Increasing the intelligibility of speech through multiple phonemic restorations

Abstract

Outside ofthe laboratory, listening conditions are often less than ideal, and when attending to sounds from a particular source, portions are often obliterated by extraneous noises. However, listeners possess rather elegant reconstructive mechanisms. Restoration can be complete, so that missing segments are indistinguishable from those actually present and the listener is unaware that the signal is fragmented. This phenomenon, called temporal induction (TD, has been studied extensively with nonverbal signals and to a lesser extent with speech. Earlier studies have demonstrated that TI can produce illusory continuity spanning gaps of a few hundred milliseconds when portions ofa signal are replaced by a louder sound capable of masking the signal were it actually present. The present study employed various types of speech signals with periodic gaps and measured the effects upon intelligibility produced by filling these gaps with noises. Enhancement of intelligibility through multiple phonemic restoration occurred when the acoustic requirements for TI were met and when sufficient contextual information was available in the remaining speech fragments. It appears that phonemic restoration is a specialized form of TI that uses linguistic skills for the reconstruction of obliterated speech.In everyday life, portions of speech and other acoustic signals are often obliterated by extraneous transient noises such as coughs, slamming doors, and traffic sounds. However, we possess sophisticated mechanisms that can restore the segments of these signals that have been obliterated by noise. The rule governing restoration is simple, and its consequences manifold. Restoration can occur when an extraneous sound is capable of masking a signal and when there is sufficient contextual information to establish the identity of an obliterated fragment. This restoration, which has been called temporal induction (TI), is complete, so that when appropriate conditions are met, listeners cannot distinguish restored sounds from those actually present.A review of the literature has indicated that there are three basic types ofTI (Warren, 1984). The simplest form of restoration has been called homophonic continuity. If three levels of the same sound (e.g., 60-, 70-, and 80-dB levels of a noise each lasting 300 msec) are presented in succession and repeated without pauses, the 6O-dB level appears to be on continuously with the other levels heard as intermittent pulsed additions. Apparent continuity of the faintest sound occurs despite the fact that this level is actually present for less than half of the time. Homophonic continuity is a subtractive process. Thus, when Portions of this work were supported by grants from the National Institutes of Health (DC00208) and the Air Force Office of Scientific Research (88-0320). These experiments were outlined in a paper presented at the 119th meeting of the Acoustical Society of America, with a briefabstract appearing in the Ioumal of the Acoustical Society (VoL 87, Supp1. 1,1990, S71). The authors are w...

Phonemic restoration effect reversed in a reverberant room

“…Each mean was based on data pooled between the two testing Fig. 2(a)], the intelligibility of the interrupted speech was higher when the interrupted gaps are filled with noise (filled circles) relative to silence (open circles) [F (1, 11) ¼ 104.79, p < 0.0001, M se ¼ 12.943], which is consistent with previous demonstrations of the phonemic restoration effect (Bashford et al, 1992;Shinn-Cunningham and Wang, 2007). In addition, intelligibility was found to increase as the interruption rate increased [F(2, 22) ¼ 89.90, p < 0.0001, M se ¼ 9.44], presumably due to the shorter duration of the interruption intervals at higher interruption rates.…”

“…Because past studies of phonemic restoration effects with similar silent intervals resulted in decreased speech intelligibility (Bashford et al, 1992;Shinn-Cunningham and Wang, 2007), the notion that interruptions to the at-the-ear signal are in some sense implausible under natural listening conditions may contribute to the phonemic restoration effect. It is also important to note that the noise interruptions appear to be unaffected by the room manipulation, as evidenced by the lack of a significant difference in intelligibility performance between the two room conditions (Pre-vs Post-Room) [t(35) ¼ 1.732, M se ¼ 1.33126, p ¼ 0.092].…”

“…The effect has been shown to produce objectively higher intelligibility (Bashford et al, 1992;Shinn-Cunningham and Wang, 2007) and can be enhanced when multimodal input is made available to the listener (Shahin and Miller, 2009). Overall, the phonemic restoration effect may be considered as a special case of interrupted speech perception (Miller and Licklider, 1950) in which top-down compensation factors such as context, expectation, and linguistic rules are thought to account for relatively intact word recognition even when significant portions of the speech are missing (Bashford et al, 1992). One aspect of the phonemic restoration effect that has not been previously considered is the role of the listening environment, which can under certain situations cause significant changes to the speech and interruption signals reaching the ears.…”

Auditory induction: Reciprocal changes in alternating sounds

“…A series of studies has been undertaken to test the validity of this suggestion. These studies have found that, in keeping with the spectral requirements for nonverbal auditory induction, the greatest increase in apparent continuity of narrowband speech passages interrupted periodically by narrowband noise occurred when the center frequencies of the speech and noise bands were the same (Bashford & Warren, 1979, 1987, and also that the intelligibility of narrowband sentences was greatest when the interrupting noise bands matched the center frequency of the speech (Bashford et al, 1992). Bregman (1990) has suggested that phonemic restoration represents a "schema-driven stream segregation" that is fundamentally different from the auditory induction of nonverbal sounds such as tones and noises.…”

Comparing human and automatic speech recognition in a perceptual restoration experiment

“…Perceptual restoration studies indicate that the restoration effect depends on acoustic cues that relate to the additive noise (Powers and Wilcox, 1977;Samuel, 1981;Warren et al, 1997) and context cues that relate to the speech material (Warren and Sherman, 1974;Verschuure and Brocaar, 1983;Bashford et al, 1992). The current section describes the stimuli and evaluation metric used in the present work.…”