The article presents a novel method based on Kullback-Leibler divergence integrated within the Kalman filter framework, aimed at improving the estimation of input parameters and system states. This approach specifically addresses the challenge of uncertainties arising from initial parameter guesses, which can hinder accurate system identification. By leveraging available data more effectively, the method enhances the precision of input-parameter-state estimation. The focus on combining divergence measures with filtering techniques represents a significant step in refining estimation processes under uncertain conditions. The research highlights how this method systematically utilizes observed data to mitigate the impact of initial uncertainties. Overall, the study contributes to advancing input-system-state identification by proposing a data-driven solution within an established estimation framework. This development aligns with ongoing efforts to improve robustness and accuracy in dynamic system modeling.