经验模态分解在直升机谱特征提取中的应用

Application of empirical mode decomposition in helicopternoise spectrum feature extraction

为了更有效地提取直升机辐射噪声的频谱特征,提出了一种基于经验模态分解(EMD)的特征提取算法.仿真研究表明EMD具有突出信号局部特征且余量可表征信号变化趋势的特性,该算法将目标辐射噪声的频谱构成和EMD的特性相结合,将EMD应用于变换域(信号频谱序列)信号分析,有效的实现了信号连续谱的提取.相比于平滑滤波器法须人为选择最优的α值,且提取的连续谱有滞后性的不足,文中所示方法可自适应的提取信号的连续谱,且在准确无滞后地估计连续谱的同时,对线谱有一定的"增益",为后续的目标识别提供更全面的信息.

In order to better extract identifying features in the sound spectra of helicopter-radiated noise,a new feature extraction algorithm based on empirical mode decomposition（EMD） was proposed.Simulation results showed that the EMD algorithm can emphasize a signal＇s instantaneous characteristics and that the residual trend information was the mutative trend of the original signal.The proposed algorithm combined the characteristics of the EMD algorithm and spectrum composition of helicopter-radiated noise.Application of EMD to the signal analysis in transform domain（frequency domain） allowed effective extraction of the continuous spectrum.Using the smoothing filter method,when choosing the optimal α value the extracted continuous spectra are hysteretic to some extent.To overcome this disadvantage,the proposed algorithm can adaptively extract a signal＇s continuous spectrum,and the extracted continuous spectrum has no lag.In addition it can provide gain for a line spectrum,which can supply more useful information for subsequent target identification.