Assessing the effects of hearing-aid dynamic-range compression on auditory signal processing and perception

The consequences of a sensorineural hearing loss include decreased sensitivity to low-intensity sounds as well as supra-threshold processing deficits, such as loudness recruitment and degraded spectral and temporal resolution. To counteract this, most hearing aids apply level-dependent amplification, such as wide dynamic-range compression (WDRC).

WDRC can improve audibility of speech while avoiding loudness discomfort. However, it can also lead to many distortions, such as the reduction of temporal and spectral contrasts in the signal. Therefore, reports on the benefit of WDRC have been mixed and the choice of hearing-aid parameters, such as compression time constants, have remained controversial.

The aim of the presented work was to study the effects of hearing-aid amplification on auditory signal processing and perception and its influence on speech recognition. Based on these findings, new compensation strategies were proposed. In the first main study of this PhD project, the effects of linear vs multichannel compressive hearing-aid amplification on psychoacoustic metrics of temporal and spectral resolution were investigated.

A group of normal-hearing (NH) and hearing-impaired (HI) listeners was considered. A computational model of auditory signal processing was used to evaluate the effects of hearing loss and hearing-aid processing on temporal and spectral masking. It was found that the decay of forward masking, an indicator of temporal resolution, can be improved in the HI listeners when amplification is applied.

No differences were found between linear amplification and compression. The amount of spectral masking, on the other hand, was reduced when compression was applied, but not with linear amplification. This was due to an enhancement of the signal-tonoise ratio (SNR) across frequency. The second study investigated the effects of compression release time on consonant recognition by HI listeners in temporally fluctuating noise.

Two compression conditions were considered: fast-acting (10 ms release time) and slow-acting (500 ms release time). Several objective metrics were used to evaluate the effects of compression on stimulus audibility, SNR and temporal envelope distortions. It was found that fast-acting compression provided a higher consonant output level (i.e. improved audibility), and a higher output SNR than the slow-acting system, while introducing negligible envelope distortion.

This, in turn led to improved consonant recognition performance in the HI listeners. The third study focused on the distortions that fast-acting compression can introduce when applied in realistic speech-in-noise scenarios. Rapid gain fluctuations applied by such processing introduce modulations to the background and amplify the noise, thus reducing the output SNR.

A novel compression scheme was proposed, that selectively applies fast-acting compression to the speech-dominated time-frequency units and linearizes the processing of the background by applying compression with a long release time. When compared to conventional fast-acting processing, the proposed strategy was found to provide equally effective compression of the speech signal (often linked to improved audibility), while reducing the distortion in the modulation domain and improving the output SNR.