Perceptual Effects of Adjusting Hearing-Aid Gain by Means of a Machine-Learning Approach Based on Individual User Preference.

The hearing aids were mounted on a manikin, so participants did not hear their own voice amplified, which likely influenced low-frequency gain preferences. The study only adjusted static gain in three bands; dynamic processing features like compression and noise reduction were fixed, limiting the potential for benefits in speech clarity or complex noise. The sample size was small (N=20) and consisted of experienced hearing-aid users who were familiar with listening tests, which may not generalize to new users. Individual variation was high, and some participants experienced disadvantages with the self-adjustment method, suggesting it may not be suitable for all users. The study design imposed specific listening intentions on participants, which may not have matched their actual real-world needs in those scenarios.

The study was conducted at SenseLab, FORCE Technology. The paper does not explicitly state the funding source or declare conflicts of interest in the provided text, but the experimental design was inspired by a prior study in Widex's own lab, and the commercial hearing aid (Widex EVOKE) and its feature (SoundSense Learn) were used.

Do not assume that machine-learning self-adjustment will improve speech clarity or listening comfort in noisy environments with speech. The study found no significant benefit for these domains, and individual responses varied widely, with some users experiencing worse outcomes.