Abstract
The influence of several compression parameters on speech intelligibility in stationary and fluctuating speech-shaped noise was systematically investigated. The experiment was designed to investigate possible interaction effects between compression parameters on the speech reception threshold (SRT) in noise. Experimental conditions included all combinations of compression ratio (CR=1/2, 2/2, 2/3, 3/3
... read more
for low/high frequencies), attack and release times (Ta/Tr=4/4, 4/40, 40/40, 4/400, and 40/400 ms), and number of channels (NC=1, 2, and 6). Twenty subjects with moderate sensorineural hearing loss took part in the extensive experiment. For stationary noise the results indicate that most compression settings yielded lower scores than those obtained with linear amplification. Single-channel compression resulted in the worst speech intelligibility whereas the results for two-channel and six-channel compression were roughly equal to each other. A compression ratio of 3/3 (CRlow/CRhigh) gave worse speech intelligibility than lower compression ratios. The combination of largest time constants (Ta/Tr =40/400 ms) gave best results. For stationary noise two significant interaction effects were found: NC* T and NC*CR* T. Overall, the best result was obtained for two-channel compression CP( 2, 2/3, 40/40) (_SRT =-0.7 dB). For six-channel compression the best speech intelligibility was achieved with CP( 6, 2/2 + 2/3, 40/400) (_SRT =-0.4 dB) and the best result with single-channel compression was equal to that with linear amplification. For fluctuating noise, the results showed that compression had only a small influence on speech intelligibility. Although multi-channel compression is known to reduce spectral and temporal contrasts, its effect on speech intelligibility in fluctuating noise was limited. For fluctuating noise, only one significant result was obtained: overall, six-channel compression led to significantly worse speech intelligibility than compression with NC=1 or 2. No significant interaction effects were found. For fluctuating noise, the detrimental effects of compression were smaller than for stationary noise and the variance in the SRT results was larger. This suggests that stationary noise is more suited than fluctuating noise for measuring the effect of compression on speech intelligibility. Analyses for the data of individual subjects showed that for our subjects the effect of compression was rather small with respect to the within subject variability. Additionally the measurement error was larger for subjects with worse speech intelligibility in noise. The effect of compression was not significantly related to standard audiometric characteristics such as hearing threshold or dynamic range. However, for stationary noise, a significant correlation was found between the effect of compression and the SRT for linear amplification for the lowest two compression ratios (CR=1/2 and CR=2/2) only. Moreover, for this noise type, the best results obtained with compression improved significantly for subjects with a worse SRT. These correlations tentatively suggest that speech in stationary noise might be more useful for determining a possible benefit of compression than pure-tone audiometric threshold or dynamic range. In contrast, most current fitting rationales are threshold based.
show less
