Abstract
A new signal decomposition algorithm is developed in this paper. The goal of signal decomposition is to decompose a complex and nonstationary signal into a set of basic functions, usually called intrinsic mode functions (IMFs). In the proposed algorithm, it is assumed that each IMF has a limited bandwidth, and the frequency bands of various IMFs are disjoint. Based on this assumption, it is suggested to estimate the IMFs using a set of ideal filters designed in the frequency domain. The ideal filter bank is estimated in the proposed algorithm using penalized least-squares, where the weightedℓ0-norm is utilized as the penalty term. In addition, the weighting parameters (used with the weightedℓ0-norm) and the regularization parameter are calculated using techniques based on energy detection over short windows. It was found that the optimum value of the regularization parameter depends on the noise variance and the size of the windows over which the energy is calculated. To maintain the adaptivity of the proposed algorithm, two techniques are suggested for estimating the values of these two parameters from the analyzed signals. The numerical results on simulated and real data show that the proposed algorithm is fast, robust against noise, and outperforms some state-of-the-art signal decomposition algorithms.