earticle

영어

This study proposes a novel methodology for quantitatively evaluating the contribution of input signals in the time domain using Mutual Information (MI). Traditional contribution analysis methods based on Pearson correlation coefficients are limited by their assumption of linearity, making them inadequate for systems with time-varying characteristics or nonlinear transfer paths. To address this, we construct simulation data comprising transient, non-stationary input signals and nonlinear transfer functions, and compute time-local mutual information by adopting the windowing approach commonly used in Short-Time Fourier Transform (STFT). The results demonstrate that the proposed MI-based method outperforms conventional linear techniques in capturing the contributions of inputs under nonlinear and time-varying conditions. Notably, the MI approach provides accurate quantitative assessment even when the system's transfer path responds nonlinearly to input amplitude. This study shows that MI-based contribution analysis is a powerful and effective tool for evaluating input influence in nonlinear, non-stationary, and multi-input systems, and lays a foundation for future applications to experimental data and integration with alternative MI estimation methods.