A Hybrid Core Temperature Estimation Method Integrating Physiological and Environmental Factors

Accurate estimation of core body temperature ($T_{c}$) is vital for monitoring thermoregulation and preventing heat-related illnesses under extreme environmental conditions. Invasive measurements are unsuitable for continuous use, motivating noninvasive computational alternatives. This study proposes a hybrid $T_{c}$ estimation method that integrates physiological signals, including heart rate ($\text{HR}$) and skin temperature ($T_{s}$), with environmental temperature ($T_{a}$) and humidity. The approach combines two differential equation-based thermoregulation models with a random forest regression model to improve predictive performance. Model parameters were jointly optimized, and performance was validated using leave-one-out cross-validation. The hybrid method achieved an RMSE of 0.24 $^\circ \text{C}$ and a 95% limits of agreement of [−0.47 $^\circ \text{C}$, 0.47 $^\circ \text{C}$].