Measurement Matrix Optimization For MIMO Compressive Sensing Radar Based On Information Theory

Recently, wide bandwidth, high resolution and multi-channel multi-target have gained more and more popularity in radar signal processing, which has brought some problems, including high computational complexity, large storage capacity, large transmission data volume and difficulty in engineering application. Limited by Nyquist sampling theorem, big data processing has become the bottleneck for traditional radar. Due to the direct sampling, Compressive Sensing Radar can reduce computational complexity and improve the resolution, so it has a bright prospect in radar research.The incoherence between measurement matrix and redundancy matrix is the precondition of effective signal recovery. Measurement matrix can not only affect the information sampling and transmission, but also influence whether it can or not recover the original signal. Traditional random measurement matrix need large storage capacity and computational complexity, it is not the optimum measurement matrix. Therefore, in order to improve the performance of radar target information extraction and imaging ability, this thesis considers the problem of measurement matrix optimization in MIMO Compressive Sensing Radar, analysis and shows the effectiveness of measurement matrix optimization method based on maximizing SNR. What’s more, a measurement matrix optimization method based on mutual information is proposed. First, a Bernoulli-Gaussian Signal Model is established in MIMO Compressive Sensing Radar imaging scene. Then, measurement matrix optimization object function and constraint based on maximizing the mutual information and minimizing the coherence of the sensing matrix according to information theory and convex optimization is derived, software packages are used to solve the problem. Finally, in the case of the optimizing measurement matrix, SLIM algorithm is used to recover the target information extraction and imaging.Computer simulation shows that the method of measurement matrix optimization based on mutual information can achieve smaller mutual coherence of each column in sensing matrix and get better measurement matrix. This method not only can overcome the bad performance of the existing random measurement matrix and some determinate measurement matrix, but also can largely degrade the error between recovery target and original target, thus enhance the performance of target information extraction and imaging.