PP505 - Effects of the Spectral and Temporal Character of Noise on Speech Recognition

Abstract:
Detecting acoustic signals in background noise is an important communicative task. However, the question of how that is accomplished is not fully understood. Consequently, the most appropriate choice of speech-in-noise tests used clinically has not reached consensus. Of significance is that speech recognition ability sharply declines in patients with even mild to moderate hearing loss. We examined the role of more precisely defined spectrotemporal noise compositions on speech recognition in normal and hearing-impaired listeners. Findings include large differences in speech intelligibility between stationary and fluctuating noises depending upon the spectrotemporal noise composition in subjects with normal and mildly impaired hearing.

Summary:
Rationale/Purpose
How does speech recognition in background noise depend on the spectro-temporal composition of the noise envelope? Several noise-types are in use by audiologists to assess speech recognition ability in patients including stationary and fluctuating noises. Some of them take into account characteristic temporal aspects of typical variations in speech envelopes but few take into account the varying spectral modulations as imposed, for example, by the speech format structure. Few systematic analyses of the combined role of spectral and temporal modulations of the noise envelope have been performed. Our objective was to compare the sensitivity and reliability of speech intelligibility testing for structurally different noises in normal subject as an initial step for assessing their sensitivity and reliability in detecting early signs of hearing impairment.

Methods
We designed innovative broadband dynamic moving noises (DMRs) with well-defined temporal and spectral modulations of the stimulus envelope. The overall DMR spectrum matched that of speech. Two ranges of spectral (0-0.3 rip/oct and 0.3-2 rip/oct) and temporal modulations (0-5 Hz and 5-20 Hz) were selected and speech recognition in normal hearing subjects was tested for all four possible combinations plus a noise containing the full range of modulations. Depth of modulation was 40 dB and the overall stimulus level was 65 dB SPL. Normal hearing subjects (age 22-45 years; 6 female, 7 male) were tested with five DMR noises and seven noises commonly used in audiology. We used the American English matrix sentence test with a fixed noise level to determine the 50% correct response performance (Kollmeier et al., 2015). Statistical analysis included repeated measure ANOVA and post-hoc t-tests with Bonferroni correction.

Results
The 50% sentence intelligibility level varied significantly between the noises. Fluctuating noises with low temporal modulations had up to 20 dB better detection thresholds than stationary noises and noises with faster temporal modulations. Low-frequency temporal modulation noises, however, showed a wide range of inter-subject variability despite a high test-retest reliability for the individual subjects. It remains to be determined whether this variability is due to cognitive aspects of the test or an indication of some early signs of underlying hearing disorder not reflected in the audiograms (e.g., synaptopathy). Lower thresholds for noises with the lower range of spectral modulations was also evident but had a smaller effect size on the performance thresholds than differences in temporal modulations.

Brief Summary of Clinical Takeaways:
Despite a wide range of choices available for speech-in-noise testing, it is still not evident what type of noise would be most sensitive and reliable to determine early signs of hearing impairment. Our comparison between different noise types shows significant differences in patient performance and across-patient reliability. Potential benefits of DMR noises in diagnosing early signs of hearing impairment in patients with normal or borderline audiograms remains to be seen.

Kollmeier, B., A. Warzybok, S. Hochmuth, M. A. Zokoll, V. Uslar, T. Brand, and K. C. Wagener. The Multilingual Matrix Test: Principles, Applications, and Comparison across Languages: A Review. International Journal of Audiology 54 (Suppl. 2): 3–16, 2015.