A definition of lyric intelligibility by hearing aid users: A focus group study

Authors: Scott Bannister¹, Alinka Greasley¹, Michael Akeroyd², Jennifer Firth², William Whitmer², Jon Barker³, Gerardo Roa Dabike³, Trevor Cox⁴, Bruno Fazenda⁴, Simone Graetzer⁴, Rebecca Vos⁴

¹University of Leeds
²University of Nottingham
³University of Sheffield
⁴University of Salford

Hearing aids (HAs), optimized for speech, can be problematic for music. A key difficulty relates to hearing lyrics in music, which is a different experience from hearing spoken speech. To improve the music listening experiences of HA users, such as through the machine-learning challenges of the CADENZA project, it is vital to evaluate improvements in ways relevant to HA users. To do so, we must first comprehensively understand what lyric intelligibility means to this heterogeneous population. Ten bilateral HA users (5 females, 5 males), with hearing loss ranging from mild to moderately severe, discussed their experience of lyrics in music, across two online focus groups. Discussions involved listening to music excerpts and developing a definition and measurement of lyric intelligibility. Participants led the discussions, with researchers facilitating the sessions. Lyric intelligibility was defined by participants as “how clearly and effortlessly the words in the music can be heard”. Good lyric intelligibility involved clarity in lyrics and lack of interference; poor lyric intelligibility referred to poor balance between lyrics and music, masking of lyrics by other voices or instruments, or mumbled or quick singing styles. Musical genres, singing styles and balance in production/arrangement were considered important factors for lyric intelligibility. Additionally, lyric intelligibility was distinct from ‘understanding’ lyrics (e.g., comprehension of lyrics in another language). Results advance our understanding of how HA users experience and perceive lyrics in music. This has implications not only for evaluating potential improvements, such as in the CADENZA machine-learning challenges, but also in developing novel signal processing strategies for hearing-assistive technologies to improve music listening experiences for people with hearing loss.