Large Scale Radio Frequency Wideband Signal Detection & Recognition

Abstract

Applications of deep learning to the radio frequency (RF) domain have largely concentrated on the task of narrowband signal classification after the signals of interest have already been detected and extracted from a wideband capture. To encourage broader research with wideband operations, we introduce the WidebandSig53 (WBSig53) dataset which consists of 550 thousand synthetically-generated samples from 53 different signal classes containing approximately 2 million unique signals. We extend the TorchSig signal processing machine learning toolkit for open-source and customizable generation, augmentation, and processing of the WBSig53 dataset. We conduct experiments using state of the art (SoTA) convolutional neural networks and transformers with the WBSig53 dataset. We investigate the performance of signal detection tasks, i.e. detect the presence, time, and frequency of all signals present in the input data, as well as the performance of signal recognition tasks, where networks detect the presence, time, frequency, and modulation family of all signals present in the input data. Two main approaches to these tasks are evaluated with segmentation networks and object detection networks operating on complex input spectrograms. Finally, we conduct comparative analysis of the various approaches in terms of the networks' mean average precision, mean average recall, and the speed of inference.

An Analysis of Radio Frequency Transfer Learning Behavior

“…Therefore, the synthetic data experiments use a simple CNN architecture, shown in Figure 5, that is based on the architecture used in [28], with a reduction in the input size. Although many works have found success using larger input sequences [28,38,39], works such as [16,17] have found 128 input samples to be sufficient. Recognizing that longer input sequences result in increased computation and training time, in this work, 128 raw IQ samples are used as input corresponding to approximately 16-32 symbols depending on the symbol rate of the example.…”

Signal analyzing network tool using deep learning

“…The synthetic data is also cross-compared against other sources, some generated for this project and many cross-compared against libraries such as pytorch. 1,7,8 The model accuracy and confusion matrices with augmentations is shown below, divided into "family" and "format" groups. The family group generally represents the same mechanism producing the modulation, PSK being phase-shift keying, ASK amplitude-shift keying, FSK frequency-shift keying, GFSK gaussian frequency-shift keying, QAM quadrature amplitude modulation, APSK amplitude-phase-shift keying, OFDM orthogonal frequency division multiplexing, FM frequency modulation, AM amplitude modulation.…”