Effects of wide dynamic-range compression on the perceived clarity of individual musical instruments

Sara Miay Kim Madsen, Michael A. Stone, Martin F. McKinney, Kelly Fitz, Brian C. J. Moore

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

The effects of wide-dynamic-range compression (WDRC) on the ability of hearing-impaired subjects to hear out individual instruments or voices (called “sources”) in a mixture were explored. On each trial, the subjects were asked to judge the relative clarity of the target in two repetitions of the same music excerpt (mixture of sources) that were processed in different ways. The stimuli were processed via a five-channel simulated WDRC hearing aid, using individual insertion gains and compression ratios recommended by the CAM2 fitting procedure. Both fast- and slow-acting WDRC and a condition with linear amplification and frequency-response shaping were used. To investigate the role of cross-modulation (the partial correlation of the envelopes of different sources caused by the time-varying gain applied by the compressor), conditions were included where the sounds from different sources were compressed before being added together and where the sounds were added together before being compressed. The results showed no effect of cross-modulation, lower clarity with WDRC than with linear amplification, and no significant overall effect of compression speed, although some subjects consistently rated clarity as greater with slow compression. The deleterious effect of WDRC may be related to changes in temporal-envelope shape or reduced spectral contrast produced by WDRC.
Original languageEnglish
JournalJournal of the Acoustical Society of America
Volume137
Issue number4
Pages (from-to)1867–1876
ISSN0001-4966
DOIs
Publication statusPublished - 2015
Externally publishedYes

Fingerprint Dive into the research topics of 'Effects of wide dynamic-range compression on the perceived clarity of individual musical instruments'. Together they form a unique fingerprint.

Cite this