Effects of pre-processing filters on a wavelet packet-based algorithm to identify speech transients

Daniel M. Rasetshwane, J. Robert Boston, Ching Chung Li, John D. Durrant

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Scopus citations

Abstract

Speech transients have been shown to be important cues for identifying speech sounds, and amplification of transients can improve the intelligibility of speech in noise. We have developed a time-frequency approach to identify transients that uses a pre-processing filter, but optimal filter parameters are difficult to determine due to the large number of possibilities. This paper describes the use of the Articulation Index (AI) to evaluate the effects of different high-pass filters on our algorithm. The best filter was found to depend on signal-to-noise ratio (SNR). AI results should be interpreted with caution, but they appear to provide a reasonable approach to selecting a pre-processing filter.

Original languageEnglish (US)
Title of host publication2007 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA
Pages179-182
Number of pages4
DOIs
StatePublished - 2007
Event2007 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA - New Paltz, NY, United States
Duration: Oct 21 2007Oct 24 2007

Publication series

NameIEEE Workshop on Applications of Signal Processing to Audio and Acoustics

Other

Other2007 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, WASPAA
CountryUnited States
CityNew Paltz, NY
Period10/21/0710/24/07

ASJC Scopus subject areas

  • Signal Processing

Fingerprint Dive into the research topics of 'Effects of pre-processing filters on a wavelet packet-based algorithm to identify speech transients'. Together they form a unique fingerprint.

Cite this