Phoneme Adjustment in Enhanced Speech

Abstract

A system was developed to enhance the quality and intelligibility of speech which had been pre-processed by a Speech Enhancement Unit (SEU) at RADC Griffis AFB. The system processes the speech in the frequency domain. A Hamming window with 50% overlap was applied to the time waveform and a 512-point Discrete Fourier Transform (DFT) was computed. The amplitude spectrum of voiced regions was smoothed in order to reduce the effects of noise. Frequencies above 2.5 KHz were enhanced as they had been attenuated by SEU. Harmonics of the glottal pitch frequency of voiced speech and peaks of unvoiced speech were selected to further reduce the noise effects. The harmonics of the glottal frequency. The two neighboring frequency points were checked and the maximum of those three points was selected instead of the exact glottal harmonic. Speech was reconstructed using amplitude phase, and frequency of the harmonics/peaks selected. The reconstructed speech had much better quality and improved SNR. SPIRE (Speech and phonetics Interactive Research Environment) and ILS (interactive Laboratory System) software packages were used for visual analysis of the amplitude spectrum. The system was implemented in FORTRAN 77 on a VAX 11/ 780 machine. Keywords: Speech processing, Theses.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 1989
Accession Number
ADA206357

Entities

People

  • Nadeem A. Bashir

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Abstracts
  • Air Force
  • Classification
  • Computer Programs
  • Discrete Fourier Transforms
  • Electrical Engineering
  • Engineering
  • Environment
  • Frequency Bands
  • Frequency Domain
  • Intelligibility
  • Language
  • Operating Systems
  • Recognition
  • Signal Processing
  • Speech
  • Waveforms

Fields of Study

  • Engineering

Readers

  • Acoustics.
  • Approximation Theory.
  • Speech Processing/Speech Recognition.