高斯混合噪声下微弱线谱检测的EM-静态非线性系统方法

Em Nonlinear System Method of Weak Line Spectrum Detection in Gaussian Mixture Distribution Noise

采用3-水平量化器随机共振对非高斯水下噪声中的线谱信号进行检测,为使该随机共振类型达到最优的检测效果,利用EM算法进行了高斯混合分布的参数估计.通过仿真和实测数据对参数估计方法进行了检验,实验中算法能够正确收敛,拟合性能较好,在此基础上可以得到量化器的最优阈值.随机共振系统处理前后的功率谱对比表明,3-水平量化器随机共振是检测高斯混合分布中微弱线谱的有效手段.

The 3 level quantizer stochastic resonance is utilized to detect line spectrum buried in non Gaussian underwater noise. To achieve optimal detection result by the stochastic resonance, EM algorithm is proposed to estimate parameters of the Gaussian mixture distribution. Simulative and real data are all used to test the estimation method. Experiments show that the algorithm has proper convergence and fitting property. Based on the parameters got by EM method, the optimal threshold of quantizer could be calculated. The comparison between the power spectrum of the signals preceded and passed stochastic resonance implies that, the 3 level quantizer stochastic resonance is good for the detection of weak line spectrum buried in Gaussian mixture noise.