Ensemble ML for RF Signal Classification Benjamin J Gilbert College of the Mainland Robotic Process AutomationDownload
We study lightweight ensembles that mix deep and
traditional models for RF modulation recognition. We compare
majority vs. confidence-weighted voting, with optional feature
fusion and classical models, and report accuracy, macro-F1,
latency and calibration (ECE) across SNR. Our reproducible
pipeline evaluates seven modulation classes across SNR conditions
from-5 to +15 dB.
