Temporal and speech processing skills in normal hearing individuals exposed to occupational noise

Introduction

One of the major causes of cochlear hearing loss especially in adults is noise. Prolonged exposure to high levels of noise results in noise-induced hearing loss. The prolonged exposure to high levels of noise causes damage to the hair cells in the cochlea and results in permanent noise-induced cochlear hearing loss. Noise damage to hearing is a health risk which is associated with both civilian and military occupations as well as certain leisure activities. Psychophysical evidence indicates that presence of cochlear hearing loss causes deficits in temporal processing. [1] Temporal processing encompasses a wide range of auditory skills including temporal resolution or temporal discrimination (i.e., gap detection and fusion), masking (i.e., backward and forward masking), temporal integration (i.e., temporal summation), and temporal ordering (i.e., temporal sequencing), as well as localization and pitch perception. [2] Normal perception of the temporal aspects of the stimulus is crucial for understanding speech in quiet and adverse listening conditions. [3] It has been shown that individuals with cochlear hearing loss perform poorly on tasks such as gap detection, modulation detection, and temporal integration. [1] One of the important factors that contribute to the poor performance hearing-impaired listeners on temporal processing tasks is audibility of high-frequency signals. [1] Apart from audibility, the suprathreshold distortions can also contribute to the poor performance of hearing-impaired listeners on temporal processing tasks. These suprathreshold distortions may be caused by changes in the central auditory system secondary to cochlear damage. Long-term exposure to high levels of occupational noise is shown to result in hemispheric reorganization for speech processing. [4] Brattico [5] showed that well-known left hemisphere dominance in speech discrimination became right hemisphere preponderant in individuals exposed to occupational noise. Chang et al[6] in animal model showed that when infant rat pups were reared in moderate levels of noise, it resulted in delay of organizational maturation of the auditory cortex.

Recently, it has been shown that cochlear damage can occur without significant reduction in sensitivity. Fneg et al[7] measured the temporal resolution and speech perception in adults with sloping high-frequency sensory neural hearing loss in regions of normal hearing. Temporal resolution was evaluated through amplitude modulation detection and gap detection in noise. Speech perception was assessed trough hearing in noise test. Results showed that individuals with sensory neural hearing loss performed poorly on both gap detection and modulation detection even though the stimulus was restricted to the regions of normal hearing. They hypothesized that the deterioration in temporal resolution in individuals with high-frequency sensory neural hearing loss is caused by central auditory deficits secondary to cochlear damage or subclinical cochlear damage in the lower frequencies. Evidences from the animal research also suggest that the cochlear lesion can occur without significant change in the hearing sensitivity. [8]

Consequences of cochlear hearing loss on speech perception and psychophysical abilities have been well documented. The aim of this study was to evaluate the speech recognition and temporal processing abilities of train drivers with normal hearing sensitivity who were exposed to continuous noise of more than Leq8h of 80 dBA of the engine for a period of more than 10 years.

Methods

We evaluated (1) gap detection in noise; (2) modulation detection for sinusoidally amplitude modulated noise at 8, 20, 60, and 100 Hz; and (3) duration pattern and speech recognition in the presence of multi-talker babble at -5 dB signal to noise ratio in individuals who were exposed occupational noise of more than Leq8 h of 80dBA. Results were compared with matched control group. Contribution of temporal processing skills to speech perception in adverse listening situation was also measured.

Participants

A total of 118 participants took part in this research. Participants comprised three groups of train drivers in the age range of 30-40 (n = 13), 41 50 ( = 13), 41-50 (n = 9), and 51-60 (n = 6) years and their non-noise-exposed counterparts (n = 30 in each age group). Because both the temporal processing and speech perception are influenced by age, participants were subgrouped depending on age. Our earlier study has shown that temporal processing skills of individuals below 40 years were significantly different from those of elderly group. [9] All the members of the experimental group were exposed to railway engine noise for about 8-10 hours a day for a period of more than 10 years. The amount of leisure time varied between 1 and 3 hours. Noise survey showed that the engine noise at the level of the train driver's ear was Leq of 86 dBA. Participants of all the groups including the train drivers had hearing sensitivity within 25 dB HL in the octave frequencies between 250 and 8 kHz. None of the subjects reported any otological or neurological problems, usage of ototoxic drugs, or exposure to organic solvents.

Stimulus and procedure

Psychophysical tests

Gap detection and modulation detection were measured using "mlp" tool box which implements a maximum likelihood procedure in Matlab. [10] The maximum likelihood procedure uses a large number of candidate psychometric functions and after each trial calculates the probability (or likelihood) of obtaining the listener's response to all of the stimuli that have been presented given each psychometric function. The psychometric function yielding the highest probability is used to determine the stimulus to be presented on the next trial. Within about 12 trials, the maximum likelihood procedure usually converges on a reasonably stable estimate of the most likely psychometric function, which then can be used to estimate threshold. [11],[12] Stimuli were produced at 44,100 Hz sampling rate. A two-interval alternate forced choice method using a "maximum likelihood procedure" was used to track an 80% correct response criterion. During each trial, stimuli were presented in each of the two intervals: One interval contained a reference stimulus and the other interval the variable stimulus. The participant indicated after each trial which interval contained the variable stimulus. Stimuli for the duration pattern test were generated using Audacity software 1.3.5 (beta version 2008). In all of the psychophysical tests, stimuli were presented binaurally at an intensity of 80 dB SPL. Stimuli were presented via a laptop computer (Lenovo 3000 G530) connected to EAR-3A earphones. Output of the earphones was calibrated at the beginning of the experiment and regularly thereafter to produce 80 dB SPL for a 1 kHz pure tone in a 2cc coupler. For this purpose, a 1 kHz pure tone was generated at the same root mean square (rms) level as the test signal. Output of the ear phone was routed to a 2cc coupler which was connected to a sound level meter (Quest 1800) and a microphone (Quest 4180). The volume control of the computer was adjusted to produce 80 SPL on the sound level meter. Subjects were given three to four practice trails before the commencement of each test. All psychophysical tests were carried out in a quiet room.

Gap detection in white noise

In this subjects, ability to detect a temporal gap in the center of 750 ms broadband noise was measured. Duration of gap was varied according to the listener performance using maximum likelihood procedure. The noise had 0.5 ms cosine ramps at the beginning and end of the gap. In two interval alternate force choice tasks, the standard stimulus was always a 750 ms broadband noise with no gap, whereas the variable stimulus contained the gap.

Modulation detection thresholds

Temporal modulation refers to a reoccurring change (in frequency or amplitude) in a signal over time. A 500 ms Gaussian noise was sinusoidally amplitude modulated at modulation frequencies of 8, 20, 60, and 200 Hz. Noise stimuli had two 10-ms raised cosine ramps at onset and offset. The subject had to detect the modulation and determine which interval had the modulated noise. Modulated and unmodulated stimuli were equated for total rms power. Depth of the modulated signal was varied according to the participant's response up to an 80% criterion level. The modulation detection thresholds were expressed in dB by using the following equation:

Modulation detection thresholds in dB = 20 log 10 m

where m = modulation detection threshold in percentage

Duration pattern test

The duration pattern test was administered in the manner described by Musiek et al. [13] A 1000 Hz pure tone was generated at 44,100 sampling frequency with two different durations (i.e., short 250 ms and long 500 ms) using Audacity software (ver. 1.3.5). By combining these two durations in three-tone patterns, six different patterns were generated (Short Short Long, Short Long Short, Long Long Short, Long Short Short, Short Long Long, Long Short Long). Interstimulus interval was 250 ms within a tone sequence and 6 s between two tone sequences. Following practice trails, 30 test items were administered. Participants were asked to verbally repeat the sequence.

Speech recognition with multi-talker babble

Speech recognition ability of the subjects was tested using a set of custom-made sentence material. Eighty sentences were constructed with a reasonable degree of homogeneity, complexity, and sentence length. Each sentence had four to five key words. These 80 sentences were provided to five native Kannada speakers (age range: 20-40 years) who had at least 15 years of formal education in Kannada. These five Kannada speakers served as judges. They were instructed to underline the key words and also to rate the complexity levels of the sentences on a five point rating scale (1, very difficult; 2, moderately difficult; 3, difficult; 4, easy; 5, very easy). Ten sentences were selected from this pool based on judge's ratings. The selected sentences had a 100% agreement among all the five judges in terms of the key word identification and the complexity level ratings. There were a total of 44 key words. Sentences used in the test are provided in Appendix I. A 23-year-old native male Kannada speaker spoke these sentences. Using the Praat software the spoken sentences were digitally recorded on a Presario C700 Compaq laptop in a sound treated room via a JVC MV 40 microphone. The sampling frequency was set at 44,100 Hz. Using a custom-written Matlab code, four-talker babble was added to the sentences at -5 dB SNR. The Matlab program first calculated the rms amplitude of the speech stimulus and then adjusted the rms of the four-talker babble to achieve the desired signal to noise ratio. These sentences were randomly presented to the subjects at 80 dB SPL intensity through an insert EAR-3A earphones which were connected to a computer. Subjects were instructed to repeat the sentences. The repeated sentences were recorded and used for further analysis. Each correctly repeated key word was given a score of "1" and the total number of correct responses was calculated for each subject separately.

Results

Psychophysical tests

Gap detection in noise

[Figure 1] shows the mean and 1 SD error bars for gap detection thresholds in the experimental and control groups across different age groups. Two-way ANOVA was done to test the significance of differences between the mean gap detection thresholds of different groups. ANOVA did not reveal a significant main effect of noise exposure on the gap detection thresholds [F(2,124) = 1.08, P > 0.05].{Figure 1}

Modulation detection thresholds

[Figure 2]a shows the mean modulation detection thresholds for 8, 20, 60, and 200 Hz modulation frequencies along with the 1 SD of variation in experimental and control group in the age range of 30-40 years. The mean and standard deviation values of the modulation detection thresholds for 41-50 years and 51-60 years group are shown in [Figure 2]b and c, respectively. MANOVA was done to find the significance of differences between the means of modulation detection thresholds between experimental and control groups. MANOVA showed the significant main effect of participant groups on the modulation detection thresholds [F(4,122) = 4.78, P0 < 0.01]. Post hoc analysis with Bonferroni correction revealed that noise-exposed group had significantly poor modulation detection thresholds in higher frequencies (at 60 and 200 Hz modulation frequencies) in 30-40 years and 41-50 years.{Figure 2}

Duration pattern test

[Figure 3] shows the mean and 1 SD error bars for duration pattern scores. From [Figure 3], it can be seen that the experimental group had poorer duration pattern scores compared with control group in all the three groups. Two-way ANOVA showed a significant main effect of subject group on the DPT scores [F(1,125) = 49.9, P < 0.01]. Follow-up non-parametric independent sample t test revealed that the experimental group had significantly poorer duration pattern scores than their age-matched control group across all the three age groups [(t = 2.159 and = 2.159 and P0 < 0.05 for 30-40 years age group), (t =2.323 and = 2.323 and P < 0.05 for the 40-50 years age group), and ( t = 2.906 and = 2.906 and P < 0.05 for the age group of 50-60 years)].{Figure 3}

Speech recognition with multi-talker babble

The speech recognition scores were converted into rationalized arcsine transferred scores [14] and all the statistical analysis was done on arcsine-transferred scores. [Figure 4] shows mean and standard deviation values for speech identification scores in experimental and control group. Two-way ANOVA revealed a significant main effect of subject groups on the speech perception scores [ F(1,125) = 51.8, P < 0.01]. Follow-up non-parametric independent sample t test showed that the individual's ability to recognize speech in presence of noise was poor in noise-exposed participants compared with their age-matched control groups [(t = 4.836 and = 4.836 and P< 0.05 for 30-40 years age group), (t = 3.481 and = 3.481 and P< 0.05 for the 40-50 years age group), and (t = 2.422 and = 2.422 and P< 0.05 for the age group of 50-60 years)].{Figure 4}

To study the relationship between speech recognition scores and temporal processing abilities, Pearson's Product Moment correlation analysis and multiple regression analysis were performed. For this purpose, data from all the age groups were combined. [Table 1] and [Table 2] show the results of multiple regression and correlation analysis. From the [Table 1] and [Table 2], it can be seen that gap detection in noise, modulation detection threshold at 200 Hz and duration pattern scores are significant predictors of speech recognition scores.{Table 1}{Table 2}

Discussion

Primary goal of this research was to explore temporal processing and speech perception skills individuals who are exposed to occupational noise and not yet incurred clinically significant threshold shifts. Results revealed that both the temporal processing and speech perception skills were adversely affected in noise-exposed group. Given the present data, the observed deterioration in the temporal and speech processing skills in the noise-exposed individuals, in the presence of normal hearing sensitivity probably due to changes in the central auditory system was caused due to prolonged exposure to occupational noise. It has been reported that long-term noise may have persistent effect on brain function and behavior, even when the peripheral hearing sensitivity is within normal range. Persistent effects of long-term noise exposure on central auditory system were evaluated using auditory evoked potentials. Kujala et al. [4] assessed the performance in visuo-motor target tracking task and simultaneously recorded the mismatch negativity for /pa/ and /ka/ contrasts on healthy individuals who were exposed to high levels of occupational noise. All their subjects had hearing thresholds that were comparable to the control group. Results showed impaired syllable-discrimination in the left hemisphere of noise-exposed individuals in silence and increased N2b complex for the novel sounds. Furthermore, attention control and ability to focus on visuo-motor tasks were aberrant in noise-exposed group. These results suggest that long-term exposure to occupational noise effects both sound discrimination mechanism and attention control mechanism. Brattico et al[5] measured the neural responses in normal hearing, noise-exposed, and non-exposed participants to speech and non-speech deviants. Brain electrical source modeling suggested that speech sound contrast was lateralized to left hemisphere in non-noise-expose group but in right hemisphere in noise-exposed group. This group differences were not found for the non-speech deviants. These studies show that long-term occupational noise can have a detrimental effect on the central auditory system. This detrimental effect has been observed even when the peripheral hearing sensitivity is intact. The observed deficits in the temporal and speech processing abilities in normal hearing, noise-exposed individuals in this study may be due to compromised central auditory system in these individuals. However, we cannot totally exclude the deleterious effects of distorted cochlear input as a factor. Normal hearing sensitivity does not necessarily mean the normal functioning of the cochlea in noise-exposed individuals. Evidences from the animal research suggest that cochlear functioning can be affected even in the presence of normal-hearing sensitivity. [8] Since in this study otoacoustic emission testing was not carried out due to technical and logistic issues, it is difficult to rule out the possibility of subtle cochlear dysfunction. Kujawa and Liberman [8] reported a rapid and irreversible degeneration of spiral ganglion cells by the noise exposure which resulted in the temporary threshold shifts. Even after, hair cells and hearing sensitivity were recovered and neuronal loss persisted. Effects of such neuronal losses on auditory and speech processing are detrimental.

In general, a trend of reduced temporal processing skills was observed in individuals with noise exposure. Although gap detection thresholds and modulation detection thresholds for low modulation frequencies were not statistically different between noise-exposed and non-exposed group, nevertheless mean modulation and gap detection thresholds were slightly lower in the noise exposed group. Modulation detection thresholds for the high modulation frequency and duration pattern scores were significantly poorer in noise-exposed group compared with control group. In the auditory system, modulations are represented by phase locked neural discharges of the auditory nerve fibers to individual cycles of modulation frequency. Data from the animal research have shown that acoustic over exposure can cause acute loss of afferent nerve terminals and degeneration of cochlear nerve. This might cause disruption in the phase locking and synchronization in the discharge patterns of auditory nerve fibers causing poor modulation detection thresholds. Poor modulation detection thresholds for higher modulation frequencies suggest that noise-exposed individuals had difficulty in perceiving rapid fluctuations in the stimulus. Any complex broadband signals such as speech can be decomposed by auditory filters into relatively slow variations in the amplitude over time called envelop and relatively rapid oscillations called temporal fine structure. Importance of slowly varying temporal envelope in speech perception is well documented. [15] Recently, it has also been demonstrated that temporal fine structure plays a crucial role in hearing in the presence of background noise. [16] It is necessary to perceive the rapid oscillations to derive benefits of temporal fine structure cues. Difficulty in perceiving rapid amplitude fluctuations in noise exposed group may also pose problems in coding temporal fine structure. This may be one of the reasons for poor performance of noise-exposed individuals in speech perception measures. It is also been suggest that speech is comodulated at the rate of fundamental frequency (in this study mean fundamental frequency of the target stimulus was 211 Hz). It is important to perceive these comodulations to perceptually separate target speech and background babble as different acoustic streams. Difficulty of noise-exposed individuals in perceiving the rapid amplitude fluctuations may limit their ability perceptually segregate target and background babble. [17] Duration pattern assess the auditory sequencing abilities. Auditory short-term working memory and auditory attention are crucial for good performance on duration pattern test. There is good evidence that noise exposure impairs working memory and attention. [18] These deficits in the attention and working memory capacities might have caused poor performance in duration pattern test in noise-exposed individuals.

The second goal of this study was to see whether there is a relationship between temporal processing abilities and speech perception measures. The overall speech recognition scores were poorer in noise-exposed group compared with control group. Multiple regression analysis with speech recognition scores as dependent variable showed that temporal processing skills can account for the 26% of the variability seen in the speech recognition scores. Gap detection thresholds, modulation detection threshold for 200 Hz modulation frequency, and duration patter scores were significantly related to speech perception in noise. Overall, results suggest that speech recognition in noise was adversely affected in individuals who are exposed to occupational noise, although the peripheral hearing sensitivity was intact and this deficit in speech recognition in noise was partially accounted by the poor temporal processing abilities. While interpreting the results of this study, it should be kept in mind that we did not subgroup the participants based on the duration of noise exposure as this would reduce the number of participants in each group. Duration of noise exposure along with other conditions such as cardiological problems (arrhythmia, anemia), hypertension, hyper-cholesterol, and heavy smoking could have potentially influence the results.

In summary, results of this study indicated deterioration in temporal and speech processing abilities of individuals who were exposed to occupational noise. These deficits were observed despite normal peripheral hearing sensitivity. These results suggest that noise can cause significant distortions in the processing of suprathreshold temporal cues which may add to difficulties in hearing in adverse listening conditions.

References

1. Moore BC. Perceptual consequence of cochlear hearing loss. Cambridge: Oxford University Press; 2007.

2. American speech language and hearing association. Central auditory processing disorders: Current status of research and implications for clinical practice. Am J Audiol 1996;5:41-54.

3. Strouse A, Ashmead DH, Ohde RN, Grantham DW. Temporal processing in the aging auditory system. J Acoust Soc Am 1998;104:2385-99.

4. Kujala T, Brattico E. Detrimental noise effects on brain's speech functions. Biol Psychol 2009;81:135-43.

5. Brattico E, Kujala T, Tervaniemi M, Alku P, Ambrosi L, Monitillo V. Long-term exposure to occupational noise alters the cortical organization of sound processing. Clin Neurophysiol 2005;116:190-203.

6. Chang EF, Merzenich MM. Environmental noise retards auditory cortical development. Science 2003;300:498-502.

7. Feng Y, Yin S, Kiefte M, Wang J. Temporal resolution in regions of normal hearing and speech perception in noise for adults with sloping high-frequency hearing loss. Ear Hear 2010;31:115-25.

8. Kujawa SG, Liberman MC. Adding insult to injury: cochlear nerve degeneration after "temporary" noise-induced hearing loss. J Neurosci 2009;29:14077-85.

9. Kumar AU, A V S. Temporal processing across different age groups. J Am Acad Audiol 2011;22:5-12.

10. Grassi M, Soranzo A. MLP: a MATLAB toolbox for rapid and reliable auditory threshold estimation. Behav Res Methods 2009;41:20-8.

11. Green DM. Stimulus selection in adaptive psychophysical procedures. J Acoust Soc Am 1990;87:2662-74.

12. Green DM. A maximum-likelihood method for estimating thresholds in a yes-no task. J Acoust Soc Am 1993;93:2096-105.

13. Musiek FE, Baran JA, Pinheiro ML. Duration pattern recognition in normal subjects and patients with cerebral and cochlear lesions. Audiology 1990;29:304-13.

14. Studebaker GA. A "rationalized" arcsine transform. J Speech Hear Res 1985;28:455-62.

15. Drullman R, Festen JM, Plomp R. Effect of temporal envelope smearing on speech reception. J Acoust Soc Am 1994;95:1053-64.

16. Moore BC. The role of temporal fine structure processing in pitch perception, masking, and speech perception for normal-hearing and hearing-impaired people. J Assoc Res Otolaryngol 2008;9:399-406.

17. Snyder JS, Alain C. Toward a neurophysiological theory of auditory stream segregation. Psychol Bull 2007;133:780-9.

18. Salame P, Baddeley AD. Disruption of short-term memory by unattended speech: implications for the structure of working memory. J Verb Learn Verb Beh 1982;21:150-4.