Publication detail

Cepstral Vocal Tract Modelling for Fixed-Point Digital Signal Processor Implementation

SMÉKAL, Z., VÍCH, R.

Original Title

Cepstral Vocal Tract Modelling for Fixed-Point Digital Signal Processor Implementation

English Title

Cepstral Vocal Tract Modelling for Fixed-Point Digital Signal Processor Implementation

Type

conference paper

Language

en

Original Abstract

The choice of a given type of model is given by the technical possibilities of the facility for speech synthesis, and also by the complexity of the program and its time demand while being implemented. New families of fixed-point digital signal processors usually have dual Harvard architecture. This architecture best supports the “multiply and accumulate” instructions. In the paper, the properties of canonic and non-canonic models of the zero-pole vocal tract model are described from the viewpoint of implementation on dual Harvard architecture of the fixed-point digital signal processor.

English abstract

The choice of a given type of model is given by the technical possibilities of the facility for speech synthesis, and also by the complexity of the program and its time demand while being implemented. New families of fixed-point digital signal processors usually have dual Harvard architecture. This architecture best supports the “multiply and accumulate” instructions. In the paper, the properties of canonic and non-canonic models of the zero-pole vocal tract model are described from the viewpoint of implementation on dual Harvard architecture of the fixed-point digital signal processor.

Keywords

Cepstral Vocal tract model, formant and antiformants

Released

09.10.2000

Location

Ilmenau

ISBN

3-8007-2564-9

Book

Proc. of the Inter. Conf. Konvens 2000

Edition number

první

Pages from

135

Pages to

139

Pages count

5

BibTex


@inproceedings{BUT17864,
  author="Zdeněk {Smékal} and Robert {Vích}",
  title="Cepstral Vocal Tract Modelling for Fixed-Point Digital Signal Processor Implementation",
  annote="The choice of a given type of model is given by the technical possibilities of the facility for speech synthesis, and also by the complexity of the program and its time demand while being implemented. New families of fixed-point digital signal processors usually have dual Harvard architecture. This architecture best supports the “multiply and accumulate” instructions. In the paper, the properties of canonic and non-canonic models of the zero-pole vocal tract model are described from the viewpoint of implementation on dual Harvard architecture of the fixed-point digital signal processor.",
  booktitle="Proc. of the Inter. Conf. Konvens 2000",
  chapter="17864",
  year="2000",
  month="october",
  pages="135",
  type="conference paper"
}