Perception of Temporally Processed Speech by Listeners with Hearing Impairment

Abstract

Although SPC-processed speech was not preferred by listeners with hearing loss, the listeners with a more moderate degree of impairment could not differentiate the unprocessed sentences from the SPC processed sentences. Speech intelligibility was not improved by SPC processing.

Analysis of Spatiotemporal Pattern Correction Using a Computational Model of the Auditory Periphery

“…We found that SPC increased the error in the absolute spatiotemporal response patterns and synchronized rate patterns, which could explain the poor experimental results obtained by Shi et al (2006) and Calandruccio et al (2007). However, these degradations were smallest at low SPLs and SPC provided the best peak-lag cross-correlation improvement at low levels.…”

“…The delays introduced by the SPC scheme were shown by Calandruccio et al (2007) to be near the range of 0.4 to 0.6 milliseconds for a vowel; these small values, along with the listening results described earlier, might call into question the importance of correcting for group delay because the literature cites larger detection thresholds for across-frequency delay differences. Specifically, Kates and Arehart (2005) found that it is unlikely that listeners with hearing impairment would detect a difference in group delay of around 4 milliseconds between the low and high frequencies.…”

“…This suggests that SPC disrupts certain aspects of the fine temporal structure of the AN response pattern, particularly as sound level increases. Shi et al (2006) presented their stimuli at 65 dB SPL, which in some cases could be lower than the 30 dB re: speech reception threshold level at which Calandruccio et al (2007) presented their stimuli. This, together with the fact that fine temporal structure appears disrupted by SPC as sound level increases, may have contributed to relatively better intelligibility results found by Shi et al (2006).…”

“…Temporal shifts are applied in integer sample numbers, where the sampling frequency is Fs = 32 kHz. Shi et al (2006) and Calandruccio et al (2007) inserted delays at various SPC strengths, which scaled the amount of delay inserted based on the assumed percentage of remnant cochlear nonlinearity. Since we used the auditory model to characterize the hearing loss, the retained nonlinearity was known and we inserted the full difference in group delay as prescribed by the healthy and impaired C1 filters at each filterbank channel.…”

“…They found that the difference in hearing-impaired listeners' preference level for minimally processed versus unprocessed sounds was not significant for the group of subjects, but the two listeners who had the greatest degree of hearing loss preferred some amount of SPC processing. Subsequently, Calandruccio et al (2007) evaluated a version of the SPC scheme that used rectangular filters in place of the gammatone filters when implementing the analysis-synthesis filterbank. Their study found that listeners preferred unprocessed sounds to SPC processed sounds.…”