Supra-threshold hearing loss and wide dynamic range compression

Abstract

Cochlear hearing loss leads to deficits in many perceptual aspects, including threshold elevation (loss of sensitivity), reduction of auditory dynamic range (loudness recruitment) and problems with resolving sounds (loss of supra-threshold hearing). An ongoing debate is whether assessment of the individual supra-threshold hearing provides information which is likely to have an important influence on clinical decisions on signal processing strategies. The aim of this thesis was to investigate different methods of assessing the supra-threshold hearing and on this basis motivate configurations in a wide dynamic range compression hearing aid.

A method to assess the supra-threshold hearing has previously been developed; Psychoacoustical Modulation Transfer Function (PMTF). This method measures masked short-tone thresholds at the peak and in the valley in a 100 % intensity modulated octave band of noise. The effects of duration and level were also investigated for a simultaneous masking task, analogous to the condition of the threshold at the peak. The masked thresholds and temporal integration for both normal-hearing and hearingimpaired listeners vary as a function of level. This supports the view that the task is highly influenced by peripheral compression. By measuring PMTF as a function of the perceived loudness of the modulated noise a comparison was possible at equal loudness categories between a group of listeners with presumed good supra-threshold hearing and normal-hearing listeners. Listeners with hearing loss showed lower than normal peak-to-valley threshold difference at loudness ratings below their comfortable loudness level. An individual fitting rationale was defined and evaluated in a field trial using an experimental hearing aid configured as a slow-acting compressor. With regard to speech recognition in noise and sound quality ratings and self-assessment, the result does not support the use of a fitting rationale based on the residual nonlinearity in the inner ear instead of a threshold-based gain prescription.

In a series of four experiments, the effects of presentation level, compression characteristics and time constant on the benefit from compression compared to linear were studied using a real-time simulated three-channel compressor. Speech recognition was measured with sentences in fluctuating speechshaped noise. Results showed that only one-third of the hearing-impaired listeners showed benefit from fast-acting compression across all presentation levels. An inverse relationship was found between benefit from fast-acting compression and the severity of the supra-threshold hearing loss. The choice of compression speed affected speech recognition in a fluctuating noise for both hearing-impaired and normal-hearing listeners. The normal-hearing listeners showed the largest benefit from fastacting compression and the effect of compression speed for this group was predicted from acoustical factors.

In three subsequent experiments hearing impairment was simulated in normal-hearing listeners to explore the mechanisms behind the lack of benefit for most listeners. Evidence was found that moderate fast-acting compression could be beneficial in fluctuating noise when the speech was not limited by hearing threshold, was presented at negative signal-tonoise ratios or contained only minimal spectral cues. In conclusion, the data indicates that fast-acting compression should be used with caution. Compressors of intermediate and slow speed are more appropriate because it will preserve the temporal contrast.