Research On Sparse Signal Detection Method Based On Compressive Sensing

Many applications in radar system are based on achieving signal detection task,such as radar reconnaissance, radar imaging and so on. In modem radar systems,Chirp signal is commonly used. In order to achieve some functions, it is necessary torealize the signal detection of broadband chirp signal. However, if taking on thetraditional Nyquist sampling theorem, the high bandwidth of chirp signal means thatthere will be a problem on collecting, storing, transmissing and processing largeamount of data. Compressive sensing theory appears to provide a new way to solvethis problem.This thesis focuses on the research of bandwidth chirp signal detectionproblem.Combined with compressive sensing theory, we will analysis and implementa detection algorithm of chirp signal. The main contents include the sparserepresentation of chirp signal, the signal detection model based on compressivesensing theory,the design and implementation of detection algorithm based oncompressive sensing theory. The sparse representation of chirp signal focuses onresearching the adaptive decomposition in the wave-form delay dictionary andtime-frequency dictionary(Gabor dictionary and Chirplet dictionary).Our researchincludes two kinds of detection model, one is under channel condition with whiteGaussian noise, the other is under the channel condition with Gaussian noise andstrong narrowband interference.There are some existing detection algorithms based on compressive sensingtheory, mainly including incoherent detection and estimation algorithm and detectionalgorithm based on the digital signature of measurement values. The former algorithmuses the maximum projection coefficient obtained by partial reconstruction algorithmto complete detection, which is simple to achieve but shows poor detection result; thelatter uses the deviation of actual sampling values from the expectations undercorresponding hypothesis as criterion to accomplish detection, which presentsexcellent capacity under low signal to noise radio. Based on the analysis of existingalgorithms, we propose a correlation detection algorithm based on measurementvalues and verify the algorithm’s detection performance by simulations. Thesimulation results show that the detection success rate of the proposed algorithm ismuch bigger than that of the incoherent detection and estimation algorithm.Meanwhile, with respect to the incoherent detection and estimation algorithm andetection algorithm based on the digital signature of measurement values, the proposed algorithm has lower computation complexity. In addition, under channelcondition with narrowband interference, the proposed algorithm presents gooddetection performance when the signal-to-interference ratio is certain.