Sound source localization from laser vibrometry recordings

Sammanfattning: Acoustic holography deals with a typical inverse problem, to obtain the unknown from something that is known. From measurements of sound, most often using microphone arrays, reconstructions of the sound field are made in an attempt to find out what causes the sound. The successfulness of such analyses is highly dependent on the quality of the reconstructed images, where factors like noise and spatial resolution are of importance. Also, when using microphones for measurements of the sound it is inevitable that the sound field is disturbed. Such disturbances will always have an impact on the reconstructions, even though in most cases the effects are negligible. In this thesis laser vibrometry for the purpose of sound measurements has been investigated and evaluated. Since it is an optical method it has the great advantage of allowing the sound to be measured without being disturbed. The 1D quantitative data it provides can be used without any preparation for the 2D numerical propagation of sound fields. Further on, the measurement data can also be used in order to obtain a 2D or 3D tomographic reconstruction of the sound field. A multi-wavelength method is also presented, which efficiently suppresses noise and increases the spatial depth resolution in sound field reconstructions. The sound is measured and reconstructed for several frequencies. When adding these reconstructions they will only add up constructively in positions corresponding to a sound source. This way noise in the resulting image is efficiently suppressed and the information of interest enhanced. Besides this "cleaning" of the image a large improvement of the depth-resolution is also achieved. While the in-plane size of the imaged sound source is diffraction-limited, the spatial depth-resolution can be improved by applying a filter based on the standard deviation calculated over the phase for the different reconstructions. Using this filter, sound sources can be imaged with a depth-size as small as only two wavelengths. Another advantage with this multi-wavelength technique is that it is possible to make a selective imaging of sound sources. When the measured sound originates from several different sound sources, a reconstruction can be made where only sound sources of a specific type are imaged. The other sound sources are treated as noise and are efficiently suppressed. This multi-wavelength method for imaging and localization of sound sources has so far proven to be very efficient and to have a low sensitivity to noise in the measurements.