Fourier based high-resolution near-field sound imaging

Noise pollution is a generally acknowledged problem in modern day society. The current tendencies towards lightweight and cheaper product design are primarily responsible for increasing nuisance, annoyance and environmental problems caused by acoustic noise. There are several reasons for research towards technologies that facilitate acoustic noise reduction. Nowadays, low noise design of consumer electronics, high-tech systems and automotive are restricted to increasingly stringent regulations and quality aspects. Effective countermeasures in order to reduce sound radiation are only taken when the source of sound is known. "Inverse Acoustics" is a very effective method to visualize and quantize the sound sources, which reconstructs source information based on measurements away from the source, yet in the near-field. In fact, the system is able to reconstruct the entire acoustic message that a source radiates in the direction of interest. The current methods for source reconstruction produce sound images with very little detail, they often require cumbersome numerical calculations and models, and they are often highly impractical for industrial applications. This research focuses on fast and accurate measurement and signal processing methods for inverse acoustics that are applicable in practical situations which require high resolutions under hazardous acoustic conditions. The inverse process is based upon spatial and wavenumber domain Fourier techniques, also referred to as Near-field Acoustic Holography. More in detail, spatial properties with respect to aliasing, leakage, signal-to-noise ratio and sensor set-ups are investigated and explicit methods and rules are developed to assist in proper determination of the acoustic holograms. In order to correctly transform the spatial hologram data into the wavenumber domain or k-space, a method called border-padding is developed. This method, which is an alternative to spatial windowing, is highly accurate without slowing down the processing time considerably. Another important factor is regularization, which is required since the inverse process is highly ill-posed. Without proper filtering action taken, noise blows up as the hologram-source distance or the wavenumber grows. In this research project a method is developed to automatically determine the proper filter function and filter parameters, which is a near-optimal trade-off between noise blow-up and deterioration of useful source information. These important properties are combined in a fully automated near-field sound imaging system design. At the Technical University of Eindhoven two versions of this system were developed and built; a large version that is based in the semi-anechoic room of the laboratory and a portable system that is suitable for small electronic devices and high-tech systems. A number of practical cases are used to qualitatively as well as quantitatively validate the improvements with respect to existing methods and illustrate the possibilities for industrial application.