A Disturbance Suppression Framework Based on Online Prediction for the Dynamic Measurement of Skin Temperature

Body temperature usually measured through skin plays a critical role in assessing human health status. The response time in conventional wearable temperature monitors is relatively slow due to the requirement of thermal equilibrium. Dynamic measurement of skin temperature before thermal equilibrium suffers from disturbance due to sensor noise. This paper presents a suppression framework that improves the performance of the theoretical parametric method on heat conduction. By using least squares regression of limited temperature samples, variables in theoretical models could be corrected to produce a fast and early prediction of the final equilibrium temperature. Experiments were conducted with 16 subjects to evaluate the performance of this method. Error comparison experiment results show that the error suppression method achieved a mean absolute error (MAE) of 0.046°C with 3 data samples at 90-second intervals, and an MAE of 0.054°C with 5 samples at 30-second sampling intervals. The effective reduction of error demonstrates the feasibility of this framework. The method is simple to implement in embedded devices, showing promising potential in the online prediction of dynamic skin temperature.