<p>Postoperative elbow adhesions and contractures are common complications that can severely restrict joint mobility and impair patients’ daily functioning. Building upon an in-depth investigation and optimization of the Cole-Cole model and variational mode decomposition–Hilbert-Huang transform (VMD-HHT) model, we developed a novel elbow joint orthosis and its corresponding control algorithm using bioelectrical impedance spectroscopy technology. The system incorporates genetic algorithm–back propagation (GA-BP) neural networks for deep learning-based prediction of subsequent motion patterns. The efficacy of the control system was validated through rabbit experiments and simulation studies based on experimental data. Results demonstrated that the system significantly improved elbow range of motion and reduced the incidence of postoperative adhesions and contractures, thereby enhancing rehabilitation outcomes. Furthermore, this work not only provides a theoretical foundation for developing control algorithms in portable rehabilitation devices but also offers robust algorithmic support for pathological signal analysis.</p>