To reduce the noise in the signal in nonstationary signals, a new technique based on EEMD(Ensemble Empirical Mode Decomposition) was studied for decomposing signal. With the EEMD the extremums of the signal was used as scalse to decompose data in the time domain and get the IMFs (Intrinsic Mode Function) with their frequencis from high to low. Then based the frequency spectrum from FT(Fourier Transform) of each IMF a new filter can be establish to denosing signals. Forthermore the nonstationary signals with different SNR (Signal to Noise Ratio) were simulated to validate this filter and compared denosing method based on wavelet analysis. The compareing indexs showed that the denosing effect with EEMD filter excelled the wavelet method appreciably, but avoid the diffecuty to select a proper wavelet. With this method, the signal was decmpsed based on the scales in itself and have the most adaptivity.
展开▼