The First Cadenza Challenge: Perceptual Evaluation of Machine Learning Systems to Improve Audio Quality of Popular Music for Those with Hearing Loss.
Scott Bannister, Jennifer Firth, Gerardo Roa-Dabike and 8 others
PMID 41618122WHAT IT FOUND
No machine learning system improved music audio quality for hearing aid users beyond the original unprocessed signal.
Clarity and low distortion drove quality ratings. Systems that boosted volume or bass for severe loss often added harshness, making simple or baseline processing preferable.
Key findings
01The best-performing machine learning systems matched the baseline and the 'do nothing' system, meaning no entrant system surpassed the original unprocessed audio in perceived quality.
02Perceived audio quality was driven primarily by clarity and low distortion, with harshness, frequency balance, and liking playing smaller roles.
03Systems that increased volume or altered bass for more severe hearing loss often introduced clipping and distortion, causing quality ratings to drop as hearing loss worsened.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
Participants listened at self-adjusted volumes, which could vary between individuals and affect ratings. Seven participants heard samples processed for a similar, rather than their exact, audiogram due to data availability. The study used only popular music genres from a specific dataset, so results may not apply to classical or other musical styles. The 'do nothing' system was not amplified, which may have influenced comparisons with systems that applied hearing loss compensation.
Declared interests
Funded by the Engineering and Physical Sciences Research Council (Project Reference: EP/W019434/1). The authors declared no potential conflicts of interest.
The easy way to misread this
Do not conclude that machine learning remixing is useless for hearing aid users. The study tested specific algorithms on popular music without hearing aids; newer systems or different genres may yield different results. Also, note that the 'do nothing' system lacked amplification, so its high ratings do not mean unamplified audio is better for all listening scenarios.
Summarised by AI from the full paper, without a clinician reviewing it. Check it against the source before it changes what you do. Read it on PubMed →