Chapter 1: Invited Addresses and Tutorials on Signals, Coding,
  Systems and Intelligent Techniques

Title: Speech Period Detection Using Neural Network Classification

Author(s): Andrej Vrábel, Gregor Rozinaj

Address: Slovak University of Technology, FEEIT, Dept. of Telecommunications Ilkovicova 3, 812 19 Bratislava | Slovak University of Technology, FEEIT, Dept. of Telecommunications Ilkovicova 3, 812 19 Bratislava

Reference: 12th International Workshop on Systems, Signals and Image Processing pp. 147 - 150

Abstract/Summary: The implementation of the Slovak speech synthesizer, based on the TD PSOLA algorithm in the EMU database system [1], has proven the strong relation between accurate pitch period estimation and the quality of synthesized speech. Although the problem of pitch detection is well known for a long time, there has not been developed a method yet, which works well in general. In this work the neural network is used as a statistical classifier, generalizing outputs of various pitch detection algorithms. Further it describes its usage in the text-to-speech syn-thesizer and discusses achieved results.

Order a copy of this article