High-resolution source imaging with bio-inspired sensing systems

lateral line;beamforming;Capon;Microelectromechanical systems (MEMS);reassignment;l1-minimization;sparse beamforming;weakly electric fish;electrosense;source localization.

Nguyen, Nam

关键词: lateral line;    beamforming;    Capon;    Microelectromechanical systems (MEMS);    reassignment;    l1-minimization;    sparse beamforming;    weakly electric fish;    electrosense;    source localization.;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/24161/Nguyen_Nam.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Source localization is ubiquitous in nature. It is a survival skill in many species to help them find food, avoid predators, or navigate. For example, blind cave fish use their lateral lines to swim in dark water by sensing flows, weakly electric fish generate an electric field to detect electric distortionscaused by nearby objects, and bats emit ultrasound and listen to echoes to capture insects. It is always the desire and challenge for engineers to buildman-made systems that can deliver such capabilities.In this thesis, two new bio-inspired, man-made sensing systems are developed.Using new hair-cell sensors built from the Micro-Electro-Mechanical-Systems (MEMS) technology, we develop an artificial lateral line system similar to the one of fish. An adaptive beamforming algorithm is used to provide high-resolution images of source locations. The other system is built based onthe principle of weakly electric fish. As it is an active sensing system, signals from multiple sources are coherent, and the previous adaptive beamformingfails. We then introduce the concept of sparse beamforming by exploiting the fact that objects to be localized are sparse in space. It is shown that thesparse beamforming technique is capable of resolving coherent sources.We not only devise those man-made sensing systems, but we also develop new algorithms to process the input sensor signals and enhance the output images. First, we provide a new l1-minimization algorithm using a backward basiselimination technique. The algorithm outperforms the well-known l1magic package for small-scale problems. This algorithm can be used in the sparsebeamforming application. Second, we introduce the reassignment method into the source localization problem to sharpen output images. The algorithmis verified in both the artificial lateral line with a vibrating sphere and the weakly electric sensing system with an insulating plastic ball.Overall, we have demonstrated the practical possibility of constructing novel man-made sensing systems that can imitate several source localization capabilities previously found only in nature.