intro algorithm sounds

Coherence-based dereverberation - sound examples

The following sound examples illustrate the speech enhancement capabilities of the binaural coherence filtering based approach. Three sound files are presented :

  • the unprocessed sound file, which contains the signals captured by the front microphone of the left and right hearing aid. These signals are the input to the binaural coherence filtering algorithm.
  • a processed sound file that is the output of the binaural coherence filtering algorithm using a high exponent to achieve efficient filtering.
  • a processed sound file that is the output of the binaural coherence filtering algorithm using a medium setting leading to less aggressive filtering.

Cafeteria with babble noise in the background


It can be heard that the diffuse babble noise is reduced, mainly at high frequencies. At low frequencies, babble artifacts remain in the signal, sounding a bit like water in a small creek. The speaker signal distortion can be perceived as increased roughness added to the signal.

In the example 'processed1' with efficient filter settings, it can be noticed that not only the target speaker signal remains in the output, but also some competing speakers which are close to the listener. This effect distinguishes the coherence filter from directive microphones, where the target signal is passed through and competing (point) sources coming from other directions are suppressed.