Publication detail

Optimized Speech Synthesis in Digital Signal Processing using the Cepstral Model of Vocal Tract

SMÉKAL, Z., VONDRA, M.

Original Title

Optimized Speech Synthesis in Digital Signal Processing using the Cepstral Model of Vocal Tract

English Title

Optimized Speech Synthesis in Digital Signal Processing using the Cepstral Model of Vocal Tract

Type

conference paper

Language

en

Original Abstract

In the paper, a parametric speech synthesizer, realized on a Motorola DSP56307 digital signal processor, is described, which forms part of a vocoder or a single-purpose system of text-to-speech synthesis. An advantage of this parametric model in TTS systems is the possibility of simple control of prosodic parameters (melody, speech speed and stress, microprosody, time division, etc.). Another possibility consists in totally replacing the speaker. If, for example, a male-voice inventory has been prepared, the male voice can be changed to the female or child voice by altering the parameters of the model and excitation. This parametric synthesizer forms the terminal part of the whole TTS system that should be implemented on the digital signal processor for automatic TTS synthesis.

English abstract

In the paper, a parametric speech synthesizer, realized on a Motorola DSP56307 digital signal processor, is described, which forms part of a vocoder or a single-purpose system of text-to-speech synthesis. An advantage of this parametric model in TTS systems is the possibility of simple control of prosodic parameters (melody, speech speed and stress, microprosody, time division, etc.). Another possibility consists in totally replacing the speaker. If, for example, a male-voice inventory has been prepared, the male voice can be changed to the female or child voice by altering the parameters of the model and excitation. This parametric synthesizer forms the terminal part of the whole TTS system that should be implemented on the digital signal processor for automatic TTS synthesis.

RIV year

2003

Released

31.07.2003

Location

Ilmenau

ISBN

1619-4098

Book

Proceedings of the 48th International Scientific Colloquium

Edition

Neuveden

Edition number

první

Pages from

121

Pages to

122

Pages count

2

BibTex


@inproceedings{BUT8144,
  author="Zdeněk {Smékal} and Martin {Vondra}",
  title="Optimized Speech Synthesis in Digital Signal Processing using the Cepstral Model of Vocal Tract",
  annote="In the paper, a parametric speech synthesizer, realized on a Motorola DSP56307 digital signal processor, is described, which forms part of a vocoder or a single-purpose system of text-to-speech synthesis. An advantage of this parametric model in TTS systems is the possibility of simple control of prosodic parameters (melody, speech speed and stress, microprosody, time division, etc.). Another possibility consists in totally replacing the speaker. If, for example, a male-voice inventory has been prepared, the male voice can be changed to the female or child voice by altering the parameters of the model and excitation. This parametric synthesizer forms the terminal part of the whole TTS system that should be implemented on the digital signal processor for automatic TTS synthesis.",
  booktitle="Proceedings of the 48th International Scientific Colloquium",
  chapter="8144",
  edition="Neuveden",
  year="2003",
  month="july",
  pages="121",
  type="conference paper"
}