Research And Application On Low SNR Signal Detection Based On Chaotic Oscillator And Wavelet

Low signal-to-noise ratio (SNR) signals exist widely in secure communication, signal reconnaissance, ground penetrating signal processing, deep sea sonar exploration, electricity & machine system appearance monitor, medical signal processing etc. Thus, the detection of low SNR signals becomes the first issues to be addressed in these tasks.Duffing oscillator is sensitive to weak signals yet immune to noise when system changes from chaotic to outer periodic orbits. Using this character the weak periodic signal in strong noise can then be detected and the improved SNR signal will be output. In this dissertation, the detections of periodic signal and linear frequency modulated (LFM) signal at low SNR by Duffing oscillator are investigated.The wavelet coefficients of signal and noise after wavelet have the different natures. Using the linear or nonlinear denoising methods, the most signal coefficients will be preserved while the noise coefficients are reduced. Therefore the output signal with the improved SNR is achieved. In this dissertation, some wavelet denoising methods are improved in low SNR signal detection.The main research works and contributions of this dissertation are as follows:1. The output Statistics natures in detection of periodic signal by Duffing oscillator under the Gauss white noise are researched. And then we presented an evaluation index for detection ability of chaotic oscillator system using periodic mean of output SNR improvement ratio (pmoSNRir). The validity of pmoSNRir is demonstrated through theory analysis and simulation experiments. We also give pmoSNRirs in different SNRs.2. The detection ability of Duffing oscillator system on low SNR LFM signal is investigated, and then influence of modulating parameters to detectable minimum input SNR of LFM signal is acquired.3. The soft threshold denoising method based on minimum mean square error (MMSE) is investigated. The optimal shrinkage of wavelet coefficients under the Gauss white noise is deduced and verified.4. The adaptive wavelet threshold denoising method based on SURE estimation is investigated and a novel set of threshold functions, which is a uniform expression of differentiable threshold functions, is presented. The function family with the adjustable parameters is close to Donoho soft threshold function and has continuous derivative. The advantages of the new threshold function family make it possible to construct an adaptive algorithm to look for the best threshold.5. The selection of optimal wavelet basis function is investigated. The cross correlation coefficients between periodic signals and db wavelets are calculated and the optimal wavelet to analysis LFM and sine signal is found.6. A low SNR signal detection system is founded by synthesizing Duffing oscillator, wavelet denoising and correlation technique, which is applied to signal receiving and processing in radar bistatic-based system. The ability of detection low SNR signal of corresponding system is improved.