Back to list

Evaluation of various Algorithms to detect acoustic Feedback

Authors Braun, S.
Year 2012
Thesis Type Audio Engineering project
Topic Audio Signal Processing
Keywords audio reproduction
Abstract Annoying howling caused by acoustic feedback is an omnipresent problem in amplified live-sound situations. There are some algorithms that detect feedback howling frequencies automatically and try to eliminate these via notch filters. For the detection of the feedback frequencies, several different approaches and criteria are available, that can be used separately or in combination. The proposed algorithms shall be implemented and evaluated in a suitable simulation environment. In this work, a new evaluation method is designed, that is more objective than the proposed evaluation in literature. Two evaluation methods are introduced, but the final evaluation focuses on the new developed method. Further, the improvement of the detection algorithms by using a Constant-Q- Analysis, and the additional use of a second microphone are investigated. This second microphone can gather additional information, such as the estimation of maximum stable gain (MSG) of the amplification system and might improve the performance. The evaluation results of several detection criteria form a basis for an advanced design of a feedback detector.
Supervisors Pomberger, H.