Evaluation of Speaker-Conditioned Target Speaker Extraction Algorithms for Hearing-Impaired Listeners.
Ragini Sinha, Ann-Christin Scherer, Simon Doclo and 2 others
PMID 40785510WHAT IT FOUND
One speaker-extraction algorithm consistently improved hearing-impaired listeners' speech recognition and reduced listening effort across noise levels, while a second algorithm offered no benefit.
These gains were identical with and without hearing aid amplification, suggesting the algorithm does the work.
Key findings
01Algo-2 consistently improved speech recognition thresholds and reduced perceived listening effort for hearing-impaired listeners, whereas Algo-1 provided no benefit over unprocessed audio.
02The benefits of the successful algorithm were independent of hearing loss compensation, with no significant difference between aided and unaided conditions.
03Hearing-impaired listeners experienced greater reductions in listening effort and improvements in speech recognition with the successful algorithm compared to normal-hearing listeners.
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
The study included only 15 participants with mild to moderate, relatively symmetric hearing loss, so results may not generalize to those with severe, profound, or asymmetric hearing loss. The algorithms were trained on English speech but evaluated on German speech, which may affect the generalizability of the artifact profiles. The study focused on static mixtures of one or two interfering speakers and did not evaluate performance in the presence of background noise or reverberation, which are common in real-world clinical and social settings. The successful algorithm (Algo-2) currently has high computational complexity, meaning it is not yet suitable for implementation in standard hearing aid hardware.
Declared interests
The authors received funding from the Lower Saxony Ministry of Science and Culture, the Volkswagen Foundation, and the Deutsche Forschungsgemeinschaft. No commercial conflicts of interest regarding the specific algorithms were declared.
The easy way to misread this
Do not assume that hearing aid amplification is necessary for these specific AI extraction features to work. The study found that the benefit of the successful algorithm was identical with and without hearing loss compensation, suggesting the algorithm's effectiveness relies on its ability to separate voices rather than on restoring audibility.