时频面上基于瞬频估计的信号提取方法

INSTANTANEOUS FREQUENCY ESTIMATION BASED SIGNAL EXTRACTION METHOD

针对多分量非平稳信号的噪声抑制和信号分离问题,提出了一种新的时频滤波法。采用FMmlet自适应分解将被分析信号在时频空间内展开,得到无交叉项干扰、时频聚集性很强,且能反映信号的线性或非线性结构的时频特性的时频分布。根据时频面内能量脊与瞬时频率的对应关系,采用谱峰检测和时频加窗轮流迭代的方法,估计出信号分量的瞬时频率。在瞬时频率精确估计的基础上,设计适当的时频滤波函数,对被分析信号的Wigner-Ville分布(WVD)进行时频加窗处理,得到单分量信号的修正WVD,然后采用WVD反变换进行信号分量的时域重构,达到从非平稳信号中分离有用分量的目的。理论分析和仿真实验表明,利用该方法从复杂非线性时变信号中提取出的信号分量失真度小。该方法在非平稳信号的深层特征提取中具有良好的应用前景。

A novel technique is presented to isolate non-stationary vibration signal component using time-frequency (TF) filter based on instantaneous frequency (IF) estimation. FMmlet transform is introduced to display the vibration signal on a joint TF plane due to that the atoms of FMmlet transform can match both the linear and nonlinear time-varying structures and the TF distribution achieved would be free of cross-terms between components with high TF resolution. Then, according to the corresponding relation between energy ridges in TF plane and the IFs, the IF of the useful signal component is estimated with the alternate iteration of peak-detection and utilization of time-frequency window function. Based on the IF estimated, a TF filtering function is designed for modifying the Wigner-Ville distribution of the analyzed signal to get the TF signal of the useful component. Whereafter, the 2-D TF signal filtered out is reconstructed to form a 1-D vibration signal by the inverse transform of Wigner-Ville. Finally, a synthetic signal is analytically formulated and used as a test signal for evaluating the technique performance. The results demonstrate the effectiveness of the technique for signal component isolation and in-depth feature extraction for non-stationary multi-components signals.