最大似然频谱估计法与周期图最大值法的性能比较分析

Performance Analysis of Maximum Likelihood Spectral Estimator Compared with PM Estimator

研究宽带信噪比(RSNW)和有效功率谱宽度对最大似然(ML)法多普勒频率估计结果的影响,采用探测概率和正确估计频率标准差来描述ML算法对仿真信号和实验数据的处理性能,并与周期图最大值(PM)法所得结果进行比较。经理论仿真和对中国科学院上海光学精密机械研究所1.5μm全光纤相干测风激光雷达实测数据的反演处理,结果表明,高谱宽度下,ML算法的正确估计频率标准差与相比PM可降低0.5 MHz,达到90%的探测概率所需宽带信噪比ML算法与PM算法相比低2 d B。表明ML算法的正确估计频率标准差能与PM算法相比不高于1.1 MHz,探测概率高9%。要获得风速精度小于1 m/s同时探测概率在80%以上,所需宽带信噪比大于-14 d B。

The effect of the wideband signal-to-noise ratio （RSNW） and the effective power spectral width on the Doppler frequency estimation performance using maximum likelihood （ML） estimator have been studied. The performance of the ML algorithm is applied to simulation signal. Experimental data are described by the detection probability and the standard deviation （SD）, and also compare with periodogram maximum （PM） estimator. The data are acquired by 1.5 μm all fiber coherent Doppler lidar in Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences. The results indicate that ML′s SD can enhance 0.5 MHz higher than PM′s under the same spectral width. ML is close to PM under the narrow spectral width. To satisfy the 90% detection probability, the required RSNW is 2 dB smaller than PM′s. In the real-process, ML′s SD can be 1.1 MHz lower than the PM, and the detection probability is 9% lower than PM. To obtain the wind speed accuracy less than 1 m/s while the detection probability is above 80%, the required RSNW is larger than -14 dB.