Digital Signal Processing in Real-time for Magneto-electrical Sensor Systems

点击翻译按钮获取中文摘要

This research proposal focuses on real-time processing for new magneto-electrical sensor systems, which are able to operate at room temperature. With those sensors (and the corresponding signal processing techniques) medically relevant research questions in the area of neurology should be addressed. The objectives are to enhance already existing signal detection, estimation, and enhancement algorithms but also to design new adaptive processing schemes. The focus of these algorithms should be on real-time implementations. The entire proposal is structured in three parts:1. A multi-channel real-time system should be built up, which is capable of connecting a multitude of magneto-electrical, electrical, and acoustic as well as acceleration sensors. All of these inputs should be equalized in an optimal manner. Background, amplifier, and other noise types should be suppressed. Exogenous noise sources should be cancelled using a combined approach consisting of a beamforming and a cancellation stage.2. The enhanced signals obtained from part 1 will be processed by a second stage in order to remove endogenous artifacts as they appear due to heart beats, eye blinks, or muscle activity. This processing stage is designed based on state-space approaches.3. Based on the DICS algorithm (DICS is a shortcut for Dynamic Imaging of Coherent Sources) an innovative source localization in real time will be investigated. For this purpose the further enhanced signals of part 2 will be used.The results of all three parts of this proposal interact. If, e.g., the coherence analysis and source localization detect the origin of a neurologic network in a certain area, the magneto-electrical sensors can be adjusted and leveled towards this origin after a short measurement. As a result the spatial resolution of the analysis will improve. This also applies to the optimal use of the bandwidth of the magneto-electrical sensors. Currently the bandwidth of the sensors is between 10 and 30 Hz and therefore not comparable to conventional (SQUID based) sensors. With the help of a frequency shift the maximum sensor sensitivity can be aligned adaptively. If a frequency range of interest is determined based on a preliminary analysis, the demodulation of the sensors can be adapted. Consequently the frequency range of interest will be investigated with the optimal signal-to-noise ratio. This leads to the development of a closed control loop, which justifies the real-time requirement of this research proposal.