Continuous respiratory rate (RR) monitoring is critical for managing respiratory conditions such as asthma, lung cancer, and chronic obstructive pulmonary disease, especially in rehabilitative care. Traditionally confined to Intensive Care Units, recent advancements suggest the efficacy of using Photoplethysmogram, Electrocardiogram, and Impedance Pneumography signals for RR estimation outside clinical settings. This study introduces a novel approach combining these signals, using an IP signal as the supervised target, to develop an accurate RR estimation model tailored for rehabilitative applications. Twelve machine learning models are tested, where an Ensemble Stacking Regression model delivers the highest accuracy. Hyperparameter tuning and residual analysis refined that model further. Shapley Additive exPlanations-based interpretation increased the transparency of the prediction. This approach offers a robust, scalable solution for continuous RR monitoring, suitable for clinical and wearable rehabilitative technologies.