Remote monitoring method for human body electrostatic potential based on symbolic regression machine learning

Abstract

Remote real-time monitoring of the human body electrostatic potential is of great value to the investigation, analysis, and prevention of electrostatic hazard accidents. The non-contact measurement method inverses the body electrostatic potential by detecting the surrounding electrostatic field. The distribution of electrostatic fields around the human body is easily influenced by the placement of metal equipment and the architectural structure in the application scenario. Therefore, physical modeling-based inversion lacks generality. Field-measured electrostatic signals and symbolic regression machine learning are used to remotely monitor body electrostatic potential. In a 25 m2 laboratory, four non-contact electrostatic sensors, a contact-type body voltage measuring system, and an ultra-wideband positioning system were used to establish the experiment setting. Sixty sets of on-site test data from three participants were used for model training and performance evaluation. The results indicate that the normalized root-mean-square errors of the body electrostatic potential ranged from 0.01 to 0.22. The optimal results satisfy the IEC 61340-4-5:2018 criteria for the precision of the body potential measuring system.