ShuffleNet-inspired lightweight neural network design for automatic modulation classification methods in ubiquitous IoT cyber–physical systems

Abstract

Automatic modulation classification (AMC) is one of the most important technologies of cognitive radios and ubiquitous internet of things (IoT) cyber–physical systems, and it can be adopted to recognize unknown signals. Recently, deep learning (DL) has been applied into AMC for the advanced classification performance. However, DL-based AMC methods generally have high computation complexity and large model sizes, which means that these methods can be rarely implemented into some IoT devices. In this paper, inspired by ShuffleNet, we design a lightweight convolutional neural network (CNN), which is named as ShuffleCNN, and a ShuffleCNN-based AMC (ShffuleAMC) method is proposed for the ubiquitous IoT cyber–physical systems with orthogonal frequency division multiplexing (OFDM). Besides, we also introduce fast Fourier transform (FFT) to pre-process the OFDM signals for the classification performance improvement, and apply ℓ2 regularization to avoid overfitting. It is demonstrated by simulation results that our proposed ShuffleAMC method has little performance loss, when compared with the common CNN-based AMC methods. More importantly, our proposed ShuffleAMC method also has the strengths of low computation complexity and few model sizes.