Discussion sessions II
Chairpersons
Discussion sessions II
Friday, June 13th 2025, morning
I. Hardware & Acoustics
d) Hearing devices: Performance & Benchmarks
Proposed session chairs: Stefan Launer (Phonak), Sébastien Santurette (Oticon)
Short description: For many years, hearing aids have been available across various price segments that differ in terms of technology levels. While expensive models come with more advanced algorithms and functionality, the extra benefit they are meant to provide has often been absent in independent research studies. Arguably, the use of ineffective test procedures has played a role for this. For the sake of promoting more effective studies, can we define better methods and establish meaningful benchmarks?
Agenda (will be coming soon)
II. Audiology
e) Auditory profiles and Hearing devices: Feasibility of individualization through Big Data.
Proposed session chairs: Kirsten Wagener (Hörzentrum Oldenburg), N.N. (GN)
Short description: Big-data generated auditory profiles allow for classification of patient cases into a limited number of distinct classes of hearing impairment and needs for aural rehabilitation. Can we use these classes for a better fitting of hearing aids? How to weigh the individual factors (e.g., experience with hearing aids, Noise haters vs. distortion haters vs. amplification junkies vs. bass lovers) against the pure audiological facts? Can EMA, portable hearing labs, and a virtual hearing clinic help us to achieve a better prescription/first fit than to rely on the current “try and error” approach?
Agenda (will be coming soon)
III. Signal Processing and AI
f) Build the HA as a Neural Net: Applying ML for gain models, feedback suppression, and other functions of the Hearing Device.
Proposed session chairs: Lars Bramsløw (Eriksholm), N.N.
Short description: The use of machine learning (ML) methods in hearing devices challenges classical approaches to hearing instrument development. ML methods promise advancements for various applications, including gain prescription, feedback suppression, and noise reduction. However, some important limitations remain, including the need for adequate training data, computational requirements, and power consumption. The quickly evolving landscape invites a look at the status quo and a discussion of new developments and perspectives.
Agenda (will be coming soon)