Intelligibility of temporally interrupted speech

Abstract

The variations in intelligibility of a temporally interrupted connected discourse passage that result from changes in both switching conditions and message rate were studied. Six normal-hearing listeners estimated the intelligibility of ten, 15-sec speech samples presented monaurally at 36 interruption conditions (12 interruption rates × 3 speech-time fractions) for each of two speaking rates. Results indicated that for a given speech-time fraction, changes in intelligibility as a function of interruption rate depend on the precise temporal correspondence of switching and message parameters. Specifically: (1) for slow interruption rates, where on-time and off-time are greater than average word duration, temporal features of the message are relatively unimportant and intelligibility varies directly with speech-time fraction; (2) intelligibility functions will begin to recover from minimum at an interruption rate where off-time is equal to, or just shorter than, average word duration; (3) differences in performance along the recovery portions of the functions for the two message rates are negligible when intelligibility is expressed in terms of percent off-time relative to average word duration. It is apparent that a trading relation exists between speech-time fraction and interruption rate. Within certain limits, decrements resulting from lowering the speech-time fraction can be overcome by increasing the switching rate.

Psychometric functions for sentence recognition in sinusoidally amplitude-modulated noises

“…According to this hypothesis, when the duration of individual envelope peaks in the masker is longer than the duration of crucial speech features, these speech features might be presented at a far more disadvantageous TMR than the nominal TMR during the masker peaks. Other studies of the recognition of periodically interrupted speech also demonstrated non-monotonicity in performance as the interruption rate increases (e.g., Miller and Licklider, 1950;Powers and Speaks, 1973;Nelson and Jin, 2004;Jin and Nelson, 2010;Shafiro et al, 2011;Bhargava and Başkent, 2012). These studies provided further evidence that important speech features are lost by gating the speech on and off at a low rate (either using a noise masker or interruptions).…”

Recognition Performance of Interrupted Monosyllabic Words: The Effects of Ten Interruption Locations

“…iller (1947) and Miller and Licklider (1950) provided the early work on the effects that silent or noise interruptions have on speech recognition. Subsequently, a multitude of investigations using a variety of interruption paradigms have been reported that involve both listeners with normal hearing for pure tones and listeners with sensorineural hearing loss (e.g., Cherry, 1953;Huggins, 1964;Dirks et al, 1969;Wilson and Carhart, 1969;Powers and Speaks, 1973;Howard-Jones and Rosen, 1993;Wang and Humes, 2010;Wilson et al, 2010;Kidd and Humes, 2012). The study reported here is the third in a series of investigations from our laboratory examining the effects that the location of the interruption pattern has on word recognition performance (Wilson 2014;Wilson and Irish, 2015).…”

The Intelligibility of Interrupted Speech: Cochlear Implant Users and Normal Hearing Listeners

“…These two conditions thus result in good intelligibility. At intermediate rates (2–4 Hz), almost every or every other word is obliterated, causing a drop in intelligibility (Powers and Speaks 1973 ; Huggins 1975 ; Nelson and Jin 2004 ; Shafiro et al 2011 ). Miller and Licklider ( 1950 ) and Wang and Humes ( 2010 ) reported that the intelligibility of single words was lowest at the rate of around 4 Hz because entire phonemes were eliminated at this rate.…”