|
Guest >> Sign In |
|
|
DIRECTED
ENERGY
PROFESSIONAL
SOCIETY
Abstract: 24-Symp-012
|
|
|
|
UNCLASSIFIED, PUBLIC RELEASE
RF Receiver Component Identification and Characterization using Deep Transfer Learning
High power microwave (HPM) sources have become increasingly prevalent in directed energy applications. Sensitive RF receivers are now at risk of permanent damage due to their exposure to these higher strength signals that are able to be produced, making the ability to identify vulnerable receivers important. In this research, a detailed method is presented to identify RF receiver chain constituent components from distorted reflections produced by extremely short, high power input signals. Using these reflected signals, we can train a convolutional neural network to identify and characterize receiver components to determine if a particular receiver is vulnerable to permanent damage. This research begins to develop a method to identify constituent components and characteristics of RF receivers from reflections of high incident in-band signals. The reflections are used to train a convolutional neural network to classify images of spectrograms of the normalized reflected signals. The trained network achieved an accuracy rate over 98% for classifying validation data, verifying this method.
UNCLASSIFIED, PUBLIC RELEASE
|