一种基于声阵列信息融合及改进EEMD的信号降噪方法

A signal de-noising method for multi-microphone array based on information fusion and improved EEMD

针对声阵列多通道信号的去噪问题,提出一种基于多传声器信息融合辅助的改进总体平均经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)的被动声信号去噪方法。对标准EEMD进行改进,通过多通道信号频谱分析,选取多传声器信号最小有效频率作为各通道信号EEMD分解的筛选截止频率,采用改进的EEMD算法将原始信号快速分解为完备的IMF分量,有效抑制了模态混叠现象并提高信号分解效率;引入声阵列时延矢量封闭准则(Time Delay Vector Close Rule,TDVCR)概念,结合多传声器数据一致性融合及信号相关性理论,对各IMF分量进行相应的权重计算,再由已确定权值对各IMF分量进行加权重构得到去噪信号;最终通过半实物仿真试验以及同传统EMD去噪的比较验证了该算法在多通道信号去噪中的有效性和实用性。

Aiming at multi-microphone signal de-noising problem in acoustic array,a signal de-noising method based on information fusion and improved ensemble empirical mode decomposition( EEMD) was proposed here. Firstly,an improved EEMD algorithm taking the ensemble minimum effective frequency of multi-channel signals as the cut-off frequency of EEMD was introduced. Original signals were decomposed rapidly into complete IMF components with the improved EEMD algorithm to suppress effectively mode mixing phenomena and false IMF components. Secondly,through introducing the time delay vector close rule( TDVCR) and the data consistency fusion theory,the weight matrix of IMFs was computed according to the correlation of corresponding IMFs,and then the de-noising signals were reconstructed with the weighted IMF components. Finally,through the semi-physical simulation tests of acoustic array and comparing with the traditional EMD signal de-noising,the effectiveness and applicability of the proposed method were verified.