The mobile crowd sensing technology in the environment integrating human, machines and things is an emerging direction in social computing. In kinematics research, continuous blood pressure monitoring and calibration are the basis for revealing the correlation between athlete motor function and blood pressure. At the same time, in the field of medical research, hypertension can be more easily controlled, thus improving the effectiveness of hypertension treatment. This paper presents the design principle of a human-machine fusion system based on CrowdOS, a mobile crowd sensing platform. The system innovatively establishes the correlation between blood pressure and exercise, improves the accuracy of cuffless blood pressure measurement, and verifies the feasibility of calibrating continuous cuffless blood pressure measurement based on exercise information. Using our system and electronic cuff sphygmomanometer, we measured 65 groups of data in walking, running, sitting and climbing stairs, each group lasting about 10 minutes. Based on these data, we established a regression analysis model for blood pressure measurement calibration. The accuracy of blood pressure calibration was improved from the original systolic root mean square error of 13.43mmHg and diastolic root mean square error of 8.35mmHg to 9.76mmHg and 5.56mmHg. The design method proposed in this paper provides a feasible solution for continuous cuffless blood pressure measurement and calibration, and shows broad application prospects in the fields of athlete scientific training and medical care.