Publication detail

CEPSTRAL SPEECH SYNTHESIS OPTIMISED FOR DUAL HARVARD ARCHITECTURE OF DSP

SMÉKAL, Y., VÍCH, R.

Original Title

CEPSTRAL SPEECH SYNTHESIS OPTIMISED FOR DUAL HARVARD ARCHITECTURE OF DSP

English Title

CEPSTRAL SPEECH SYNTHESIS OPTIMISED FOR DUAL HARVARD ARCHITECTURE OF DSP

Type

conference paper

Language

en

Original Abstract

The paper describes the realisation algorithm of the cepstral vocal tract model with poles and zeros, which is optimised from the viewpoint of implementation on a fixed-point digital signal processor with dual Harvard architecture. Cepstral coefficients for speech synthesis are in fact coefficients of type FIR digital filter, which is introduced in the first canonic form of type IIR digital filter. The order of the two filters depends on approximation accuracy and on sampling frequency

English abstract

The paper describes the realisation algorithm of the cepstral vocal tract model with poles and zeros, which is optimised from the viewpoint of implementation on a fixed-point digital signal processor with dual Harvard architecture. Cepstral coefficients for speech synthesis are in fact coefficients of type FIR digital filter, which is introduced in the first canonic form of type IIR digital filter. The order of the two filters depends on approximation accuracy and on sampling frequency

Keywords

Cepstral vocal tract model, digital signal processor

Released

05.05.2000

Publisher

IEEE

Location

Mexico

ISBN

968-36-7762-2

Book

Propc. of the Inter. Conference ICT 2000

Edition number

první

Pages from

244

Pages to

248

Pages count

5

Documents

BibTex


@inproceedings{BUT17863,
  author="Zdeněk {Smékal} and Robert {Vích}",
  title="CEPSTRAL SPEECH SYNTHESIS OPTIMISED FOR DUAL HARVARD ARCHITECTURE OF DSP",
  annote="The paper describes the realisation algorithm of  the cepstral vocal tract model with poles and zeros, which is optimised from the viewpoint of implementation on a fixed-point digital signal processor with dual Harvard architecture. Cepstral coefficients for speech synthesis are in fact coefficients of type FIR digital filter, which is introduced in the first canonic form of type IIR digital filter. The order of the two filters depends on approximation accuracy and on sampling frequency",
  address="IEEE",
  booktitle="Propc. of the Inter. Conference ICT 2000",
  chapter="17863",
  institution="IEEE",
  year="2000",
  month="may",
  pages="244",
  publisher="IEEE",
  type="conference paper"
}