High-Reliability Envelope-Based Automatic Modulation Classification Using Deep Learning

This paper presents a novel method for deep learning-based automatic modulation classification (AMC). The method relies on coupling the input signal, and extracting the envelope and the frequency information, as input features to the neural network. The main advantage of this method, compared to traditional IQ-based detection, is that it is more robust since there is no demodulation process which can suffer from non-linear imperfections. This is critical if the target signal is a high-power interferer. The presented method is tested, with hardware implementation, and is capable of detecting the modulation type and rate with an average of 99.6% accuracy. The time required for the modulation classification is approximated to be 9.7 μs. The behavior of the system in the presence of noise is also tested and presented. Compared to the state-of-the-art, this method provides more information about the interferer (modulation and rate), in a shorter time, without compromising the accuracy.