Threshold exploration via particle swarm optimizer at profitable wavelet decomposition for noise reduction

Recently, the research which based on wavelet representation to reduce noise has gotten a lot of attention. The most typical method, universal threshold proposed by Donoho, and its derivative methods have verified their efficiency on varied applications. Some settings which are signal-dependence are critical; however, they were usually given by trial and error or a rough estimate in existing algorithms. This paper addresses the assumption that source signals and additive noise are mutually independent to deal with de-noisy problem without extra prior knowledge. First, the objective function developed from a technique of blind source separation (BSS) is applied on particle swarm optimizer (PSO) to guide the determination of wavelet threshold. Second, the evaluation which is able to determine the most profitable decomposition scale for de-noisy is proposed. Therefore, the best performance of de-noisy could be obtained. In order to confirm the validity and efficiency of the proposed algorithm, several simulations which include four benchmarks with different noise degree are designed. The performance of proposed algorithm further compared with that of other existing algorithms.