Evaluation of Personal Radiation Exposure From Wireless Signals in Indoor and Outdoor Environments

With the development of wireless technology, the public is exposed to electromagnetic fields (EMF), which has led to concerns about the potential health effects of EMF exposures. This paper aims to evaluate personal EMF exposures from wireless signals in indoor and outdoor micro-environments in Malaysia. According to the influencing factors, four different types of micro-environments are selected. A radiation exposure meter called ExpoM-RF 4 is used to measure the electric field strength across these micro-environments. From the measurement campaigns, three machine learning (ML) techniques are simulated to model the Electric Field Strength in each micro-environment. The ML techniques are Fully connected neural network (FCNN), eXtreme Gradient Boosting (XG Boost), and Linear Regression (LR) to predict the RMS and Maximum radiation exposure. From the ML models, Total Emission Ratio (TER), Root Mean Square Error (RMSE) and Coefficient of Determination (R2) are evaluated to measure the performance of ML. By comparison, it is found that LR performs well with single and simple data set, while XG Boost and FCNN demonstrate superior capabilities in handling multiple types of data sets. The FCNN model provides the most accurate predictions, particularly in urban and suburban areas where extreme values are observed. Finally, the measured data and the predicted radiation exposure levels are compared against public exposure limit by International Commission on Non-Ionizing Radiation Protection (ICNIRP), Malaysian Communications and Multimedia Commission (MCMC) and Federal Communications Commission (FCC). The results demonstrate that typically personal radiation exposure is lower than the exposure limit (61.4 V/m), which is similar to the most research results. However, in areas with dense population and numerous base stations, the maximum exposure can approach 56.7365 V/m (measured data), which is close to the exposure limit.