A new approach of denoising the regular and chaotic signals using Empirical Mode Decomposition: Comparison and application

The Empirical Mode Decomposition has been used to present a new approach for denoising of regular and chaotic time series originating from the nonlinear systems. The proposed filtering approach is based on the frequency distribution of different Intrinsic Mode Functions (IMFs). It provides complete frequency and noise strength information present in the data set. The actual frequencies in an experimental data set are distributed among different IMFs which are subsequently disentangled through the algorithm resulting into filtered as well as noise components. The filtered signal is produced only after a suppression (hard threshold) of noise components. The noise strength present in the data set is produced by the identified noisy IMFs. This innovative approach has been tested for linear, nonlinear, and chaotic types of calculated data sets with various levels of simulated noise strengths and results have been compared with the existing methods. The validity of the present approach is confirmed by experimentally observed astrophysical data as well as data from a neutron detector. A sharp improvement in the signal strength has been observed with the proposed method in comparison to the existing methods. (C) 2013 AIP Publishing LLC.