Chair of
Multimedia Communications and Signal Processing
Prof. Dr.-Ing. André Kaup

Binaural noise suppression techniques for assistive listening under extreme industrial noise conditions

Supervisor:Dr.-Ing. Klaus Reindl
Faculty:Prof. Dr.-Ing. Walter Kellermann
Student:Michael Bürger
Begin:2011-12-01
End:2012-05-31
File:These-PDF
Info:

Under extreme industrial noise conditions, where the sound pressure level can be higher than 95dB(A) as observed, e.g., in the mining industry, at construction work, etc., the use of hearing protection is compulsory to prevent workers from significant hearing damage. However, not only the extreme noise conditions but also the use of hearing protection makes speech communication impossible in such environments. To make speech communication possible, the protection systems are equipped with so-called listening assistants.

Listening assistants for hearing protection can be realized as follows: At each ear, the hearing protector is equipped with multiple microphones (in general two microphones). The two microphones at each side are aligned horizontally and parallel to the look direction. The available microphone signals are processed in such a way that a desired signal assumed to be located in front of the user can be extracted by suppressing all undesired noise and interference components and at the same time preserving the binaural information of all sources. As bilateral processing is limited in its noise and interference suppression capacity and may significantly distort binaural information, a binaural processing strategy is considered.

In this thesis, a comparative study of existing binaural processing strategies should be carried out. It can be assumed that the capacity of the binaural data link allows for full-band bidirectional exchange of a single audio signal. The most promising strategy that optimally exploits the available microphones for the problem at hand and does not distort binaural cues should be analyzed theoretically as well as by Matlab simulations. The performance for stationary, non-stationary, and impulse-like industrial noises should be analyzed. For evaluation, the speech distortion, the SINR gain, as well as the influence on the binaural cues at the outputs of the signal extraction scheme should be validated. Well-documented and well-structured software is important.

TypeMaster Thesis
Status:Terminated