Wireless threat detection device, system, and methods to detect signals in wideband RF systems and localize related time and frequency information based on deep learning

The present invention comprises a novel system and method to detect and estimate the time-frequency span of wireless signals present in a wideband RF spectrum. In preferred embodiments, the Faster RCNN deep learning architecture is used to detect the presence of wireless transmitters from the spectrogram images plotted by searching for rectangular shapes of any size, then localize the time and frequency information from the output of the FRCNN deep learning architecture.

Recommended citation: Koppisetti, N. R. S. V. P., Dsouza, K. B., Boostanimehr, H., & Mallick, S. (2022). U.S. Patent Application No. 17/825,304. Full Paper