基于盲源分离的混合信号时频域特性提取仿真

Time-Frequency Domain Feature Extraction of Mixed Signals Based on Blind Source Separation

针对传统混合信号时频域特性提取中,普遍存在着提取完成时间较长、成本消耗较大、准确度较低等问题,提出一种基于盲源分离的混合信号时频域特性提取方法。对混合信号时频域特性进行分析,利用经验模式分解方法对混合信号进行分解,得到模式分量混合信号数量估计,并获取最优观察信号,将源信号与混合信号的最优观察信号构成新的观察信号,提取混合信号时频域特性。实验结果表明,所提出方法提取准确度较高、完成时间较短、成本消耗较大。

Due to long extraction time,high cost,and low accuracy in traditional feature extraction of mixed-signal time-frequency domain,this article puts forward a method for extracting time-frequency domain feature of mixed signal based on blind source separation.Firstly,the frequency domain feature of mixed signal was analyzed.Secondly,the empirical mode decomposition method was used to decompose mixed signal and obtain the number estimation of mixed signal of mode component.Thirdly,the optimal observation signal was obtained.Moreover,the optimal observation signal between source signal and mixed signal formed a new observation signal.Finally,the frequency domain feature of mixed signal was extracted.Simulation results show that the proposed method has higher accuracy of extraction,shorter completion time and higher cost.