HELIUM SPEECH TRANSLATION USING HOMOMORPHIC TECHNIQUES.

Abstract

The application of advanced digital-processing techniques has great potential for systems that transmit voice or utilize information coded in the form of speech. The report employs a digital process that offers a new approach for general use in speech synthesis, and is an application of homomorphic methods to the problem of correcting the distorted speech of talkers in pressurized helium-oxygen atmospheres. The vocal-tract impulse response of such speech was extracted by the homomorphic deconvolution technique, and its frequency components were moved downward in frequency according to correction formulas given in a study by Gerstman (1966). Both linear and nonlinear frequency corrections were used. Speech samples taken at 800-foot pressure depth in a 96 percent helium, 4 percent oxygen atmosphere were processed in this way, using a digital simulation of Oppenheim's (1969) analysis-synthesis system. Results indicate considerable promise for the technique as a tool for further study of helium speech, and perhaps as a future on-line translation method. (Author)

Document Details

Document Type
Technical Report
Publication Date
Jul 31, 1970
Accession Number
AD0712345

Entities

People

  • Roy F. Quick Jr

Organizations

  • Air Force Cambridge Research Laboratories

Tags

DTIC Thesaurus Topics

  • Atmospheres
  • Frequency
  • Translations

Readers

  • Image Processing and Computer Vision.
  • Speech Processing/Speech Recognition.
  • Underwater engineering and Marine Technology.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference