Machine Learning Based Source Reconstruction for RF Desense

Abstract

In radio frequency interference study, equivalent dipole moments are widely used to reconstruct real radiation noise sources. Previous reconstruction methods, such as least square method (LSQ) and optimization method are affected by parameter selections, such as number and locations of dipole moments and choices of initial values. In this paper, a new machine learning based source reconstruction method is developed to extract the equivalent dipole moments more accurately and reliably. Based on the near-field patterns, the proposed method can determine the minimal number of dipole moments and their corresponding locations. Furthermore, the magnitude and phase for each dipole moment can be extracted. The proposed method can extract the dominant dipole moments for the unknown noise sources one by one. The proposed method is applied to a few theoretical examples first. The measurement validation using a test board and a practical cellphone are also given. Compared to the conventional LSQ method, the proposed machine learning based method is believed to have a better accuracy. Also, it is more reliable in handling noise in practical applications.

Department(s)

Electrical and Computer Engineering

Research Center/Lab(s)

Electromagnetic Compatibility (EMC) Laboratory

Sponsor(s)

National Science Foundation (U.S.)

Comments

This work was supported in part by the National Science Foundation under Grant IIP-1440110.

Keywords and Phrases

Antennas; Artificial intelligence; Dipole moment; Electromagnetic compatibility; Feature extraction; Learning systems; Least squares approximations; Magnetic levitation vehicles; Magnetic moments; Partial discharges; Personnel training; Radio interference; Radio waves; Support vector machines; Telephone sets; Cell phone; Desense; Histogram of oriented gradients (HOG); Integrated circuit modeling; Radio frequencies; Radio frequency interference; Learning algorithms; Cellphone; Electromagnetic compatibility (EMC); Machine learning; Radio frequency interference (RFI); Support vector machine (SVM)

International Standard Serial Number (ISSN)

0018-9375; 1558-187X

Document Type

Article - Journal

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2018 Institute of Electrical and Electronics Engineers (IEEE), All rights reserved.

Publication Date

01 Dec 2018

Share

 
COinS