Automatic modulation classification using Expert features and Convolutional Neural Networks.
The network consists of 2 Convolutional Feature Detection layers . The first layer consists of 256 filters (kernels) each of size 3X1 , and the 2nd layer consists of 80 filters each of size 3X2 . We input 2X128 IQ (In phase & Quadrature) samples to the first layer. Convolution layers are followed by a Dense layer with 256 nodes in hidden layer and 11 nodes in the last layer.
