An Application Investigate Of Data Fusion In Broadband Electric Field Probe Development

The expanding spectrum of human activities in electromagnetic field has resulted in electromagnetic pollution,leading to a corresponding increasing demand in electromagnetic radiation monitoring.As a meter on quantitative analysis of strength field of electromagnetic waves,electric field probes are widely used in a variety of radiation monitoring scenarios and its working frequency ranges from 1MHz to 18 GHz in general.In recent years,with the application of electromagnetic wave,the demand to measuring the strength field in higher frequency has been developing.It makes a new challenge to the R&D of electric field probes.However,the wider the detecting frequency range is,the more difficulty to ensuring the frequency flatness of the probe.That makes the design and development of wideband electric field probes a difficult problem.In the previous work,our research group had designed a probe with three antennas.Based on this,the author implemented decision fusion on the measured data from three sensor antennas.In this way,it concentrates the effective frequency bands and makes up the local defects of the three antennas,finally achieve the goal of optimization flatness of the E-field probe.In the thesis,a data fusion framework used in broadband e-field probe measurement is designed,which consists of data reading,feature extracting,fusion deciding.The observed data from three sensor antennas is obtained by single chip and used as input to the framework.The ratios among the input data are interpreted as the data feature.On decision level,the extracted data is classified,then is performed the fusion calculation to get E-field measured value.The article highlighted how to take decision evidence effectively,offering four kinds of decision methods based on statistics judgment,BP neural net,KNN and SVM algorithm,as well as comparing result analysis.The result shows that statistical decision method is more precise and stable to determine the classification of the decision and performs better than the other methods in optimizing the flatness of the broadband E-field probe.BP net,as well as KNN and SVM method,may be more intelligent and they can be corrected on time,however they all underperformed in recognizing the “resonance point”.