Highly Efficient Two-Stage Filtering-Based Maximum Likelihood Stochastic Gradient Algorithm for Multiple-Input Multiple-Output Systems

The article delves into the challenges of system identification for complex multiple-input multiple-output (MIMO) systems, which are often characterized by intricate structures and numerous parameters. It introduces a novel two-stage filtering-based maximum likelihood stochastic gradient (TS-F-ML-SG) algorithm designed to enhance estimation accuracy and computational efficiency. The algorithm leverages advanced filtering techniques and the maximum likelihood principle to handle systems interfered by autoregressive moving average (ARMA) noises. A detailed comparison with an existing auxiliary model-based maximum likelihood stochastic gradient (AM-ML-SG) algorithm is provided, highlighting the superior performance of the proposed method. The article includes a thorough analysis of computational complexity and presents simulation results that demonstrate the algorithm's effectiveness under various noise conditions. Additionally, it discusses the potential applications of the proposed method in diverse stochastic systems and suggests avenues for future research to further improve identification techniques.