Back to list

Wechselwirkung von Lautsprecher-Mikrofon Anordnungen in Fahrzeugen

Authors Noisternig, M.
Year 2017
Thesis Type Doctoral thesis
Topic Audio Signal Processing
Abstract This work studies broadband signal enhancement methods with the aim to suppress noise and interference signals and, at the same time, minimise target signal distortions. In general, they can be subdivided into single-channel and multichannel solutions. It can be shown, that combining a microphone array beamformer with a single-channel noise suppression post-filter can significantly improve the signalto-noise ratio at the output of the system. To achieve almost real-time capability the overall latency must not exceed a certain threshold. Spectral subtraction is one of the most popular single-channel approaches to noise suppression and signal enhancement. The proposed approach improves the spectral weighting rules to eliminate musical noise. The fast response time leads to a significant reduction of the distortion of transient sounds. For the time-frequency transform a computationally efficient analysis-synthesis filter bank is developed based on masking properties of the human auditory system. Hence less emphasis is given to frequencies where noise is inaudible. This reduces the target signal distortions and improves the auditory impression. It is further shown that beamforming with microphone arrays can be used to significantly reduce the effects of noise and interference. A robust adaptive beamformer for small arrays with a low number of microphones is developed, which is based on the Generalised Sidelobe Canceler principle. It provides a high array gain over a broad frequency range and is robust against microphone mismatch, position errors, and multipath propagation. Modal microphone array beamformers provide signal enhancement within both a large temporal and spatial frequency band. The spatial sampling grids have a strong influence on the error tolerance of the systems. It is shown that only a few microphones are needed in the interior of the considered microphone array to ensure a low interpolation error in the frequency band of interest, and that most of the microphones have to be located on the boundary of the domain, with a non-uniform density depending on the shape of the domain. It is demonstrated that practical constraints can be included in the optimization process in order to ease the implementation of an array. Comparisons for some particular array geometries with design methods known from the literature are given, showing that the proposed approach results in significantly lower errors over an extended frequency range.
Supervisors Höldrich, R.