Abstract
The process of automatic modulation classification (AMC) has gained importance in recent years. AMC enables receivers to classify an intercepted signal's modulation scheme without any prior information about the signal. Likelihood, Distribution Test- and Feature-based approaches have been proposed in the literature for AMC. All these approaches suffer from very high execution time for the AMC process. This prevents the AMC component from operating in real-time alongside other components in the receiver. Thus, in this work, we propose the use of spiking neural network (SNN) in the labeling procedure of feature-based AMC. Implementing SNN in the labeling procedure will significantly decrease the required execution time for AMC while its classification accuracy remain highly competitive with other approaches such as convolutional and recurrent neural networks (CNN and RNN). We furthermore provide a comparison in terms of the required execution time and classification accuracy between our proposed classifier and two recently published approaches that are built upon the aforementioned neural network platforms. We utilize 16QAM, 32QAM and 64QAM modulation schemes from the real-world RadioML dataset in this comparison. The analysis demonstrates that our proposed classifier compared to RNN- and CNN-based classifiers requires on average 19.52% and 14.9% less execution time while sacrificing only 4.7% in classification accuracy on average when compared to the RNN-based classifier, and achieves on average a 15.7% higher classification accuracy compared to the CNN - based classifier.