Study on Methods for Improving LMD End Effect by Gaussian Process Based on the Particle Swarm Optimization Algorithm

摘要

The LMD is a new method for analyzing non-stationary signals. It can decompose complicated signals into a set of single-component signals, each of which has physical sense. However, performing the LMD will produce end effects which make results distorted. After analyzing the reasons for these, the article takes advantage of the Gaussian process algorithm to overcome the end effects of LMD. To improve the precision of GP algorithm of endpoint extension, the authors use the particle swarm algorithm to optimization the GP hyper parameter and select the optimal covariance function. Experimental results showed that the GP algorithm of particle swarm optimization (PSO) can predict the two ends of the data signal more accurately, improve the accuracy of LMD and avoid the adverse effects caused by end effect according to the internal characteristics of the signal. Therefore the PSO-GP algorithm is a better method to improve the end effect.