自适应Alpha-beta修正的线谱检测后置处理方法

A Post-processing Method for Line Spectrum Detection Corrected by Adaptive Alpha-beta Filter

矩形包络的单频信号时域频率方差等于频域有效带宽,且频率方差为信噪比的函数,在信噪比下降时,线谱方位权值计算误差变大,频率方差线谱检测器丢失目标。传统Alpha-beta滤波的增益恒定,不能有效修正频率估计。该文提出后置处理方法,将每个方位下多次求得的功率谱叠加后得到功率谱-频率曲线作为该方位下各频率对应的增益。当频率观测值对应的增益大时,新息在递推过程中占的比重大,当频率观测值对应的增益小时,预测值在递推过程中起的作用大,对滤波后频率估计值计算方差来修正方位谱输出。仿真和实验处理结果证明在低信噪比情况下,修正后的线谱检测后置处理算法抗干扰性强,能够实现弱线谱的提取,同时压低了干扰背景。

The frequency variance of single-frequency signal with rectangular envelope in time domain equals to the effective bandwidth in frequency domain, and the frequency variance is a function of Signal to Noise Ratio （SNR）. When the SNR drops, the calculated frequency variance error of azimuth of target line spectrum enhances. As a result, the Variance of frequency detector which weights bearing spectrum using the variance of peak frequency of each azimuth losts its target. The gain of traditional Alpha-beta filter is constant, thus it fails to correct frequency estimates effectively. The paper puts forward a post-processing method which sums the power spectrum of each azimuth several times and puts it as a gain corresponding to each frequency. When the gain corresponding to the frequency observation is large, the new information in the recursive process counts more, whereas the gain corresponding to the frequency observation is small, then the predicted value counts more. At last the bearing spectrums are rectified by using the modified variance of estimated frequency of each azimuth as weight. Simulation and experiment results show that under circumstance of low SNR, the modified post-processing method can extract weak line spectrum target and suppress interference background.