A Mixed-Mode Decomposition Denoising Algorithm Based on Variance Estimation

Conventional denoising algorithms have poor effect when dealing with nonlinear or non-stationary signals, and it is also difficult to select appropriate threshold parameters for denoising. This paper proposes a mixed-mode decomposition denoising algorithm based on variance estimation, which combines empirical mode decomposition (EMD) and variational mode decomposition (VMD). The algorithm performs a single EMD decomposition on an original signal, calculates an energy sequence of the first-order intrinsic mode functions, and estimates a noise variance in the original signal according to the noise energy attenuation characteristics, and the estimated variance value serves as the threshold for VMD denoising completion. Nonlinear signals and actual speech signals are selected to carry out experimental comparisons of various denoising algorithms. The results show that the mixed-mode decomposition denoising algorithm is not constrained by the selection of threshold parameters, and its denoising performance is better than that of conventional mode decomposition threshold denoising algorithms.