基于贝叶斯频谱估计的频率分辨率研究

Study of Frequency Resolution Based on Bayesian Spectrum Estimation

研究了贝叶斯频谱估计的基本理论和基本方法,分别对频率扫描法和频差扫描法进行了研究和计算。仿真结果显示,采用加窗方法可有效提高频率扫描法的频率识别精度。在采样信号相同的条件下,频率扫描法的频率采样点数为1000,得到的频率分辨率为1%。差频扫描法的采样点数为100,得到的频率分辨率为0.1%。贝叶斯频谱分析中模型的选取是决定性的因素。

It discusses the rationale and principle of Bayesian spectrum analysis, studies and calculates the frequency scanning method and the frequency difference scanning method respectively. The result of the simulation reveals that the precision of frequency identification can be raised effectively when using frequency windowing method. Under the same signal sampling condition, the resolution of frequency is about 1% with the frequency sampling point equaling to 1000 by making use of the frequency scanning method, and is about 0. 1% of 100 sampling point by making use of the frequency difference scanning method. The most important factor would be the selection of the model in Bayesian spectrum estimation, that is the conclusion.