Algorithm Evaluation

Test conditions

All signal enhancement algorithm categories were evaluated in a wide range of simulated listening scenarios, using objective (physical) performance measures intended to mimic normal hearing. The following environments have been considered :

  • a low-reverberant room
  • a typical living room
  • a large cafeteria room
  • the inside of a car
  • a street corner

Depending of the type of room, a number of real-life recordings were made with a manikin head wearing two multi-microphone behind-the-ear hearing aid devices :

  • male and female speech
  • a music signal
  • speech-shaped noise, i.e. a stationary noise signal having the average spectrum of speech
  • cafeteria babble noise
  • noise inside a car cabin
  • noise recorded beside a street

The most promising algorithm variants were selected within each category as candidates for further study. All algorithms have been implemented on the same PC-based real-time Master Hearing Aid platform. The processing is done at 16 kHz sampling frequency.
Expected performance has also been evaluated for users with different auditory profiles, i.e. different types of hearing loss. As it is basically impossible to let many users in each auditory profile evaluate every algorithm, a preliminary evaluation has been performed by simulating each auditory profile using an established type of auditory model that accounts for some of the most well-known effects of hearing impairment.

Subjective listening tests

In the near future each implemented algorithm will be evaluated through subjective listening tests by a number of selected normal hearing and hearing-impaired user groups.