Multi component signal decomposition based on chirplet pursuit and genetic algorithms

In this paper, we will propose a method based on genetic algorithms, chirplet atoms and the matching pursuit algorithm that by using prior knowledge about the chirplet parameters can do higher quality decomposition of multi component signals. The Matching pursuit is an iterative greedy algorithm that can be used for decomposing of the biological signals into basis functions in time and frequency domain. Decomposition of a non-stationary multi component biological signal by using chirplet basis functions in the matching pursuit algorithm is an optimization problem. We will use a genetic algorithm for solving this optimization problem. We will compare the mean and variance of estimated parameters of chirplets with Cramér–Rao Lower Bounds (CRLB) and the actual values. As we will see, the algorithm is robust against noise, and can do efficient decompositions. In addition, the results show that the algorithm can improve the SNR of the multi component signals. The traditional Adaptive Chirplet Decomposition (ACD) method is commonly used to decompose the multi-component signals to the chirplets. The ACD method uses the Quasi–Newton optimization method and expectation maximization refinement. The comparison between the results of the proposed method with the ACD method for real experimental data shows the improvement in “Signal to Reconstructed Error Ratio (SRER)” using the proposed method.