Existing automatic control of building thermal environments do not consider the individual's real-time thermal sensation, which could reduce the occupants' thermal comfort. Therefore, it is very important to accurately obtain an individual's thermal sensation and real-time reflect on the control logic of air-conditioning systems. Current thermal sensation estimation models mostly apply to sedentary condition without considering human sensation in different activity states, which caused these models have critical limitations in accurately predicting human thermal sensation. In this paper, an intelligent wristband device is used for online monitoring of human thermal characteristics in different active states. The wrist skin temperature and its time differential as well as the heart rate are used for the evaluation index of human thermal sensation, and a series of environmental chamber experiments are carried out to obtain the relationship between the wrist skin temperature and thermal sensation in different activity states in summer. The correlation models of human thermal sensation, wrist skin temperature and its time differential, and heart rate has been formulated by statistical analysis and correlation analysis. In order to verify the feasibility of correlation models in the unstable environmental condition, several tests were conducted in the actual built environment. This study indicates that the wrist skin temperature and its time differential and heart rate can be used for estimating human thermal sensation with a high degree of accuracy in the different activity states. In addition, results of this study also demonstrate the promising applicability of obtained correlation models in the unstable environmental condition. (C) 2018 Elsevier B.V. All rights reserved.