Advanced CNN-RNN Model Based Automatic Modulation Classification on Resource-Constrained End Devices

In this paper, we propose a hybrid convolutional recurrent neural network (CNN-RNN) model to operate on resource-constrained end (RCE) devices for automatic modulation classification (AMC). It extracts spatial features and temporal dependencies from data signals by exploiting CNN layer and RNN layer respectively in dynamic communication environments. Further, the dense layer is connected to cluster output of the RNN layer to robust the classification. Furthermore, the softmax layer is used to provide probability distribution to each modulation scheme. Our proposed hybrid model outperforms the available existing models in the literature. In addition, we provide promising future research directions.