Enzyme action optimizer based infinite impulse response filter identification through a comprehensive benchmark across full and reduced orders.

Due to the nonlinear and multimodal nature of infinite impulse response filter coefficient spaces, achieving stable and accurate identification remains a challenging task in digital signal processing. This study aims to evaluate the capability of the recently developed bio-inspired metaheuristic algorithm, called the enzyme action optimizer, for adaptive identification of infinite impulse response (IIR) filter models with varying structural complexities. The algorithm is tested on four benchmark systems with varying levels of order and structural complexity. For each system, both full-order and reduced-order identification scenarios are examined, resulting in eight independent experiments. In every case, enzyme action optimizer is directly compared with several well-known optimization algorithms including starfish optimization algorithm, hippopotamus optimizer, and grey wolf optimization. The results are evaluated using multiple performance criteria, including mean squared error, mean absolute error, standard deviation, and convergence speed, providing a comprehensive assessment. The results clearly show that enzyme action optimizer provides highly precise identification across all cases. In low-order systems with full-order modelling, the algorithm achieves perfect reconstruction with zero mean squared error, demonstrating its ability to exactly match system dynamics. In reduced-order setups, where structural simplification introduces modelling challenges, enzyme action optimizer consistently delivers the lowest errors and most stable convergence profiles. Even in high-order and asymmetric systems, enzyme action optimizer maintains its advantage, outperforming all comparative methods both in terms of accuracy and convergence speed. The success of enzyme action optimizer stems from its biologically inspired structure, where two control parameters, adaptive factor and enzyme concentration, regulate the balance between exploration and exploitation throughout the optimization process. This dynamic control enables enzyme action optimizer to search broadly in early iterations and focus on fine-tuning in later stages. The study establishes a unified benchmarking framework that uniquely demonstrates how a biologically inspired optimizer can be effectively adapted to complex infinite impulse response identification problems, highlighting its novelty and potential for broader optimization applications.