Multichannel Signal Denoising Using Multivariate Variational Mode Decomposition with Subspace Projection

Abstract

This paper describes a novel multichannel signal denoising approach based on multivariate variational mode decomposition (MVMD). MVMD is the extended version of the variational mode decomposition (VMD) algorithm for multichannel data sets. Unlike previous MEMD (multivariate empirical mode decomposition)-based denoising methods, the proposed scheme not only has a precise mathematical framework but also can better align the common frequency modes of the signals. Therefore, it has good robustness for non-stationary signals with low SNR. Based on the similarity measurement between the probability density function (pdf) of the input signal and each mode by Hausdorff distance(HD), the interval thresholding and partial reconstruction denoising of band-limited intrinsic mode functions (BLIMFs) are performed in the algorithm. Besides, to take advantage of the characteristics of channel diversity, the subspace projection method is used to further denoise the multivariable signals. We demonstrate the effectiveness of the proposed approach through results obtained from extensive simulations involving test (synthetic) and real-world multivariate data set.