Simulation of the effect of threshold elevation and loudness recruitment combined with reduced frequency selectivity on the intelligibility of speech in noise Yoshito Nejimea) and Brian C. J. Moore Department of Experimental Psychology, University of Cambridge, Downing Street, Cambridge CB2 3EB, England ~Received 28 August 1996 revised 19 March 1997 accepted 19 March 1997! The effect of loudness recruitment and threshold elevation together with reduced frequency selectivity have been simulated to examine the combined effect of the two major consequences of cochlear hearing loss on the intelligibility of speech in speech-shaped noise. In experiment 1, four conditions were simulated: a moderate flat loss with auditory filters broadened by a factor of three ~B3R2! a moderate-to-severe sloping loss with auditory filters broadened by a constant factor of three ~B3RX! and these conditions with linear amplification applied prior to the simulation processing ~B3R21, B3RX1!. For conditions B3R2 and B3RX, performance was markedly worse than for a control condition ~normal hearing, condition R1! tested in a previous study. For conditions B3R21 and B3RX1, linear amplification improved performance considerably. However, performance remained below that for condition R1 by between 5% and 19%. In experiment 2 the broadening of the auditory filters was made more realistic by making it a function of the absolute threshold at the center frequency of the auditory filter. Three different hearing losses were simulated: a moderate-to-severe sloping loss with variable broadening of the auditory filters ~BXRX! the same moderate-to-severe sloping loss with linear amplification ~BXRX1! and the same broadening of the auditory filters but without the simulation of loudness recruitment and threshold elevation ~BX!. For condition BXRX, performance was markedly worse than in condition R1, while performance in condition BX was somewhat worse than for condition R1. For condition BXRX1, linear amplification according to the NAL procedure improved performance to a large extent but it remained worse than for condition R1. The results are consistent with previous evidence indicating that only part of the decrease of performance produced by actual cochlear hearing loss can be compensated by conventional linear hearing aids. �� 1997 Acoustical Society of America. @S0001-4966~97!02707-0# PACS numbers: 43.71.Ky, 43.71.Gv @WS# INTRODUCTION People with cochlear hearing loss often have difficulty in understanding speech, especially when the speech is ac- companied by interfering sounds. However, there is contro- versy over whether this is primarily due to reduced audibility of the speech signal ~Humes et al., 1987 Zurek and Del- horne, 1987! or to abnormalities in the perceptual analysis of the signal, even when it is well above the absolute threshold ~Plomp, 1978, 1986 Dreschler and Plomp, 1980, 1985 Glasberg and Moore, 1989!. It has been difficult to decide what factors are most important, since the elevation of abso- lute threshold that is the most obvious sign of hearing loss is usually associated with a variety of suprathreshold changes, such as reduced dynamic range ~loudness recruitment! ~Fowler, 1936 Steinberg and Gardner, 1937 Villchur, 1974!, reduced temporal resolution for narrow-band fluctu- ating stimuli ~Fitzgibbons and Wightman, 1982 Tyler et al., 1982 Glasberg et al., 1987 Glasberg and Moore, 1992!, re- duced ability to discriminate frequency differences in pure and complex tones ~Glasberg and Moore, 1989 Moore and Peters, 1992!, and reduced frequency selectivity ~Pick et al., 1977 Florentine et al., 1980 Tyler et al., 1984 Glasberg and Moore, 1986 Moore and Glasberg, 1986 Tyler, 1986!. For a review, see Moore ~1995!. Some of the methods for addressing this problem were reviewed in earlier publications ~Baer and Moore, 1993, 1994 Moore and Glasberg, 1993 Moore et al., 1995!. These publications argued that an appropriate method was to pro- cess sounds so as to simulate the effect of one specific aspect of hearing impairment, and to present those sounds to nor- mally hearing subjects for evaluation. Provided the simula- tion is accurate, this makes it possible to study the effect of that aspect in isolation. This has been applied to loudness recruitment ~Villchur, 1974 Glasberg and Moore, 1992 Moore and Glasberg, 1993 Duchnowski and Zurek, 1995 Moore et al., 1995! and to reduced frequency selectivity ~Villchur, 1977 Summers and Al-Dabbagh, 1982 Celmer and Bienvenue, 1987 Gagne �� and Erber, 1987 Moore et al., 1992 ter Keurs et al., 1992 Baer and Moore, 1993, 1994!. Previous studies from our group have used a digital simulation of threshold elevation combined with loudness recruitment to examine the intelligibility of speech in quiet, against a background of a single competing talker ~Moore and Glasberg, 1993! and against a background of noise a!Present address: Multimedia Systems Research Dept., Central Research Laboratory, Hitachi Ltd., Kokubunji, Tokyo 185, Japan. 603 603 J. Acoust. Soc. Am. 102 (1), July 1997 0001-4966/97/102(1)/603/13/$10.00 �� 1997 Acoustical Society of America

~Moore et al., 1995!. Three different types of cochlear hear- ing loss were simulated, including moderate and severe flat losses and a sloping loss. For speech in quiet, simulation of hearing loss produced a reduction in the ability to understand low-level speech. However, speech at sufficiently high levels was highly intelligible in all conditions. Linear amplification according to the National Acoustics Laboratory ~NAL! pre- scription ~Byrne and Dillon, 1986!, applied prior to the simu- lation, gave high intelligibility for speech at normal conver- sational levels. For speech presented at a fixed input level of 65 dB SPL, against a background of a single competing talker, simulation of hearing loss produced substantial decre- ments in performance. Linear amplification according to the NAL prescription improved performance markedly for the conditions simulating flat losses, but was less effective for the condition simulating a sloping loss. It was concluded that the loss of intelligibility for speech, in the presence of a single competing talker, is only partly compensated by linear amplification of the type typically used in hearing aids. For speech presented at a fixed input level of 65 dB SPL, against a background of speech-shaped noise, simula- tion of threshold elevation and loudness recruitment showed much smaller decrements in performance than those obtained using a single competing talker. When linear amplification according to the NAL prescription was applied, performance improved markedly for conditions simulating a moderate hearing loss, and did not differ significantly from that for normal hearing ~using unprocessed stimuli!. It was con- cluded that loudness recruitment and threshold elevation have larger effects for a fluctuating background sound than for a steady background, and linear amplification is more effective in the latter case. In an earlier study using a simulation of reduced fre- quency selectivity, Baer and Moore ~1993! examined the in- telligibility of speech in quiet and against a background of speech-shaped noise. They simulated reduced frequency se- lectivity using spectral smearing designed to evoke, in nor- mally hearing subjects, auditory excitation patterns resem- bling those that would be evoked by the same stimuli in subjects with broadened auditory filters. It should be noted that this simulation does not mimic the changes in the time patterns at the outputs of the auditory filters produced by broadening of the filters. Essentially, the simulation mimics the consequences of reduced frequency selectivity for place coding, but not for time coding ~Baer and Moore, 1993, 1994!. The signal processing was based on smearing of the short-term power spectrum, using broadened auditory filters defined in the frequency domain. The overlap-add technique ~Allen, 1977! was used to perform the processing. They simulated the reduced frequency selectivity associated with several different amounts of broadening of the auditory fil- ters: symmetrical broadening by factors of 3 and 6, and sev- eral degrees of asymmetrical broadening using different broadening factors for the upper branch and for the lower branch of the auditory filters. The results showed that spec- tral smearing using symmetrical broadening by a factor of 3 or 6 had little effect on the intelligibility of speech in quiet, but had a substantial effect on the intelligibility of speech in noise at speech-to-noise ratios of 0 and 23 dB. In a subse- quent study, they examined the intelligibility of speech against a background of a single competing talker using the same broadening conditions ~Baer and Moore, 1994!. The results showed that reduced frequency selectivity had slightly larger effects for a fluctuating background than for a steady background. The present study simulates the effect of threshold el- evation and loudness recruitment in combination with re- duced frequency selectivity on the ability to understand speech in a background of noise. Hereafter, we refer to this simulation as the ������combined simulation.������ The results of the earlier studies suggested that both loudness recruitment with threshold elevation and reduced frequency selectivity are sufficient to produce a substantial decrease in the ability to understand speech in the presence of background sounds. People with cochlear hearing loss usually have both loudness recruitment and reduced frequency selectivity. It remains un- clear how the two effects combine, for example whether they are additive, whether one is dominant over the other, or whether the combined effect is greater than a simple sum of the individual effects, which might happen if each removes cues that are left unaffected by the other. The two simula- tions certainly change the perception of speech in noise dif- ferently the simulation of loudness recruitment with thresh- old elevation changes the audibility of speech stimuli and their perceived dynamics, while the spectral smearing makes the speech appear ������fuzzy������ ~less clear! and subjectively harder to separate from the background. The specific questions addressed in this study are as fol- lows: ~1! How much is the intelligibility of speech affected by simulating threshold elevation and loudness recruitment in combination with reduced frequency selectivity? This was assessed by using normally hearing subjects listen- ing to signals processed via the combined simulation and comparing the results with those of earlier studies. ~2! To what extent can the deleterious effects of the com- bined factors be counteracted by linear amplification with appropriate frequency response shaping? To answer this question, speech intelligibility was measured using the combined simulation, but including frequency- dependent linear amplification according to the revised NAL procedure ~Byrne and Dillon, 1986!. I. EXPERIMENT 1 A. Method of simulation 1. Rationale Our method for combining the simulations is based on the concatenation of the two simulations which have been described in detail in earlier publications ~Baer and Moore, 1993 Moore and Glasberg, 1993!. A brief description of each simulation, and of how they were modified in order to combine them, is given next. We assume that, in an impaired ear, reduced frequency selectivity can be characterized as a broadening of the audi- tory filters. This effect can be simulated by smearing the 604 604 J. Acoust. Soc. Am., Vol. 102, No. 1, July 1997 Y. Nejime and B. C. J. Moore: Simulation of hearing loss