Modulation classification using deep learning technique

Deep Learning (DL) with Convolutional Neural Networks (CNNs) is a very potent machine-learning method, particularly for computer vision applications. This work provides a modulation recognition model that is based on CNN. MATLAB software is utilized to construct an image dataset with various signal spectrograms using the Short-Time Fourier Transform (STFT) in order to completely illustrate CNN’s picture feature extraction capability. Tests indicate that the optimal results can be obtained by employing Stochastic Gradient Descent Momentum (SGDM) as an optimization function. This work examines the recognition accuracy across a variety of configurations at various Signal-to-Noise Ratios (SNRs). The impacts of symbol rate, multi-path fading, and carrier frequency offset are taken into account. The proposed model improves the signal modulation classification accuracy significantly and achieves competitive results. The performed experiments reveal that the proposal guarantees a remarkable classification accuracy of approximately 95.07% at a 30 dB SNR over AWGN.