Compensation Method of Received Signal Power observed by Smartphone for Crowdsensed Spectrum Database

Crowdsensing is a technique for observing a wide range of radio environments at high density. Collecting radio environment information using sensors built into smartphones, a huge data set can be acquired while reducing the installation cost of new measurement equipment. By accumulating the data set in Measurement-based Spectrum Database (MSD), a wide range and high-precision Radio Environment Map (REM) can be generated. However, crowdsensing is impossible to measure the received power accurately without considering the power fluctuation due to the difference of the working scenario, e.g. specific conditions of human such as putting a wireless terminal in a pocket, and the error tendency of each terminal. In this research, we have a measurement campaign focusing on measurement errors caused by working scenarios and smartphone types in the cellular environment. Also, we propose the compensation method of observed values using simple linear regression. In the proposed method, the observed value of the terminal and one of the measurement equipment for area evaluation for LTE system are compared. Those data are classified in the MSD, and the mesh IDs are obtained. Then, those data matching mesh IDs are plotted to obtain a simple regression line. The predicted value is compensated by substituting the observed value of the terminal into the regression line. Comparing the difference between the ground-truth and the predicted value, the proposed method can improve the accuracy of REM under all working scenarios and smartphone types.