This thesis addresses the problem of improving the results of statistical parametric speech synthesis using deep learning algorithms. The subject has become more important in recent times due to the increasing presence of artificial voices in several devices and applications. In these, there is a need to refine the results...