In this work, we report on progress in building a machine learning (ML) algorithm to blindly infer the signal modality of anomalous wireless signals. The system built is designed to be robust to hardware impairments like carrier frequency offset (CFO), sample frequency offset (SFO), wireless channel, sample rate changes due to radio resampling etc. The main novelty of our work is the exploration of metric learning methods for the task of blind modality/modulation classification using cyclostationary features. We describe how the ML approach evolved, with an empirical illustration of improvement in classification accuracy.
Citation and Bibtex
Rajagopal, S., Mathuria, R., & Bharadia, D. (2024). Blind Signal Characterization: Transformers, Triplet Losses and Beyond. In IEEE International Symposium on Dynamic Spectrum Access Networks (IEEE DySPAN 24).
