基于EMD分解的盲源分离模型以及在多路径削弱中的应用

Blind Source Separation Model Based on EMD Decomposition and Its Application in Multipath Weakening

通过阐述EMD和广义特征值的盲源分离的相关理论,并严密推导了相关公式,系统地分析了2种方法的优势,在此基础上,提出一种基于EMD分解的广义特征值盲源分离算法获取多路径误差模型,可有效削弱多路径效应误差。实验结果表明:新算法得到的标准差最小,较原始坐标序列的标准差分别减小约80%、90%、77%;与EMD方法相比,N方向标准差减小了33.3%,E方向标准差减小了59.2%,U方向标准差减小了30.8%;与Wavelet方法相比,N方向标准差减小了6.8%,E方向标准差减小了16.3%,U方向标准差减小了10.7%。新模型对BDS坐标序列中多路径误差进行更为有效的削弱,为GNSS变形监测提供了有力的技术支持。

This paper expounds the related theories of EMD and generalized eigenvalue blind source separation respectively, deduces the related formulas rigorously, and analyze the advantages of the two methods systematically. On this basis, a generalized eigenvalue blind source separation algorithm based on EMD decomposition is proposed to obtain the multipath error model, which can effectively reduce the multipath effect error. The experimental results show that the standard deviation of the new algorithm is minimized, which is reduced by about 80%, 90% and 77%, respectively compared to the standard deviation of the original coordinate sequence. Compared with the EMD method, the standard deviation in N direction, E direction and U direction is reduced by 33.3%, 59.2% and 30.8%, respectively. Compared with the wavelet method, the standard deviation in N direction, E direction and U direction is reduced by 6.8%, 16.3% and 10.7%, respectively. The new model weakens the multipath error in BDS coordinate sequence more effectively, and provides strong technical support for GNSS deformation monitoring.