Robust and Efficient Modulation Recognition Based on Local Sequential IQ Features.

Wei Xiong,Petko Bogdanov,Mariya Zheleva

DOI: https://doi.org/10.1109/infocom.2019.8737397

2019-01-01

Abstract:Modulation recognition plays a key role in emerging spectrum applications including spectrum enforcement, resource allocation, privacy and security. While critical for the practical progress of spectrum sharing, modulation recognition has so far been investigated under unrealistic assumptions: (i)a transmitter’s bandwidth must be scanned alone and in full, (ii) prior knowledge of the technology must be available and (iii) a transmitter must be trustworthy. In reality these assumptions cannot be readily met, as a transmitter’s bandwidth may only be scanned intermittently, partially, or alongside other transmitters, and modulation obfuscation may be introduced by short-lived scans or malicious activity.This paper bridges the gap between real-world spectrum sensing and the growing body of methods for modulation recognition designed under simplifying assumptions. We propose to use local features, besides global statistics, extracted from raw IQ data, which collectively enable a robust framework for modulation recognition that outperforms baselines from the state-of-the-art. Specifically, we exploit the discriminative power of local patterns from consecutive IQ samples extracted based on a Fisher Kernel framework that captures non-linearity in the underlying data. With these domain-informed features, we employ lightweight linear support vector machine classification for modulation detection. Our framework is robust to noise, partial transmitter scans and data biases without utilizing prior knowledge of the underlying transmitter technology. The recognition accuracy of our approach consistently outperforms baselines in both simulated and real-world traces. We demonstrate up to a 98% accuracy and a 30% improvement over several counterparts from the literature with partial scans in a USRP testbed.