Immediate effects of signal processing on memory systems

• Traditionally, hearing instrument signal processing schemes are evaluated where the speech recognition tests in noise are most sensitive, i.e. around 50 percent correct. However, such testing typically leads to ecologically unrealistic negative signal-to-noise ratios. Therefore, memory performance has been suggested as a more ecologically valid outcome measure in conditions with interfering background noises, and where participants perform closer to 100 percent correct. In a recent study by Sarampalis et al. (2009) they used an Ephraim-Mallah noise reduction algorithm and found beneficial effects for normal hearing participants on memory function. However, such memory effects have not been shown for hearing impaired participants. The current study on hearing-impaired subjects evaluated ‘aggressive’ ideal binary masking (IBM, see Wang, et al., 2009), as a proof-of-concept not only for improving speech comprehension but also for its potential effects of masking on memory systems. Furthermore, binary masks estimated from directional microphones (Boldt et al., 2008) were used as a realistic version of binary mask processing. Results show significant main effects of noise type and noise reduction. Noise reduction techniques improved the memory performance in the four-talker babble background, while no such improvement was observed in the steady-state noise background. Participants with higher reading span scores performed significantly better in the memory task, but only in the four-talker babble background.The study demonstrates the binary-masking noise reduction technique helps freeing up cognitive resources and hence enhances memory task performance in the four-talker babble background for hearing-impaired listeners with high reading span performance. The findings suggest that there is a cognitive challenge for the hearing aid industry to improve signal processing options for individuals to improve immediate perception and encoding into memory.

Publikations- och innehållstyp

vet (ämneskategori)

kon (ämneskategori)