Does use of individual acoustical transforms in hearing aids have the potential to improve the perceived sound quality? Scan to access poster pdf
Niels Søgaard Jensen, Søren Laugesen, Filip Marchman Rønne, Julie Hefting Pedersen Eriksholm Research Centre, Rørtangvej 20, DK-3070 Snekkersten, www.eriksholm.com Contact: Niels Søgaard Jensen,
[email protected] Presented at the 12th Congress of The European Federation of Audiology Societies, Istanbul, Turkey, May 27-30, 2015
A potential soundquality benefit of using individualised rather than standardised acoustical transforms was demonstrated in this study
Introduction
Results
The (unavoidable) acoustical consequences of placing a hearing aid in the ear are normally compensated for by applying three different types of acoustical transforms (ATs) in the signal-processing chain [1]:
The paired-comparison data were organized in preference matrices and analysed using a Bradley-Terry-Luce (BTL) model [2]. The outcome was a ratio-scale BTL score for each condition, and a corresponding 95% confidence interval. Table 1 shows the total preference matrix (pooling all data), while Figure 3 shows the BTL scores for each sound sample.
• Open-Ear Gain (OEG ), a.k.a. Real-Ear Unaided Gain (REUG) • Microphone Location Effect (MLE) • Real-Ear-to-Coupler Difference (RECD) These ATs are typically standardised values, based on either manikin or (average) real-ear measurements. In this study, we investigated whether the perceived sound quality of a hearing aid may be improved if individualised rather than standardised ATs are used.
OEGind HA0stand OEGstand
OEGind
HA0stand
OEGstand
0
157 (58%)
155 (57%) 144 (53%)
113 (42%) 115 (43%) 157 (58%)
0 126 (47%) 175 (65%)
0 180 (67%)
OEGind,lowres Average 113 (42%) 95 (35%) 90 (33%)
52% 43% 41%
Table 1. Preference matrix based on all data. A significant preference for a row condition over a column condition is indicated.
Methods
OEGind,lowres
Four different test conditions, corresponding to various degrees of AT individualisation, were simulated in a sound-quality experiment where all stimuli were presented via insert phones:
• For all sound samples, OEGind,lowres is the ‘winner’ (most preferred). • OEGind is preferred over HA0stand and OEGstand, but only for sound samples including music. A different preference pattern was observed for sound samples including speech, see the left and right plots in Figure 3. • Large individual variations in preference patterns were observed.
• OEGind: Listening through own OEG, estimated individually from real-ear measurements with frontal sound incidence (~ LTM,unaided-LFF, cf. Figure 1).
Figure 1. Sound pressure levels relevant for hearing-aid fittings and real-ear measurements.
0
63%
Figure 3. Normalised BTL scores and 95% confidence intervals for the four test conditions, shown for each of the five sound samples.
• HA0stand: Listening through a hearing aid (Oticon Alta Pro RITE) set to provide 0-dB insertion gain based on use of standardised ATs (~ LTM,aided-LFF). • OEGstand: Listening through a standardised version of OEG, as used in the fitting software. • OEGind,lowres: Listening through a low-resolution version of OEGind. Figure 2. The frequency responses of the four test conditions, shown for one individual testsubject ear.
Test approach • N = 18 normal-hearing test subjects. • A/B paired-comparison test paradigm (including all 6 condition pairs). • 5 sound samples - Classical music, jazz, rock, speech in quiet, dialogue in canteen. • 3 replications (i.e., 6 x 5 x 3 = 90 comparisons per subject). • Order of condition pairs and sound samples was counterbalanced. • Test-subject task: For a given sound sample, switch between conditions A and B and select the one offering the preferred sound quality.
References
[1] Dillon (2012). Hearing Aids. New York: Thieme. [2] Wickelmaier & Schmid (2004). A Matlab function to estimate choice model parameters from paired-comparison data. Behav. Res. Methods Instrum. Comput. 36.
Discussion • The higher BTL scores observed for the conditions based on individual ATs, as compared to the conditions based on standardised ATs, indicate a potential for improving the sound quality in hearing aids by use of individualised ATs. However, the benefit will depend on the input signal. • The higher BTL scores observed for OEGind,lowres, as compared to OEGind, may indicate a preference for a ‘smoother’ frequency response. However, a more likely cause may be an (unintended) high-frequency boost introduced by the method used to generate the lowres condition. • The individual variation indicates that not all persons will benefit from use of individualised ATs, e.g. due to small differences between individual and standardised ATs or unreliable individual measurements. • Another study is needed to investigate whether the results transfer to hearing-aid users when hearing-loss compensation is included in the signal-processing chain.
