Convolutional Radio Modulation Recognition Networks

TL;DR

This paper demonstrates that deep convolutional neural networks can effectively classify radio signal modulations directly from raw data, outperforming traditional expert feature methods especially in noisy conditions.

Contribution

It introduces a convolutional neural network approach for radio modulation recognition and shows significant performance improvements over traditional expert feature methods.

Findings

Deep CNNs outperform expert feature methods in modulation classification.

CNNs are effective at low signal-to-noise ratios.

Blind temporal learning is viable for radio signal analysis.

Abstract

We study the adaptation of convolutional neural networks to the complex temporal radio signal domain. We compare the efficacy of radio modulation classification using naively learned features against using expert features which are widely used in the field today and we show significant performance improvements. We show that blind temporal learning on large and densely encoded time series using deep convolutional neural networks is viable and a strong candidate approach for this task especially at low signal to noise ratio.