lars_round
Lars Bramsløw

Research Engineer, PhD, Project Leader

> Learn more

mail@eriksholm.com

In a paired-comparison experiment, the sound quality of (simulated) hearing aids representing different degrees of acoustical-transform individualization was assessed. Hearing aids which were individualized according to a REIG target were preferred over hearing aids individualized according to a REAR target.

Background

The acoustical difference between listening through the open ear and through a hearing aid is normally compensated for by including various acoustical transforms (ATs) in the prescription of the hearing-aid gain. ATs are often referred to as the microphone location effect (MLE), the open ear gain (OEG), and the real ear to coupler difference (RECD). Even though large individual variations in ATs are known to exist, the applied ATs are typically standardized values.

Study

The aim of this study was to investigate whether use of individual ATs has the potential to improve the perceived sound quality of hearing aids. Real-ear measurements (REM) were used to take individualized ATs into account. Two strategies of REM-based fittings were included: A real-ear insertion gain (REIG) approach, which defines the SPL at the eardrum relative to the unaided ear, and a real-ear aided response (REAR), which defines the absolute SPL at the eardrum.

18 test subjects participated. They were all normal-hearing in order not to have to compensate for a hearing loss and thereby avoid possible effects caused by a given fitting rationale. Five different hearing-aid gain settings, with different degrees of individualization of ATs, were simulated and sound was presented to the test subjects via insert phones. The five settings are listed in the table below.

 

HA0REIG
 
Mimicking a 0-dB insertion gain HA fitted with individual ATs according to a REIG target.
 
HA0avg
Was meant to mimic a 0-dB insertion gain HA fitted according to average ATs. However, due to a programming mistake this condition must be disregarded.
 
HA0REAR1

Mimicking a 0-dB insertion gain HA fitted with individual ATs according to a REAR target. Based on the automatic AutoFit function in the Genie HA-fitting software together with the Affinity REM system.

HA0REAR2

Mimicking a 0-dB insertion gain HA fitted with individual ATs according to a REAR target. Derived directly from the standard OEG used in the Genie fitting software.

HA0REIGlowres 

Similar to HA0REIG, except that the individual ATs were realized with a frequency detail similar to what is available in the Genie fitting software.

The sound quality of the five settings was assessed using an A/B paired-comparison approach where all pairs of settings were compared when listening to various speech and music samples. For each A/B pair, the task of the test subject was to pick the setting with the preferred sound quality.

Results

The preference data were analysed using a so-called Bradley-Terry-Luce (BTL) model. The resulting BTL scores were normalized such that the score for a given setting could be interpreted as the likelihood of preferring that setting in comparison with another (randomly selected) condition. The result for all sound samples pooled is shown in Figure 1.

The HA0avg score should be disregarded due to an error. Hence, the main result to be taken from Figure 1 is that HA0REIG is preferred over both HA0REAR1 and HA0REAR2, which suggests that in consideration of sound quality a REIG target-approach should be preferred over a REAR-target approach. The (surprising) fact that the highest preference was observed for the low-resolution HA0REIGlowres setting was most likely due to a minor high-frequency boost at 6-8 kHz caused by (unintended) artefacts of the ‘lowres’ procedure.
  • benefits of individualized acoustical transforms fig 1 BTL
    Figure 1. BTL scores and 95% confidence intervals for each of the five test settings. Higher scores indicate better sound quality. The y-axis is logarithmic due to the ratio-scale nature of the BTL scores
    Figure 1. BTL scores and 95% confidence intervals for each of the five test settings. Higher scores indicate better sound quality. The y-axis is logarithmic due to the ratio-scale nature of the BTL scores
The trend shown in Figure 1 was even more prominent when music samples were analysed in isolation, while the trend was less clear for speech samples only. This difference was likely caused by speech intelligibility being used (by some subjects) as criterion – rather than sound quality – when assessing the speech samples. Also, it should be noted that the individual patterns of preference for some test subjects deviated considerable from the pattern shown in Figures 1. This is not surprising, due to the individual nature of the ATs, which means that the contrasts among the different settings were not the same for all test subjects

Conclusions

The study indicated that hearing aids individualized according to a real-ear insertion gain (REIG) target were preferred over hearing aids individualized according to a real-ear aided response (REAR) target. Comparison to non-individualized hearing aids was not possible due to a programming mistake. Further research is needed to investigate whether the same effects are present for listeners with hearing impairment.

Further reading

Laugesen S, Jensen NS, Rønne FM, Pedersen JH (2016). Can individualized acoustical transforms in hearing aids improve perceived sound quality? Speech perception and auditory disorders. Proceedings of the 5th International Symposium on Audiological and Auditory Research, ISAAR, Denmark.