The Local Mean Decomposition And Its Application

The analysis of the feature and the extraction of usefull information from a greatamount of irregular and complex signals is an important issure in modern scientificresearch and a hotspot among scholars both in China and abroad. The effect of thetraditional feature-extraction method such as Fourier transform, Power spectral analysisand Wavelet transform is unsatisfactory in analysing the nonlinear and nonstationarysignals. Empirical mode decomposition(EMD) is a novel timefrequency analysismethod, which results in a good effect in the application of the EEG research, themechanical fault diagnosis, etc. A lot of problems such as end effect and negativefrequency still existed because of the limitations of the algorithm itself.Local mean decomposition(LMD) is proposed on the basis of EMD. Theadvantages of EMD are kept and the defect is improved. LMD is especially suitable foranalysis of the amplitude-modulated or frequency-modulated signals. The principle ofLMD is analysed and several unsolved issures is particularly discussed. Severalreference conclusion are proposed about the value of the parameter in the judgement ofthe pure magnitude modulated signal; The end effect in the calculation of the localmean and the envelop is improved by a simple and effective method; A comparision ofthe cubic spline interpolation and the moving average is carried out. LMD, combinedwith energy feature and Support vector machine(SVM) theory, is applid in the analysisof EEG signals and earthquake precursor wave. EEG signals and earthquake precursorwave are preprocessed and decomposed by LMD, the first three PFs of the EEG signalsand the first five PFs of the earthquake precursor wave are determined as the feature andare input into paramete optimized SVM, the output of SVM are the classification resultsof different signals.The results show that LMD can improve the classification accuracy of motorimaginary BCI experiment to92.25%and the earthquake prediction to62.24%, which ismuch better than the results of the traditional methods.