A Ship Yaw-Angle Separation Method Based on Empirical Mode Decomposition and Multi-Dimensional Physical–Statistical Evaluation

The yaw-angle signal output by shipboard attitude sensors (e.g., inertial navigation systems) is a composite of the heading and wave-induced yaw (WIY). The heading reflects large-scale directional changes due to maneuvering or voyage planning, exhibiting slowly varying, trend-like characteristics; WIY is the oscillatory motion caused by random wave–hull interaction, approximately following a zero-mean normal distribution. Accurately and adaptively separating these two components from the composite yaw-angle signal is a key technical challenge in ship motion monitoring or wave parameter inversion. This paper proposes a heading–WIY separation method based on Empirical Mode Decomposition (EMD) and multi-dimensional physical–statistical evaluation. The method adaptively decomposes the composite yaw signal via EMD and automatically determines the optimal mode combination through statistical evaluation. Moreover, a short-time segment processing strategy and overlap-region continuity checks are introduced to overcome heading trend variations over long time scales. The method is validated using a multi-scenario simulation dataset encompassing four conditions and various sea states, as well as at-sea collected data. Results demonstrate high extraction accuracy without requiring a system dynamics model or scenario-specific parameter tuning across all conditions. And the proposed method performs better in comparison to the conventional methods, particularly under conditions of spectral overlapped or high sea states.