Efficient individualization of hearing aid processed sound

Jens Brehm Nielsen, Jakob Nielsen

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review


Due to the large amount of options offered by the vast number of adjustable parameters in modern digital hearing aids, it is becoming increasingly daunting—even for a fine-tuning professional—to perform parameter fine tuning to satisfactorily meet the preference of the hearing aid user. In addition, the communication between the fine-tuning professional and the hearing aid user might muddle the task. In the present paper, an interactive system is proposed to ease and speed up fine tuning of hearing aids to suit the preference of the individual user. The system simultaneously makes the user conscious of his own preferences while the system itself learns the user’s preference. Since the learning is based on probabilistic modeling concepts, the system handles inconsistent user feedback efficiently. Experiments with hearing impaired subjects show that the system quickly discovers individual preferred hearing-aid settings which are consistent across consecutive fine-tuning sessions for each user.
Original languageEnglish
Title of host publication Proceedings. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Publication date2013
ISBN (Print)978-1-4799-0356-6
Publication statusPublished - 2013
Event2013 IEEE International Conference on Acoustics, Speech and Signal Processing - Vancouver Convention and Exhibition Centre, Vancouver, Canada
Duration: 26 May 201331 May 2013
Conference number: 38


Conference2013 IEEE International Conference on Acoustics, Speech and Signal Processing
LocationVancouver Convention and Exhibition Centre
Internet address


  • Hearing aid personalization
  • Bayesian learning
  • Gaussian processes
  • Active learning
  • Preference learning


Dive into the research topics of 'Efficient individualization of hearing aid processed sound'. Together they form a unique fingerprint.

Cite this