Data noise reduction method based on improved EMD and MED

Abstract

The invention discloses a data noise reduction method based on improved empirical mode decomposition (EMD) and minimum entropy deconvolution (MED). The invention belongs to the technical field of communication, and comprises the following specific steps: 1, denoising an original signal by using MED to enhance the impact characteristic of the signal; 2, inputting the denoised signal into an improved EMD for decomposition, and drawing an envelope line on the denoised signal by using an envope function to enhance the continuity of the signal; 3, extracting characteristic variables, namely IMF components, of the signal with the drawn envelope line by using improved EMD, so that the signal processing workload is reduced; and 4, finally, inputting the extracted IMF components into other subsequent data processing models, so that the workload of subsequent model processing is reduced, and the precision of subsequent data analysis processing is improved. According to the method provided by theinvention, the limitation of MED feature extraction under strong background noise is made up, and the defect that empirical mode decomposition is insensitive to weak fault features is overcome.

Description

technical field [0001] The invention belongs to the technical field of communication, and in particular relates to a data noise reduction method based on improved EMD and MED. Background technique [0002] When the vibration signal is extracted, it is always mixed with artificial or mechanical noise, which will affect the use of the signal and even cause errors in the detection results. Therefore, denoising the signal is of great significance to ensure accurate sampling. [0003] In recent years, scholars at home and abroad have carried out extensive research on signal noise reduction, and proposed a series of signal noise reduction processing methods including wavelet analysis, blind source separation, and empirical mode decomposition. Among them, wavelet analysis needs to select different wavelet basis functions according to different waveforms. Improper selection may lead to poor accuracy; the amplitude and sequence of source signals separated by blind source separation ...

AI Technical Summary

Problems solved by technology

Among them, wavelet analysis needs to select different wavelet basis functions according to different waveforms. Improper selection may lead to poor accuracy; the amplitude and sequence of source signals separated by blind source separation are uncertain; empirical mode decomposition overcomes the traditional envelope In the analysis, it is necessary to predetermine the center frequency band of the filter. It is widely used in the analysis of nonlinear and non-stationary signals, but there is a phenomenon of modal aliasing.

In summary, the above methods are insufficient for signal noise reduction when used, so it is necessary to find a new method to suppress the interference of noise on the sampled signal

Images