Hearing aid evaluation for music: Accounting for acoustical variability within and between music genres

Authors: Christophe Lesimple1, Volker Kuehnel1, Kai Siedenburg 2

1Sonova AG
2Department of Medical Physics and Acoustics, Carl von Ossietzky University of Oldenbur

Background: Music is an important signal class for hearing aids, yet research has used musical stimuli in rather ad hoc ways. Comparisons and evaluations of different hearing aid strategies with music signals usually include different musical genres in their research design, based on the assumption that musical genre may drive some of the variability in the results. However, acoustical differences between musical genres and the way in which hearing aid processing is affected by such differences has not been investigated systematically. This raises the question whether findings for single excerpts from one song per genre are generalizable and whether hearing aid processing has different acoustical effects in different musical genres.

Methods: A database of recorded musical pieces was compiled based on previous research projects, comprising a total of 1400 pieces from 12 different genres. From each musical piece in the database, three 20-sec excerpts were drawn randomly. These excerpts were then processed by a hearing aid simulator with two different dynamic range compression approaches.

Results: Principal component analysis on the extracted acoustical features indicated a distinct positioning of classical musical genres compared to more modern genres but there was also considerable overlap. Changes in acoustical features coming from amplification affected mainly features related to energy and spectral shape. For instance, whereas spectral centroids and RMS standard deviation for genres such as piano or orchestral music were strongly affected by hearing aids amplification, amplification had markedly little effects on genres such as metal or rap.

Conclusions: Selection of audio samples by genre should be supported by an acoustical analysis of their representativeness. Accounting for within genre variability in the test design would also help to generalize findings of studies beyond the selected test material.