Product of Experts for Statistical Parametric Speech Synthesis

Citations Over TimeTop 10% of 2011 papers

Abstract

Multiple acoustic models are often combined in statistical parametric speech synthesis. Both linear and non-linear functions of an observation sequence are used as features to be modeled. This paper shows that this combination of multiple acoustic models can be expressed as a product of experts (PoE); the likelihoods from the models are scaled, multiplied together, and then normalized. Normally these models are individually trained and only combined at the synthesis stage. This paper discusses a more consistent PoE framework where the models are jointly trained. A training algorithm for PoEs based on linear feature functions and Gaussian experts is derived by generalizing the training algorithm for trajectory HMMs. However for non-linear feature functions or non-Gaussian experts this is not possible, so a scheme based on contrastive divergence learning is described. Experimental results show that the PoE framework provides both a mathematically elegant way to train multiple acoustic models jointly and significant improvements in the quality of the synthesized speech.

Related Papers

• → Focal points in coordinated divergence(2006)21 cited
• → A Generalized $$\chi ^2$$ Divergence for Multisource Information Fusion(2021)1 cited
• → A THREE-DIMENSIONAL STUDY OF THE HORIZONTAL VELOCITY DIVERGENCE(1955)11 cited
• → Indiscriminate Separation, under the Same Environment, a Cause of Divergence(1890)9 cited
• → Determinants of Economic Divergence Among Accelerating Societies(2013)