Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 Mar 1.
Published in final edited form as: Ear Hear. 2016 Mar-Apr;37(2):144–152. doi: 10.1097/AUD.0000000000000229

Effects of Reverberation and Compression on Consonant Identification in Individuals with Hearing Impairment

Paul N Reinhart 1, Pamela E Souza 1,2, Nirmal K Srinivasan 3,4, Frederick J Gallun 3,4
PMCID: PMC4767560  NIHMSID: NIHMS719907  PMID: 26462171

Abstract

Objectives

Hearing aids are frequently used in reverberant environments; however, relatively little is known about how reverberation affects the processing of signals by modern hearing aid algorithms. The purpose of this study was to investigate the acoustic and behavioral effects of reverberation and wide-dynamic range compression (WDRC) in hearing aids on consonant identification for individuals with hearing impairment.

Design

Twenty-three listeners with mild to moderate sloping sensorineural hearing loss were tested monaurally under varying degrees of reverberation and WDRC conditions. Listeners identified consonants embedded within vowel-consonant-vowel nonsense syllables. Stimuli were processed to simulate a range of realistic reverberation times and WDRC release times using virtual acoustic simulations. Additionally, the effects of these processing conditions were acoustically analyzed using a model of envelope distortion to examine the effects on the temporal envelope.

Results

Aided consonant identification significantly decreased as reverberation time increased. Consonant identification was also significantly affected by WDRC release time. This relationship was such that individuals tended to perform significantly better with longer release times. There was no significant interaction between reverberation and WDRC. The application of the acoustic model to the processed signal showed a close relationship between trends in the behavioral performance and distortion to the temporal envelope resulting from reverberation and WDRC. The results of the acoustic model demonstrated the same trends found in the behavioral data for both reverberation and WDRC.

Conclusions

Reverberation and WDRC release time both affect aided consonant identification for individuals with hearing impairment, and these condition effects are associated with alterations to the temporal envelope. There was no significant interaction between reverberation and WDRC release time.

INTRODUCTION

Reverberation is an acoustic phenomenon present in many everyday listening environments. Reverberation occurs when some acoustic energy waves are reflected off of features of an environment (e.g., a wall) and arrive at the ear after energy waves that traveled via direct trajectory. This delay of acoustic energy disrupts speech recognition. This effect is caused through distortion of the speech signal by the smearing of spectral and temporal information (Nábělek 1988; Nábělek et al. 1989; Watkins & Holt, 2000; Bidelman & Krishnan, 2010). Reverberation is acoustically distinct from other forms of distortion, such as background noise, and causes a unique pattern of perceptual errors (Nábělek 1988; Helfer & Huntley, 1991; Hedrick & Younger, 2007). Nábělek et al. (1989) describe the effects of reverberation as a function of two main components: self- and overlap-masking. Self-masking refers to the distortion occurring within each phoneme by reverberation. Overlap-masking refers to the distortion that occurs when acoustic information from previous phonemes spills over into the subsequent speech components. Both self- and overlap-masking combine to distort the spectral and temporal cues important for speech perception.

Individuals with sensorineural hearing loss are more susceptible than individuals with normal hearing to the adverse effects of reverberation on speech recognition (Duquesnoy & Plomp, 1980; Humes & Christopherson, 1991; Gordon-Salant & Fitzgibbons, 1993, 1995; Halling & Humes, 2000; Marrone et al. 2008). In one study by Gordon-Salant and Fitzgibbons (1999), speech recognition performance for elderly individuals with hearing impairment significantly decreased when presented at a relatively minimal reverberation time of 0.4 seconds. Reverberation times in real rooms vary widely from nearly zero up to several seconds based on room size and absorptive properties of the room surfaces and materials. For example, a living room with carpets, curtains, and other absorptive materials has a typical reverberation time of 0.50 seconds; whereas, a bathroom with hard tiles which reflect much of the acoustic energy may have a reverberation time as high as 2.0 seconds. Given the increased susceptibility of individuals with hearing impairment to the effects of even minimal amounts of reverberation, speech perception in reverberation is an important clinical concern when dealing with this population.

For individuals with hearing impairment, hearing aids are the most widely distributed rehabilitation device. While hearing aids have been shown to improve individuals’ speech recognition in reverberant environments (Nábělek & Pickett, 1974; Hawkins & Yacullo, 1984; Johnson et al. 2010), hearing aid users nonetheless report significant dissatisfaction with their device’s performance in reverberant environments (Kochkin 2010). The effects of reverberation on hearing aid processing have been explored for a few hearing aid features, including adaptive directional microphone technology either in isolation or combined with digital noise reduction (Hawkins & Yacullo, 1984; Ricketts 2000; Ricketts & Henry, 2002; Walden et al. 2000). However, there is still a relative paucity of information on the effects of reverberation on hearing aid processing as compared to other forms of environmental distortion, such as noise.

Of particular interest in the present study is the core processing strategy in all digital hearing aids: wide-dynamic range compression (WDRC). A WDRC amplifier applies differential amounts of gain based on the level of input, such that low-intensity inputs will receive more gain than high-intensity inputs. The speed at which the processor reacts to changes in input level depends on the attack time (AT) and release time (RT).

In consumer hearing aids AT values are typically short in order to activate rapid decreases in gain following loud impulse noises (e.g., a door slamming); however, RT values range from very short (e.g., ~20ms) to very long (e.g., several seconds). Shorter RTs provide improved audibility to speech by quickly returning gain to low-intensity phonemes directly after higher-intensity speech components; however, these short RTs have also been shown to distort the temporal characteristics of the signal which provide important linguistic information for speech recognition (Plomp 1988; Rosen 1992). Because of the complicated relationship between these factors, there is no clear consensus on how WDRC RTs should be set. Some studies have suggested longer RTs may be better for intelligibility and sound quality (Neuman et al. 1998; Hansen 2002; Moore et al. 2004); whereas other studies have suggested shorter RTs may provide better consonant audibility (van Toor & Verschuure, 2002; Jenstad & Souza, 2005). The contrasting findings of these studies demonstrate the variable nature of optimal RT settings.

One reason for this inconsistency is that optimal hearing aid processing often varies situationally (e.g., Nábělek, 1983). Most of the research investigating the effects of hearing aid amplification strategy on reverberant speech was published before WDRC was the preeminent hearing aid amplification algorithm, and so it is mostly outdated. Only two papers have explicitly investigated the interaction between compression speed and reverberation on speech intelligibility.

Nábělek (1983) investigated the effects of different amplitude compression strategies on reverberant speech. In that study, 10 participants identified nonsense syllables under linear amplification and three-channel multiband compression conditions with an attack time of 1 ms and a release time of 20 ms. When individuals with hearing impairment listened to speech in a reverberant environment (reverberation time = 0.8 seconds) using linear and WDRC amplification, there was no difference in speech recognition for either amplification strategy. This suggests that there is no benefit of WDRC. However, that study investigated a very limited range of reverberation and WDRC conditions.

Shi and Doherty (2008) expanded these findings by including more WDRC and reverberation conditions in the research design. The authors tested sentence intelligibility in 30 individuals with wearable hearing aids set with either a long release time (1500 ms) or short release time (90 ms) at reverberation times up to 3.6 seconds. Results here indicated that increasing amounts of reverberation significantly decreased sentence intelligibility. The authors also found that WDRC amplification with both long and short RTs resulted in improved speech intelligibility over linear amplification; however, there was no significant difference observed between fast and slow compression. This suggests that RT does not have an effect on reverberant speech perception. This is a puzzling result, considering the expected consequences of WDRC release time on temporal envelope fidelity and subsequent behavioral effects (Verschuure et al. 1994; Kates 2010); however, the previous study did not perform any acoustic analysis in order to examine the physical alterations occurring to the signal as a result of reverberation or WDRC.

Thus, these studies yielded discrepant results. Nábělek (1983) concluded that there was no effect of WDRC vs. linear amplification on reverberant speech intelligibility; Shi and Doherty (2008) concluded that there was an overall effect of WDRC, but there was no effect of RT on reverberant speech intelligibility. These studies were limited in their investigation of the range of reverberation and WDRC release time conditions implemented. Shi and Doherty (2008) investigated the effects of WDRC release time as a dichotomous variable of fast- vs. slow-acting WDRC. However, in real hearing aids, RT values vary widely along a continuum. Additionally, neither study investigated the acoustic effects of reverberation and WDRC on the signal. This limits the conclusions that can be drawn from the behavioral results, since it is impossible to assess what acoustic effects (if any) the signal processing manipulations had on the signal.

The purpose of the present study was to further explore the relationship between reverberation and WDRC processing. This project addressed this goal in two primary ways: (1) By investigating reverberation and WDRC across a wide range of externally valid conditions. This was achieved through reverberation and WDRC simulations that allowed for controlled and systematic implementation of variable values and contrasts. (2) By applying an acoustic model of temporal envelope alteration to the processed stimuli. The application of an acoustic model allowed for further investigation of the acoustic effects of reverberation and WDRC and how these acoustic alterations are associated with behavioral results.

MATERIALS AND METHODS

Participants

Participants included 23 adults (mean = 76.2 years, range 57 to 90 years; 13 males) with mild to moderate sloping, symmetrical, sensorineural hearing loss. All individuals were recruited from the greater Chicago area. Participants were evaluated using a standard audiometric protocol including case history, otoscopy, pure tone audiometry, and tympanometry. Participants were included based on results of pure tone audiometry: air conduction testing was performed at .25-8 kHz octave frequencies, and inter-octaves at 3 & 6 kHz; bone conduction was performed at octave frequencies .5-4 kHz. Participants with conductive loss (two air-bone gaps greater than 10 dB at any two frequencies and/or irregular tympanometry re. ASHA 1997) were excluded from the study. All participants reported good general health at the time of testing and no history of neurologic or otologic impairment. Participants all spoke English as their sole or primary language. Individuals completed an informed consent process and were compensated for their time. Participant audiograms for their test ear, and the group mean can be seen in Figure 1. The test ear was always the individual’s better ear or their right ear in cases of complete symmetry.

Figure 1.

Figure 1

Open in a new tab

Individual participant audiograms for their test ear; group mean thresholds in dotted black line

Stimuli

Test stimuli were a set of 16 vowel-consonant-vowel (VCV) syllables consisting of different consonants /b, d, g, p, t, k, f, θ, v, ∫, ð, ð, z, s, m, n/ spoken in an /aCa/ context (Turner et al. 1995). VCV stimuli were selected in order to limit the role of cognitive processes in the task. VCV stimuli were additionally selected to investigate the effects of reverberation and WDRC at a phonemic level. Each syllable was produced by 4 talkers (2 female, 2 male) in order to avoid individual idiosyncrasies. This results in 64 total VCV stimuli. All participants were instructed on /z/ vs /ᴣ/ and /θ/ vs. /ð/ distinctions before testing. The entire set of stimuli was then processed in two stages: reverberation and WDRC simulations.

Processing

  • Stage 1: Reverberation Simulator

    Binaural room impulse responses (BRIR) for different reverberation times were calculated using virtual acoustic techniques (Zahorik, 2009). Different reverberation times were obtained by changing the absorptive properties of the reflective surfaces while the size of the virtual room and the position of the source and listener were kept constant. The VCV stimuli tokens were convolved with the BRIRs to get the reverberant signals. Simulated room size and source-listener distance are specified in Table 1.

    The simulation method used an image model to compute directions, delays, and attenuations of the early reflections, which were spatially rendered along the direct-path using non-individualized head-related transfer functions (Allen & Berkley, 1979). The late reverberant energy was simulated statistically using exponentially decaying Gaussian noise. Further details of the reverberation simulator can be reviewed in Zahorik (2009). Overall, this method of room simulation has been found to produce reverberation simulations that are reasonable physical and perceptual approximation of those measured in real rooms (Zahorik 2009).

Table 1.

Parameters of the reverberation simulator used to process the reverberation stimuli.

Reverberation Simulation Parameters
Room Volume 5.7m × 4.3m × 2.6m
Source-Listener Distance 1.4m
Reverberation Times 0.0, .50, 1.0, 2.0, 4.0s
Open in a new tab

The range of reverberation times selected (0.0, 0.50, 1.0, 2.0, and 4.0 s) was selected to reflect the range of real world reverberation times. The 0.0 s reverberation time denotes an anechoic signal: signal with just the direct-path and no reflections. The relatively mild reverberation times of 0.50 & 1.0 s reflect listening in average rooms in a house; whereas, the higher reverberation times of 2.0 & 4.0 s are closer approximations of listening in highly-reverberant environments such as bathrooms, places of worship, or theaters (Bradley 1986). Processing of the 64 VCV tokens resulted in 320 unique stimuli (16 consonants * 4 talkers * 5 reverberation times) which were then subjected to the WDRC simulator.

  • Step 2: WDRC Simulator

The stimuli were next processed through a virtual hearing aid compression circuit to provide a range of envelope alteration. This program can be modified to simulate any combination of compression characteristics, such as threshold kneepoint, compression ratio, and attack and RTs or number of compression channels. Briefly, the simulator processes the signal into a six-channel filter bank followed by a peak detector that reacts to increases in the within-band signal level with the attack time and decreases in the signal level with the release time. Stimuli were individually processed through the simulator without a carrier phrase. Further details of the WDRC simulator can be reviewed in Arehart et al. (2010). All parameters except for RT were held constant in order to isolate the effects of RT. These parameters were set to common values representative of most WDRC processing in consumer hearing aids. Processing parameters for the WDRC stage of processing can be seen in Table 2. The RT conditions selected (12, 90, 800, & 1500 ms) reflect a range of WDRC RTs used in real hearing aid circuits. The input speech level was 65 dB SPL to represent a conversational speech level. The set of 320 stimuli after reverberation simulation was then processed at these 4 RT values to yield the final set of 1280 VCV stimuli.

Table 2.

Parameters of the WDRC simulator used to process the final stimuli set.

Compressor Simulation Parameters
Compression Threshold 45 dB SPL
Compression Ratio 3:1
Attack Time 10 ms
# Channels 6
Release Times 12, 90, 800, 1500ms
Open in a new tab

Behavioral Procedure

Stimulus presentation and scoring were controlled by a local Matlab program. Test protocol for each participant included a 320 stimuli subset of the total stimuli set. These stimuli were counterbalanced such that each of the four unique talkers was equally represented across the consonants, RTs, and reverberation times (16 consonants * 5 reverberation times * 4 RTs). Stimuli were blocked by the four RTs and reverberant stimuli were randomized within each block to prevent acclimatization to the reverberant characteristics of the listening environment (e.g., Brandewie & Zahorik, 2010; Srinivasan & Zahorik, 2013). The order of the 4 blocks was counterbalanced across subjects. After each trial, participants indicated the consonant perceived in a 16-alternative forced-choice task using a computer mouse to navigate a screen display. Listeners did not receive feedback during testing. Total testing time took 2 hours including audiometric protocol and consent. During testing, breaks were provided at 20-minute intervals to prevent fatigue.

Signals were presented monaurally in the listener’s better ear at 68 dB SPL to represent conversational level and account for loss of binaural summation. Listeners received individual NAL-NL1 shaping to provide sufficient audibility (Byrne & Dillon, 1986). Shaping was based on a single input level so as not to influence the effect of WDRC differentially across listeners with different audiometric losses. One-third octave band levels for the final output signal were measured in a 2 cc coupler and compared to the desired NAL-NL1 target values specified in the same 2 cc coupler using a sound level meter. Consistent with clinical practice, measured output was within +/− 5 dB of target at all third octave bands through 8000 Hz. The digital signals were converted to analog using Tucker-Davis Technologies equipment (RX6, Alachusa, FL) and played through Etymotic ER-2 insert earphones (Elk Grove Village, IL). All procedures were approved by the Northwestern University Institutional Review Board.

Acoustic Analysis

Stimuli were acoustically analyzed using the Envelope Difference Index (EDI; Fortune et al. 1994). The EDI quantifies temporal envelope alteration between two acoustic signals and was calculated using locally developed Matlab code. Two signal comparisons were made: (1) modeling the effects of reverberation by comparing the signal envelopes of the anechoic syllable with the reverberant version of the same syllable. In this instance anechoic digital files were zero padded to be the same length as the reverberant file. This comparison examines the effects of reverberation on temporal envelope. (2) Modeling the effects of WDRC by comparing the signal envelopes of the reverberant uncompressed syllable with the reverberant compressed version of the same syllable. This second comparison examines the effects of WDRC on the temporal envelope.

In both of these comparisons, first a syllable was rectified and digitally low-pass filtered using a Butterworth sixth-order filter with a 50-Hz cutoff to obtain the syllable envelope. The envelope was downsampled to a sampling frequency of 6000 Hz, and the mean amplitude of the syllable was calculated. Each sampled data point of the envelope was scaled to the mean amplitude by dividing every value by the mean. This provided a common reference for comparing the two envelopes. These same steps were performed on the second syllable. The EDI was calculated using the equation below:

EDI=(∑n=1N∣Env1n−Env2n∣)∕2N,

Where N = number of sample points in the waveforms, Env1n was the reference signal (either anechoic or reverberant uncompressed depending on the comparison). Env2n was the comparison signal (either the reverberant or reverberant compressed signal corresponding to the appropriate comparison condition):

RESULTS

Behavioral Results

Data were analyzed using a two-way repeated-measures analysis of variance (RM-ANOVA) with two independent factors (reverberation time and RT) to assess whether there were differences in consonant identification. To stabilize the variance of the dependent variable, absolute scores were transformed to rationalized arcsine units (RAU; Studebaker 1985). There was a significant main effect of reverberation time, F(4,19)=73.656, p < 0.001 with large effect size partial η2 = .939. There was also a significant main effect of RT, F(3,20)=16.268, p < 0.001. The effect of RT was smaller than that of reverberation with a partial η2 = .709. The RT × Reverberation Time interaction was not significant, F(12,11)=1.239, p = 0.365. Before data collection, an a priori power analysis was conducted using G*Power software to determine that a sample size of 24 would be necessary to provide a power of .80 for a medium effect of the interaction with an α = 0.05 criterion for significance. While the current sample falls one subject short of the recommended value for .80 power, given the current p-value of this test it is highly unlikely that an additional subject would lead to a significant result. The data for reverberation time and RT are plotted in Figure 2.

Figure 2.

Figure 2

Open in a new tab

Consonant identification scores in RAUs of reverberant VCV stimuli at 5 different reverberation times for four different release time processing conditions. The error bars represent +/− 1 standard error.

To further explore the main effects found in the RM-ANOVA pairwise comparisons were made with significance level controlled via a Bonferroni correction. Pairwise comparisons showed that consonant identification scores were significantly higher for the 1500 ms RT setting over the 12 ms (p = 0.005), 90ms (p < 0.001), and 800 ms (p = 0.045). Scores were significantly higher for the 800 ms RT setting over the 90ms RT setting (p = 0.006). There were no significant differences between the 12 ms and 800 ms RTs (p = 1.0) and 12 ms and 90 ms RTs (p = 0.185). Increasing the reverberation time significantly decreased consonant identification scores for all pair-wise comparisons (p < 0.001).

Additionally, there was evidence for individual variability in sensitivity to reverberation. The RM-ANOVA resulted in a significant interaction between individual listener and reverberation time (p < 0.001). Listeners ranged from approximately 20% loss of recognition in reverberation to upwards of 70% loss of recognition in reverberation when compared to individual baseline scores. This suggests that some individuals are more susceptible to increasing amounts of reverberation and exhibit a greater decrease in performance with increasing reverberation.

To further examine this source of individual variability a multiple regression was conducted to determine if listener age and degree of hearing loss predicted the consonant identification performance in the most reverberant condition. Degree of hearing loss was quantified by 3-frequency pure tone average (500, 1000, and 2000 Hz), and a composite score for the 4 second reverberation time condition was calculated by taking the arithmetic mean of performance across the 4 release times. Results of this analysis can be seen in Table 3. These two listener factors combined explain 60.2% of the variance in listener performance (F(2,20) = 15.123, p < 0.001, R2 = .602). The model shows that listener age was a significant predictor (Beta = −.623, t(20) = −3.966, p = 0.001); however, pure tone average did not significantly predict performance (Beta = −.263, t(20) = 0.109).

Table 3.

Model explaining consonant identification performance in reverberation by patient characteristics (age and pure-tone average).

Variable Beta P
Age −.623 .001
PTA −.263 .109
Model R2 = .602
Open in a new tab

Acoustic Results

Because there was no interaction in the behavioral data, results of the EDI were analyzed separately for reverberation and WDRC using one-way ANOVAs. Within each one-way ANOVA, results were collapsed across the variable not being investigated. There was a significant main effect of amount of reverberation on changes to the temporal envelope as measured by the EDI [F(3,60) = 334.459, p < 0.001]. Post hoc tests were conducted using a Bonferroni correction for multiple comparisons. All pair-wise comparisons were significantly different (p < 0.001). These results match the trend in the behavioral data in which participants demonstrated significantly poorer consonant identification with increasing reverberation times. The similar trends of the behavioral and acoustic results can be seen in Figure 3. The behavioral results are scaled relative to performance in the anechoic condition; the EDI results are depicting the amount of change occurring to the syllables caused by reverberation.

Figure 3.

Figure 3

Open in a new tab

Comparison of acoustic results of the EDI with behavioral results. The bars depict behavioral performance as a function of reverberation time (A; left) and release time (B; right). Behavioral performance scale for each graph is on left axis. The dotted lines depict EDI results. Acoustic result scale for each graph is on right axis. Error bars for both graphs, and for the bars and lines represent +/− 1 standard error.

EDI values for RT were similarly analyzed using a one-way ANOVA. The EDI values were obtained comparing individual consonants under the reverberant condition with the same consonant and reverberation condition after compression. The values were then transformed in order to represent the similarity between the reverberant and reverberant + compressed signals (i.e., 1-EDI). Thus in this comparison, higher numbers mean less distortion occurring as a result of WDRC. There was a significant main effect of WDRC RT on changes to the temporal envelope as measured by the EDI [F(3,316) = 41.762, p < 0.001]. Post hoc tests were conducted using a Bonferroni correction for multiple comparisons. Results of the post hoc tests matched that of the behavioral experiment. There was no significant difference between the 12 ms and 90 ms conditions (p = 1.0). The 1500 ms RT caused significantly less distortion compared to the 800 ms (p = 0.002), 90 ms, and 12 ms (p < 0.001) RTs, and the 800 ms RT caused significantly less distortion than the 90 ms and 12 ms RTs (p < 0.001). The similar trends of the behavioral and acoustic results can be seen in Figure 3. The behavioral results are plotted as raw proportion correct for each RT condition. The EDI results are depicting the amount of change occurring to the syllables caused by different WDRC RTs. The EDI results are scaled relative to behavioral performance in the 12 ms RT condition.

DISCUSSION

The current study evaluated the effects of reverberation and RT on consonant recognition in individuals with hearing impairment. Effects were assessed using behavioral testing and via the application of the acoustic EDI model which evaluates the amount of temporal distortion occurring to a signal. Consonant identification was significantly reduced as the amount of reverberation increased. There was also a significant main effect of WDRC RT on performance. This relationship was such that participants achieved the highest intelligibility with longer RTs. There was no significant interaction between reverberation and WDRC, and both manipulations had additive effects. Trends of the EDI were consistent with behavioral results: both reverberation and WDRC cause significant changes to the temporal envelope as measured by the acoustic model. These parallel trends of behavioral and acoustic results suggest that behavioral performance is associated with changes to the temporal envelope.

Reverberation

The findings of the present study are consistent with previous results suggesting that individuals with hearing impairment are significantly affected by even mild amounts of reverberation (e.g., Helfer & Wilber, 1990; Humes & Christopherson, 1991; Gordon-Salant & Fitzgibbons, 1993, 1995, 1999; Sato et al. 2007; Shi & Doherty, 2008). This vulnerability of individuals with hearing impairment to reverberation has important implications for testing. Much of the research in speech perception of individuals with hearing impairment involves anechoic presentation of sounds to the listener; however, this is not indicative of many realistic listening conditions in which some base level amount of reverberation would be present. While the normal auditory system is able to cope with these mild reverberation (e.g., Humes & Christopherson, 1991; Gordon-Salant & Fitzgibbons, 1993, 1999), as demonstrated here, listeners with hearing impairment are significantly affected. Additionally, most clinical fittings and validations are performed under idealized conditions which contain very little reverberation. These environments may not represent the different listening environments in which hearing aid users will primarily be listening through their device. Thus, it might be important to provide validation for supplemental testing of performance of hearing aid users in more realistic adverse conditions than is currently applied for functional testing. This suggests that testing, both clinical and research, should incorporate some reverberation into speech perception measures in order to better approximate real-world listening and improve external validity.

Declines in intelligibility as a function of reverberation from the behavioral task were paralleled by distortion to the temporal envelope as measured by the EDI. Reverberation is a multi-faceted source of distortion that affects both the spectral and temporal properties of speech (Nábělek 1988; Nábělek et al. 1989; Watkins & Holt, 2000). While the EDI only captures acoustic effects to the temporal envelope and does not include spectral or fast-modulation distortion, the results of the EDI still successfully modeled the behavioral effects. The EDI has not previously been used in the literature to quantify the distortive effects of reverberation. This suggests that the EDI can successfully be used clinically to model how reverberation will affect intelligibility for listeners with hearing impairment.

Interestingly, not all individuals were equally affected by reverberation; some participants were significantly more susceptible to the deleterious effects of reverberation on speech recognition. Results of this study suggest that age is a significant factor driving susceptibility to reverberation, whereas degree of hearing loss does not. This is consistent with previous work that has identified a significant effect of age on reverberant speech intelligibility (Plomp & Duquesnoy, 1980; Gordon-Salant & Fitzgibbons, 1993, 1995). Previous work has also suggested that performance in reverberation may be affected by individual factors including temporal resolution (Gordon-Salant & Fitzgibbons, 1995, 1999), degree of hearing loss (Gordon-Salant & Fitzgibbons, 1993; Marrone et al. 2008), and cognition (Kjellberg 2004). If any of these measures are sufficiently sensitive and specific for indicating reverberant susceptibility, such a measure could be used clinically in order to identify patients who would benefit most from hearing aid algorithms focused on improving reverberant speech. These patients could then be fitted with reverberation suppressing signal processing features in the initial fit to proactively address this individual deficit. However, further research is necessary to understand the underlying traits driving susceptibility to reverberation and the efficacy of these signal processing features.

WDRC

In the present study, the WDRC RT significantly affected consonant identification even in the presence of reverberation. Thus, the parameters involved in WDRC processing did have an effect on speech perception even under reverberant conditions. This finding is in contrast with the previous study investigating the effects of WDRC RT on reverberant speech (Shi & Doherty, 2008). In the previous study listeners performed significantly better with fast and slow compression over a linear system; however, there was no difference in speech intelligibility between the fast and slow compression systems. These differing results could be due to several methodological differences between the studies. The present study used a fixed compression ratio of 3:1, whereas Shi and Doherty (2008) used individualized compression ratios for each listener based on NAL-NL1 standards resulting in compression ratios between 1.07 and 2.67. It has been demonstrated that higher compression ratios exacerbate the effects of RT (Neuman et al. 1998; Olsen et al. 2004). The fixed 3:1 compression ratio used in this study represents the upper limit of the range typically used in WDRC hearing aids. While real hearing aids provide individualized compression ratios based on audiometric configuration, the uniform application of compression provided the same degree of envelope alteration in order to ensure that identical stimuli were presented to all listeners. This also allowed for the use of the EDI model to examine the acoustic effects of WDRC release time since each participation received the same amount of compression.

Another possibility for this discrepancy is due to the differences in the stimulus type used between the two studies to assess speech intelligibility. Shi and Doherty (2008) used the Speech-in-Noise Test. In this test, intelligibility is scored on the basis of whether individuals are able to identify the last word of a sentence. The present study used isolated nonsense syllables in which the task is to identify the consonant embedded between two vowels. Because reverberation has a greater effect on speech recognition than WDRC, it could be that the effect of reverberation overpowers the smaller effects of WDRC by the final word of the sentence. In other words, in a sentence there would be more overlap-masking because there are more preceding phonemes compared to a single vowel. It was suggested by Shi and Doherty that reverberation may do this by “masking” the smaller effects of RT. However, this is unlikely because the present study used a higher reverberation time that would cause a greater amount of distortion than Shi and Doherty, and the effect of RT remained even in the presence of this more extreme condition. Additionally, results of the EDI comparison demonstrated that RT significantly affects the temporal envelope even in the presence of reverberation. Thus, there are additive effects of RT on consonant recognition associated with changes to the temporal envelope that persist even under highly reverberant conditions.

Another possibility for this difference is that nonsense syllables are more sensitive stimuli to the effects of RT. Participants in Shi and Doherty (2008) were listening to sentences and had access to some context information and a longer speech stimulus to identify. However, the present study used nonsense syllables in which the task was to identify a low-intensity consonant embedded between two high-intensity vowels. In this instance, the preceding vowel triggered the compression of the hearing aid simulator. The short RT (12 ms) is sufficiently fast to restore gain to the consonant portion; whereas, the longer RTs did not allow the hearing aid simulator to fully restore gain in time to amplify the consonant. Thus, the effect of varying RT would be greatest for the consonant portion, and it may be that with a longer utterance this effect would be minimized.

This study is consistent with several other studies (for non-reverberant speech) that have demonstrated that fast compression results in worse performance when combined with high compression ratios (Verschuure et al. 1994; Stone & Moore, 2004, 2008). The effects of RT in the present study were related to distortion to the temporal envelope, and significant changes to the temporal envelope resulting from WDRC corresponded to significant changes in the behavioral results. These findings are consistent with the idea that individuals even with mild degrees of hearing loss depend to varying extents on the temporal cues of speech (Souza et al. 2015), and this relationship between envelope distortion and speech recognition is consistent with previous applications of the EDI for nonsense syllables; however, it has been noted that WDRC does not affect all consonants equally (Jenstad and Souza, 2005, 2007). Jenstad and Souza (2005) observed that voiceless consonants were more affected by WDRC. The present study did not have sufficient consonant repetition within each processing condition to analyze specific error patterns. However, presumably because there was no overall interaction between reverberation and RT for consonant recognition, these error patterns would be similar under reverberant conditions.

Given the increased susceptibility of hearing-impaired individuals to reverberation, it is unsurprising that hearing aid users report significant relative dissatisfaction in reverberant environments (Kochkin 2010). Algorithms have been specifically proposed to address the issue of environmental reverberation in hearing aid amplification (e.g., Lebart et al. 2001; Kustomoto et al. 2005; Wu & DeLiang, 2006). Currently, some hearing aid manufacturers offer features designed to suppress the effects of reverberation. These features work primarily through adjusting the time constants of the compressor. This study suggests that altering the WDRC constants can have acoustic and behavioral effects on listener speech intelligibility in reverberant environments.

General Discussion

The use of the EDI in the present study was able to model both the effects of reverberation and WDRC on the signal acoustics. The EDI results paralleled the behavioral results of the consonant identification task. This conclusion is consistent with previous studies using the EDI to model distortion effects (e.g., Jenstad & Souza, 2005, 2007; Souza et al. 2012). This study further demonstrates the validity of the EDI’s usefulness as a means of modelling distortion occurring to the temporal envelope resulting from external factors (e.g., reverberation and noise) and factors related to the hearing aid signal processing (e.g., WDRC settings). This study expands the potential clinical use of the EDI by presenting evidence that the EDI can be used to model the effects of reverberation. Moreover, the EDI can be used to examine the effects of the acoustic environment and hearing aid processing simultaneously in order to predict performance.

While these effects were observed in the present study under monaural laboratory conditions, it is not certain whether these effects will generalize to real-world reverberant environments with actual hearing aids. Clinical hearing aids provide compression characteristics individualized for each listener based on residual dynamic range. This would likely lead to smaller compression ratios than that used in the current study.

Another consideration is the role of compression and audibility. A primary goal of fast-acting WDRC is to provide customized gain suited to each syllable or phoneme in order to place the greatest amount of speech energy within the listener’s dynamic range. In the present study, speech tokens were processed through the WDRC simulator at 65 dB SPL input. This represents a conversational speech level which would have provided good audibility for the mild-to-moderate degrees of hearing loss, regardless of compression speed. Thus, having shorter release times may have had minimal effects on audibility which would not outweigh the detrimental effects of distorting the temporal characteristics of speech. It would be of further interest to examine how varying speech level would affect this relationship between reverberant speech and WDRC release time.

The results of this experiment suggest that hearing aid processing has an additive effect to reverberation that manifests itself both acoustically and behaviorally. Results also suggest that the EDI can be used to simultaneously model the acoustic effects of reverberation and WDRC processing. Previous research in the area of WDRC settings and adverse listening conditions has demonstrated a higher order interaction among WDRC, acoustic environment, and individual cognition (Lunner & Sundewall-Thorén, 2007; Rudner et al. 2009; Ohlenforst et al. 2014). The present study did not consider individual cognition effects because the nature of the stimuli were unlikely to substantially engage cognitive processes; however, based on this previous research indicating a relationship among environmental, individual, and signal processing factors, further investigation is warranted in the area of reverberation and signal processing as well as under more realistic conditions (e.g., real hearing aids, background noise which would likely be present in most reverberant environments, etc.). The present study represents a step towards understanding this interaction by investigating these effects at a phonemic level both acoustically and behaviorally. Further research is necessary to fully understand how to optimize hearing aids to support aided speech recognition in reverberation under more realistic conditions.

Acknowledgements

The authors thank James Kates for providing the compression program, and Laura Mathews for her help with data collection. Work supported by National Institutes of Health Grants R01 DC0060014 to Pamela Souza and R01 DC011828 to Frederick Gallun. Portions of this work were presented at the International Hearing Aid Research Conference, Lake Tahoe, California, August 2014, and at the Annual Scientific and Technology Conference of the American Auditory Society, Scottsdale, Arizona, March 2015.

Financial Disclosures/Conflicts of Interest:

This research was funded by the National Institutes of Health Grants R01 DC0060014 to Pamela Souza and R01 DC011828 to Frederick Gallun

References

  1. Allen JB, Berkley DA. Image method for efficiently simulating small-room acoustics. J Acoust Soc Am. 1979;65(4):943–950. [Google Scholar]
  2. Arehart KH, Kates JM, Anderson MC. Effects of noise, nonlinear processing, and linear filtering on perceived speech quality. Ear Hear. 2010;31(3):420–436. doi: 10.1097/AUD.0b013e3181d3d4f3. [DOI] [PubMed] [Google Scholar]
  3. American Speech-Language-Hearing Association Guidelines for audiologic screening. 1997 [Google Scholar]
  4. Bidelman GM, Krishnan A. Effects of reverberation on brainstem representation of speech in musicians and non-musicians. Brain Res. 2010;1355:112–125. doi: 10.1016/j.brainres.2010.07.100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bradley JS. Predictors of speech intelligibility in rooms. J Acoust Soc Am. 1986;80(3):837–845. doi: 10.1121/1.393907. [DOI] [PubMed] [Google Scholar]
  6. Brandewie E, Zahorik P. Prior listening in rooms improves speech intelligibility. J Acoust Soc Am. 2010;128(1):291–299. doi: 10.1121/1.3436565. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Byrne D, Dillon H. The National Acoustic Laboratories (NAL) new procedure for selecting the gain and frequency response of a hearing aid. Ear Hear. 1986;7(4):257–265. doi: 10.1097/00003446-198608000-00007. [DOI] [PubMed] [Google Scholar]
  8. Duquesnoy AJ, Plomp R. Effect of reverberation and noise on the intelligibility of sentences in cases of presbycusis. J Acoust Soc Am. 1980;68:537. doi: 10.1121/1.384767. [DOI] [PubMed] [Google Scholar]
  9. Fortune TW, Woodruff BD, Preves DA. A new technique for quantifying temporal envelope contrasts. Ear Hear. 1994;15(1):93–99. doi: 10.1097/00003446-199402000-00011. [DOI] [PubMed] [Google Scholar]
  10. Gordon-Salant S, Fitzgibbons PJ. Temporal factors and speech recognition performance in young and elderly listeners. J Speech Lang Hear Res. 1993;36(6):1276–1285. doi: 10.1044/jshr.3606.1276. [DOI] [PubMed] [Google Scholar]
  11. Gordon-Salant S, Fitzgibbons PJ. Comparing recognition of distorted speech using an equivalent signal-to-noise ratio index. J Speech Lang Hear Res. 1995;38(3):706–713. doi: 10.1044/jshr.3803.706. [DOI] [PubMed] [Google Scholar]
  12. Gordon-Salant S, Fitzgibbons PJ. Profile of auditory temporal processing in older listeners. J Speech Lang Hear Res. 1999;42(2):300–311. doi: 10.1044/jslhr.4202.300. [DOI] [PubMed] [Google Scholar]
  13. Halling DC, Humes LE. Factors affecting the recognition of reverberant speech by elderly listeners. J Speech Lang Hear Res. 2000;43(2):414–431. doi: 10.1044/jslhr.4302.414. [DOI] [PubMed] [Google Scholar]
  14. Hansen M. Effects of multi-channel compression time constants on subjectively perceived sound quality and speech intelligibility. Ear Hear. 2002;23(4):369–380. doi: 10.1097/00003446-200208000-00012. [DOI] [PubMed] [Google Scholar]
  15. Hawkins DB, Yacullo WS. Signal-to-noise ratio advantage of binaural hearing aids and directional microphones under different levels of reverberation. J Speech Hear Disord. 1984;49(3):278–286. doi: 10.1044/jshd.4903.278. [DOI] [PubMed] [Google Scholar]
  16. Hedrick MS, Younger MS. Perceptual weighting of stop consonant cues by normal and impaired listeners in reverberation versus noise. J Speech Lang Hear Res. 2007;50(2):254–269. doi: 10.1044/1092-4388(2007/019). [DOI] [PubMed] [Google Scholar]
  17. Helfer KS, Huntley RA. Aging and consonant errors in reverberation and noise. J Acoust Soc Am. 1991;90(4):1786–1796. doi: 10.1121/1.401659. [DOI] [PubMed] [Google Scholar]
  18. Helfer KS, Wilber LA. Hearing loss, aging, and speech perception in reverberation and noise. J Speech Lang Hear Res. 1990;33(1):149–155. doi: 10.1044/jshr.3301.149. [DOI] [PubMed] [Google Scholar]
  19. Humes LE, Christopherson L. Speech Identification Difficulties of Hearing-Impaired Elderly Persons The Contributions of Auditory Processing Deficits. J Speech Lang Hear Res. 1991;34(3):686–693. doi: 10.1044/jshr.3403.686. [DOI] [PubMed] [Google Scholar]
  20. Jenstad LM, Souza PE. Quantifying the effect of compression hearing aid release time on speech acoustics and intelligibility. J Speech Lang Hear Res. 2005;48(3):651–667. doi: 10.1044/1092-4388(2005/045). [DOI] [PubMed] [Google Scholar]
  21. Jenstad LM, Souza PE. Temporal envelope changes of compression and speech rate: Combined effects on recognition for older adults. J Speech Lang Hear Res. 2007;50(5):1123–1138. doi: 10.1044/1092-4388(2007/078). [DOI] [PubMed] [Google Scholar]
  22. Johnson JA, Cox RM, Alexander GC. Development of APHAB norms for WDRC hearing aids and comparisons with original norms. Ear Hear. 2010;31(1):47–55. doi: 10.1097/AUD.0b013e3181b8397c. [DOI] [PubMed] [Google Scholar]
  23. Kates JM. Understanding compression: Modeling the effects of dynamic-range compression in hearing aids. Int J Audiol. 2010;49(6):395–409. doi: 10.3109/14992020903426256. [DOI] [PubMed] [Google Scholar]
  24. Kjellberg A. Effects of reverberation time on the cognitive load in speech communication: Theoretical considerations. Noise Health. 2004;7(25):11. [PubMed] [Google Scholar]
  25. Kochkin S. MarkeTrak VIII: Consumer satisfaction with hearing aids is slowly increasing. Hear J. 2010;63(1):19–20. [Google Scholar]
  26. Kusumoto A, Arai T, Kinoshita K, Hodoshima N, Vaughan N. Modulation enhancement of speech by a pre-processing algorithm for improving intelligibility in reverberant environments. Speech Commun. 2005;45(2):101–113. [Google Scholar]
  27. Lebart K, Boucher JM, Denbigh PN. A new method based on spectral subtraction for speech dereverberation. Acta Acust united Ac. 2001;87(3):359–366. [Google Scholar]
  28. Lunner T, Sundewall-Thorén E. Interactions between cognition, compression, and listening conditions: Effects on speech-in-noise performance in a two-channel hearing aid. J Am Acad Audiol. 2007;18(7):604–617. doi: 10.3766/jaaa.18.7.7. [DOI] [PubMed] [Google Scholar]
  29. Marrone N, Mason CR, Kidd G., Jr The effects of hearing loss and age on the benefit of spatial separation between multiple talkers in reverberant rooms. J Acoust Soc Am. 2008;124(5):3064–3075. doi: 10.1121/1.2980441. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Miller GA, Nicely PE. An analysis of perceptual confusions among some English consonants. J Acoust Soc Am. 1955;27(2):338–352. [Google Scholar]
  31. Moore BC, Stainsby TH, Alcántara JI, Kuhnel V. The effect on speech intelligibility of varying compression time constants in a digital hearing aid. Int J Audiol. 2004;43(7):399–409. doi: 10.1080/14992020400050051. [DOI] [PubMed] [Google Scholar]
  32. Nábělek AK, Pickett JM. Monaural and binaural speech perception through hearing aids under noise and reverberation with normal and hearing-impaired listeners. J Speech Lang Hear Res. 1974;17(4):724–739. doi: 10.1044/jshr.1704.724. [DOI] [PubMed] [Google Scholar]
  33. Nábělek AK. Identification of vowels in quiet, noise, and reverberation: Relationships with age and hearing loss. J Acoust Soc Am. 1988;84(2):476–484. doi: 10.1121/1.396880. [DOI] [PubMed] [Google Scholar]
  34. Nábělek AK, Letowski TR, Tucker FM. Reverberant overlap-and self-masking in consonant identification. J Acoust Soc Am. 1989;86(4):1259–1265. doi: 10.1121/1.398740. [DOI] [PubMed] [Google Scholar]
  35. Nábělek IV. Performance of hearing-impaired listeners under various types of amplitude compression. J Acoust Soc Am. 1983;74(3):776–791. doi: 10.1121/1.389865. [DOI] [PubMed] [Google Scholar]
  36. Neuman AC, Bakke MH, Mackersie C, Hellman S, Levitt H. The effect of compression ratio and release time on the categorical rating of sound quality. J Acoust Soc Am. 1998;103(5):2273–2281. doi: 10.1121/1.422745. [DOI] [PubMed] [Google Scholar]
  37. Ohlenforst B, Souza P, MacDonald E. Interaction of working memory, compressor speed and background noise characteristics. American Auditory Society; Scottsdale, AZ: 2014. [Google Scholar]
  38. Olsen HL, Olofsson Å, Hagerman B. The effect of presentation level and compression characteristics on sentence recognition in modulated noise. Int J Audiol. 2004;43(5):283–294. doi: 10.1080/14992020400050038. [DOI] [PubMed] [Google Scholar]
  39. Plomp R. The negative effect of amplitude compression in multichannel hearing aids in the light of the modulation-transfer function. J Acoust Soc Am. 1988;83(6):2322–2327. doi: 10.1121/1.396363. [DOI] [PubMed] [Google Scholar]
  40. Ricketts T. Impact of noise source configuration on directional hearing aid benefit and performance. Ear Hear. 2000;21(3):194–205. doi: 10.1097/00003446-200006000-00002. [DOI] [PubMed] [Google Scholar]
  41. Ricketts T, Henry P. Evaluation of an adaptive, directional-microphone hearing aid. Int J Audiol. 2002;41(2):100–112. doi: 10.3109/14992020209090400. [DOI] [PubMed] [Google Scholar]
  42. Rosen S. Temporal information in speech: acoustic, auditory and linguistic aspects. Phil Trans R Soc London Ser B. 1992;336(1278):367–373. doi: 10.1098/rstb.1992.0070. [DOI] [PubMed] [Google Scholar]
  43. Rudner M, Foo C, Rönnberg J, Lunner T. Cognition and aided speech recognition in noise: Specific role for cognitive factors following nine-week experience with adjusted compression settings in hearing aids. Scand J Psychol. 2009;50(5):405–418. doi: 10.1111/j.1467-9450.2009.00745.x. [DOI] [PubMed] [Google Scholar]
  44. Sato H, Sato H, Morimoto M. Effects of aging on word intelligibility and listening difficulty in various reverberant fields. J Acoust Soc Am. 2007;121(5):2915–2922. doi: 10.1121/1.2713715. [DOI] [PubMed] [Google Scholar]
  45. Souza P, Hoover E, Gallun F. Application of the Envelope Difference Index to Spectrally Sparse Speech. J Speech Lang Hear Res. 2012;55(3):824–837. doi: 10.1044/1092-4388(2011/10-0301). [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Souza PE, Wright RA, Blackburn MC, Tatman R, Gallun FJ. Individual Sensitivity to Spectral and Temporal Cues in Listeners With Hearing Impairment. J Speech Lang Hear Res. 2015;58(2):520–534. doi: 10.1044/2015_JSLHR-H-14-0138. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Srinivasan NK, Zahorik P. Prior listening exposure to a reverberant room improves open-set intelligibility of high-variability sentences. J Acoust Soc Am. 2013;133(1):EL33–EL39. doi: 10.1121/1.4771978. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Stone MA, Moore BC. Side effects of fast-acting dynamic range compression that affect intelligibility in a competing speech task. J Acoust Soc Am. 2004;116(4):2311–2323. doi: 10.1121/1.1784447. [DOI] [PubMed] [Google Scholar]
  49. Stone MA, Moore BC. Effects of spectro-temporal modulation changes produced by multi-channel compression on intelligibility in a competing-speech task. J Acoust Soc Am. 2008;123(2):1063–1076. doi: 10.1121/1.2821969. [DOI] [PubMed] [Google Scholar]
  50. Studebaker GA. A rationalized arcsine transform. J Speech Lang Hear Res. 1985;28(3):455–462. doi: 10.1044/jshr.2803.455. [DOI] [PubMed] [Google Scholar]
  51. Turner CW, Souza PE, Forget LN. Use of temporal envelope cues in speech recognition by normal and hearing-impaired listeners. J Acoust Soc Am. 1995;97(4):2568–2576. doi: 10.1121/1.411911. [DOI] [PubMed] [Google Scholar]
  52. Tyler RS, Hall JW, Glasberg BR, Moore BC, Patterson RD. Auditory filter asymmetry in the hearing impaired. J Acoust Soc Am. 1984;76(5):1363–1368. doi: 10.1121/1.391452. [DOI] [PubMed] [Google Scholar]
  53. van Toor T, Verschuure H. Effects of high-frequency emphasis and compression time constants on speech intelligibility in noise. Int J Audiol. 2002;41(7):379–394. doi: 10.3109/14992020209090415. [DOI] [PubMed] [Google Scholar]
  54. Verschuure H, Prinsen TT, Dreschler WA. The effects of syllabic compression and frequency shaping on speech intelligibility in hearing impaired people. Ear Hear. 1994;15(1):13–21. doi: 10.1097/00003446-199402000-00003. [DOI] [PubMed] [Google Scholar]
  55. Villchur E. Signal processing to improve speech intelligibility in perceptive deafness. J Acoust Soc Am. 1973;53(6):1646–1657. doi: 10.1121/1.1913514. [DOI] [PubMed] [Google Scholar]
  56. Walden BE, Surr RK, Cord MT, Edwards B, Olson L. Comparison of benefits provided by different hearing aid technologies. J Am Acad Audiol. 2000;11(10):540–560. [PubMed] [Google Scholar]
  57. Watkins AJ, Holt NJ. Effects of a complex reflection on vowel identification. Acta Acust united Ac. 2000;86(3):532–542. [Google Scholar]
  58. Wu M, Wang D. A two-stage algorithm for one-microphone reverberant speech enhancement. IEEE Trans Audio, Speech, Language Process. 2006;14(3):774–784. [Google Scholar]
  59. Zahorik P. Perceptually relevant parameters for virtual listening simulation of small room acoustics. J Acoust Soc Am. 2009;126(2):776–791. doi: 10.1121/1.3167842. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES