Objective Evaluation of a Deep Learning-Based Noise Reduction Algorithm for Hearing Aids Under Diverse Fitting and Listening Conditions
Bibliographic record
Abstract
This study evaluated a deep-neural-network denoising system using model-based design, comparing it with adaptive filtering and beamforming across various noise types, SNRs, and hearing-aid fittings. A KEMAR manikin fitted with five audiograms was recorded in reverberant and non-reverberant rooms, yielding 1,152 recordings. Speech intelligibility was estimated using the HASPI from 1,152 KEMAR manikin recordings. Effects of processing strategy and acoustic factors were tested with model-based within-device design that account for repeated recordings per device/program and fitting. Linear mixed model results showed that the DNN with beamforming outperformed conventional processing, with strongest gains at 0 and +5 dB SNR, moderate benefits at -5 dB in low reverberation, and none in medium reverberation. Across SNRs and noise types, the DNN combined with beamforming yielded the highest predicted intelligibility, with benefits attenuated under moderate reverberation. Azimuth effects varied; because estimates were derived from a better-ear metric on manikin recordings. Additionally, this paper reports comparisons using metrics of sound quality, for an intrusive metric (HASQI) and the pMOS non-intrusive metric. Results indicated that model type interacted with processing and acoustic factors. HASQI and pMOS scores increased with SNR and were moderately correlated (r² ≈ 0.479), supporting the use of non-intrusive metrics for large-scale assessment. However, pMOS showed greater variability across hearing aid programs and environments, suggesting non-intrusive models capture processing effects differently than intrusive metrics. These findings highlight the promise and limits of non-intrusive evaluation while emphasizing the benefit of combining deep learning with beamforming to improve intelligibility and quality.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".