Detection and removal of “smacking” artefacts from lector speech records

Downloads

Authors

  • Marek Jaskuła Szczecin University of Technology, Faculty of Electrical Engineering, 26 Kwietnia 10, 71-126 Szczecin, Poland
  • Hanna Perużyńska Szczecin University of Technology, Faculty of Electrical Engineering, 26 Kwietnia 10, 71-126 Szczecin, Poland

Abstract

In the paper, analysis of “smacking” artefacts, introduced to the speech signal by a lector, was shown. A working algorithm to reduce amount of smacks in the waveform was presented. The algorithm performs detection of distortions in the time domain using differentiation of the signal. The removal routine is based on Dicrete Wavelet Transform (DWT) and Inverse Discrete Wavelet Transform (IDWT).

Keywords:

speech enhancement, wavelet transform, impulse noise.