Reducing Microphone Artefacts in Live Sound

Abstract

This thesis presents research into reducing microphone artefacts in live sound with no prior knowledge of the sources or microphones. Microphone artefacts are defined as additional sounds or distortions that occur on a microphone signal that are often undesired. We focus on the proximity effect, comb filtering and microphone bleed. In each case we present a method that either automatically implements human sound engineering techniques or we present a novel method that makes use of audio signal processing techniques that goes beyond the skills of a sound engineer. By doing this we can show that a higher quality mix of a live performance can be achieved. Firstly we investigate the proximity effect which occurs on directional microphones. We present a method for detecting the proximity effect with no prior knowledge of the source to microphone distance. This then leads to a method for reducing the proximity effect which employs a dynamic filter informed by audio analysis. Comb filtering occurs when the output of microphones reproducing the same source are mixed together. We present a novel analysis of how the accuracy of a technique to automatically estimate the correct delay of the source between each microphone is affected by source bandwidth and the windowing function applied to the data. We then present a method for reducing microphone bleed in the multiple source, multiple microphone case, both in determined and overdetermined configurations. The proposed method is extended from prior research in noise cancellation, which has not previously been applied to musical sound sources. We then present a method for simulating microphone bleed in synthesised drum recordings, where bleed enhances the realism of the output. Through subjective listening tests and objective measures each proposed method is shown to succeed at reducing the microphone artefacts while preserving the original sound source.