Predicting audible distortions in speech
Harry N. Levitt · The Journal of the Acoustical Society of America · 1990
The reproduction of speech by public address systems, or other means of amplification, is typically accompanied by electronically generated distortions. These distortions can be both linear (e.g., modification of frequency response) or nonlinear (e.g., peak clipping) and may also include the addition of background noise. Nonlinear distortion is of particular concern since it can be highly disruptive of speech quality. A generalized index for predicting distortion in amplified speech signals has been developed. It is similar in concept to the articulation index and can be used to predict the audibility of nonspeech components in the speech signal, such as noise and/or nonlinear distortion. Linear distortions are treated as changes in the speech spectrum. An earlier version of this index was used to predict binaural release from masking and gain in intelligibility [H. Levitt and L. R. Rabiner, J. Acoust. Soc. Am. 42, 820–829 (1967)].