Instrumental Quality Predictions and Analysis of Auditory Cues for Algorithms in Modern Headphone Technology.
Thomas Biberger, Henning Schepker, Florian Denk and 1 others
PMID 33739186WHAT IT FOUND
Models that track one-ear frequency changes predicted listeners' quality ratings of distorted sounds from smart headphones and hearing devices best.
Both-ear-only predictions were weak. This is an algorithm evaluation, not patient outcome evidence.
Key findings
01The two models that predicted listener quality ratings most closely were GPSM q and D.
02A model that used only both-ear cues gave weak predictions for the hear-through database, indicating that both-ear distortions were not the main driver of perceived quality there.
03Models that track changes in the frequency content of sound predicted listener quality ratings best, while models that do not track such changes performed poorly or moderately.
STILL TO COME
How it was doneWhat they found
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
The study used existing recordings and normal-hearing listeners, not patients with hearing loss, so it does not show how people who use hearing aids or hearables would rate the sounds. The three databases used different listener groups, with 15, 15, and 17 normal-hearing participants, so model performance was not checked in one common group. Several models were trained on different signal types or distortion ranges, so poor performance may reflect a mismatch between the training data and these hearable distortions. The paper reports how well models matched listener ratings, not clinical outcomes, so it cannot tell a therapist whether a device helps a patient.
Declared interests
The authors declared no conflicts of interest. The work was supported by the German Research Foundation.
The easy way to misread this
Do not use this paper to choose a hearing aid for a patient. It tested computer models against normal-hearing listeners' ratings of recorded sound quality, not clinical outcomes.