System Using Deep Belief Network

In recent years multilayer perceptrons (MLPs) with many hidden layers Deep Neural Network (DNN) has performed surprisingly well in many speech tasks, i.e. speech recognition, speaker verification, speech synthesis etc. Although in the context of F0 modeling these techniques has not been exploited properly. In this paper, Deep Belief Network (DBN), a class of DNN family has been employed and applied to model the F0 contour of synthesized speech which was generated by HMM-based speech synthesis system. The experiment was done on Bengali language. Several DBN-DNN architectures ranging from four to seven hidden layers and up to 200 hidden units per hidden layer was presented and evaluated. The results were compared against clustering tree techniques popularly found in statistical parametric speech synthesis. We show that from textual inputs DBN-DNN learns a high level structure which in turn improves F0 contour in terms of objective and subjective tests.