Abstract: 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).