A new approach to select adaptive Intrinsic Mode Functions (IMFs) of Empirical Mode Decomposition (EMD)

Abstract

In the field of signal processing an adaptive algorithm for the selection of Intrinsic Mode Functions (IMF) of Empirical Mode Decomposition (EMD) is a time demand. In this paper, we propose an effective model for adaptive selection of IMFs after decomposition. This proposed algorithm decomposes an input signal using EMD, then the resultant IMF’s are passed through a trained Support Vector Machine (SVM) for the separation of relevant and irrelevant IMF’s. The irrelevant IMF’s are then de-noised. And all IMFs are then reconstructed. The proposed model selects IMF adaptively without any human supervision and helps achieving higher Signal to Noise Ratio (SNR) while keeping Percentage RMS Difference (PRD) and Max Error low. Experiment results show up to 36.16% SNR value, PRD and Max Error are reduced to 1.557% and 0.085%, respectively.