Signal Processing

Joint Multichannel Dereverberation and Noise Reduction

The Signal Processing group is offering a Master thesis in the field of Joint multichannel dereverberation and noise reduction.

In many speech communication applications, the recorded microphone signals are of poor quality, because of background noise and reverberation (e.g., reflections
against walls).  With the continuously growing demand for high-quality hands-free speech communication in teleconferencing applications, voice-controlled systems or hearing aids, speech enhancement techniques aimed
at both noise reduction and dereverberation have become indispensable. 

While many algorithms that aim only at noise reduction or only at dereverberation have been investigated, their optimal integration to simultaneously achieve both noise reduction and dereverberation still remains an open problem.

The aim of this thesis is the analysis, development, and implementation of multichannel speech enhancement algorithms for joint noise reduction and dereverberation.

Tasks:

  • Literature review on existing multichannel noise reduction/dereverberation algorithms
  • MATLAB implementation of multichannel speech enhancement algorithms
  • Analysis of the possibly arising trade-off between dereverberation and noise reduction performance

Recommended knowledge:

  • Digital Signal Processing (e.g. by attending the courses „Advanced Topics Speech and Audio Processing“ and the Block-Praktikum Digital Signal Processing)
  • MATLAB programming