Real-time Automatic Modulation Classification using RFSoC
Stephen Tridgell, David Boland, Philip H. W. Leong, Ryan Kastner, Alireza Khodamoradi, Siddhartha
Abstract
The computational complexity of deep learning has led to research efforts to reduce the computation required. The use of low precision is particularly effective on FPGAs as they are not restricted to byte addressable operations. Very low precision activations and weights can have a significant impact on the accuracy however. This work demonstrates by exploiting throughput matching that higher precision on certain layers can be used to recover this accuracy. This is applied to the domain of automatic modulation classification for radio signals leveraging the RF capabilities offered by the Xilinx ZCU111 RFSoC platform. The implemented networks achieve high-speed real-time performance with a classification latency of ≈8 μs, and an operational throughput of 488k classifications per second. On the open-source RadioML dataset, we demonstrate how to recover 4.3% in accuracy with the same hardware usage with our technique.