Bae Systems Information and Electronic Systems Integration Inc.

Amit Bhatia

Scott A Kuzdeba

John A Tranquilli, Jr

David J Couto

Denis Garagic

Joseph M Carmack

Joshua W Robinson

The discriminability of an RF fingerprint is increased by 'abstracting,' 'enhancing,' and &#x201c;reconstructing&#x201d; a digital signal before it is transmitted, where the abstraction is a reversible nonlinear compression, the enhancement is a modification of the abstracted data, and the reconstruction is a mapping-back of the abstraction. During a training phase, for each individual RF transmitter, RF fingerprints are analyzed and candidate enhancements are modified until a successful enhancement is identified that provides satisfactory discriminability improvement with minimal signal degradation. The successful enhancement is implemented in the RF transmitter, and the RF fingerprint is communicated to receivers for subsequent detection and verification. Reinforcement learning can direct modifications to the candidate enhancements. The abstraction can implement a deep generative model such as an auto-encoder. A covert data enhancement can encode covert data onto the RF fingerprint, whereby the covert data is transmitted covertly to a receiver.

RF fingerprint enhancement by manipulation of an abstracted digital signal

An artifact-suppressing neural network (NN) kernel comprising at least one neural network, implemented in replacement of a DSP, provides comparable or better performance under non-edge conditions, and superior performance under edge conditions, due to the ease of updating the NN kernel training without enlarging its computational footprint or latency to address a new edge condition. In embodiments, the NN kernel can be implemented in a field programmable gate array (FPGA) or application specific integrated circuit (ASIC), which can be configured as a direct DSP replacement. In various embodiments, the NN kernel training can be updated in near real time when a new edge condition is encountered in the field. The NN kernel can include DCC lower layers and dense upper layers. Initial NN kernel training can require fewer examples. Example embodiments include a noise suppression NN kernel and a modem NN kernel.

Neural network kernels for signal processing in lieu of digital signal processing in radio receivers

Growing community of inventors

Apex, NC, United States of America

Amit Bhatia