Spectral triage in contested environments faces the dual challenge of rapid classification and reliable subsequent operations.
FFT-based feature extraction provides deterministic latency
and interpretable spectral characteristics, making it suitable
for real-time systems where compute budgets are constrained.
However, pure FFT approaches suffer in low-SNR conditions
where CNN methods excel through learned representations.
REFERENCES
[1] J. Smith et al., “Reliable Command Delivery in Contested RF Environments,” IEEE MILCOM, 2023.
[2] L. Chen et al., “SNR Estimation from FFT Features for Link Quality
Prediction,” IEEE Trans. Wireless Commun., vol. 23, no. 4, pp. 2156-2167,
2024.
[3] K. Lee et al., “FEC/ARQ Tuning via CSI Quintiles for Tactical Networks,”
IEEE Internet of Things Journal, vol. 12, no. 2, pp. 1023-1034, 2025.
[4] T. O’Shea and J. Hoydis, “An Introduction to Deep Learning for the
Physical Layer,” IEEE Trans. Cognitive Communications and Networking,
vol. 3, no. 4, pp. 563-575, 2017.