The intelligibility of whitened and amplitude compressed speech

Abstract

Substantial research has demonstrated that amplitude compression can be imposed on speech without degrading intelligibility. Nonetheless, there remain aspects of compression which should be investigated. Consequently, we completed an experiment which investigated (1) the effects of whitening speech prior to amplitude compression, (2) the use of a moderate compression ratio (3:1) versus one that imposed a more drastic reduction on the dynamic range of speech (10:1), and (3) the effects of amplitude compression on speech intelligibility when more than one speaker was present. A competing speech message was electrically mixed with CNC discrimination words and the composite signal whitened and amplitude compressed. Discrimination functions were then obtained for the CNC material by varying the intensity of the competition to yield signal-to-competition ratios of 0, +8, and +12 dB. Persons with normal hearing served as subjects. The results indicate that whitening the speech and the use of the 10:1 compression ratio did not alter intelligibility. Amplitude compression did consistently lower speech intelligibility when more than one talker was present, although this reduction was not substantial and was statistically significant only for the +8 dB signal-to-competition ratio. [Research completed at Auditory Research Laboratories, Northwestern University and supported by Veterans Administration.]