An efficient hybrid-based charged system search algorithm for active filter design

Active filter design is one of the complex real-world optimization problems for signal processing applications. Active filters contain essential parameters such as transistors, resistors, coils, and capacitors to calculate the output signal of the filter. These parameters can be estimated using optimization algorithms. Numerous algorithms are developed in the literature to solve optimization problems related to active filter design. Some well-known optimization algorithms are charged system search (CSS), local search (LS), and levy flight (LF) algorithms. Even though these optimization algorithms have their strengths, they still have weaknesses in stacking local minima, leading to less efficient results. Therefore, more efficient, and robust learning methods that get the most benefit from these optimization algorithms' strongest sides are required to improve the convergence ability of the optimization algorithms. This research implements LS, LF, and a Hybrid optimization method containing both LS and LF onto CSS algorithm to estimate active filter parameters. These optimization algorithms are applied to 20 different benchmark functions to compare and validate the significance of the methods. Some other common approaches such as Genetic Algorithm (GA) and particle swarm optimization (PSO) are also included in the performance analysis to enhance the comparison of the methods. According to these 20 test functions' results, CSS-hybrid is the winner by overperforming in 12 functions, whereas CSS-LS and CSS-LF is the winner by overperforming in 7 and 1 function, respectively. In addition, GA and PSO couldn’t be winner in any of the 20 benchmark functions. The proposed and common algorithms are applied to the low-pass (LP) and high-pass (HP) active filter components, which are a real-world problem for improving their exploitation and exploration balance. The obtained results demonstrate that the proposed methods have a remarkable improvement in predicting active filters parameters.