Deep learning for communication signal classification – part A

• Main Author: Wang, Chien Wei
• Other Authors: Alex Chichung Kot
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/167179
• Institution: Nanyang Technological University

Deep Learning methods have seen significant success in a variety of applications in recent years. Due to its feature extraction capability, it can be widely used to solve specific problems in different domains. One area where Deep Learning has been applied is in communication signal modulation classification. Automatic classifiers can be used to determine the types of modulated signals present between the transmitter and receiver, which has important applications in both military and commercial sectors. The focus of this project is to investigate and compare the performance of various Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN) for common communication signal classification tasks. There are ten types of signals included in this testing: Binary Phase Shift Keying, Quadrature Phase Shift Keying, 8 Phase Shift Keying, 16 Quadrature Amplitude Modulation, 32 Quadrature Amplitude Modulation, 64 Quadrature Amplitude Modulation, Gaussian Frequency Shift Keying, Broadcast Frequency Modulation, Double Sideband Amplitude Modulation, and Single Sideband Amplitude Modulation. This report would first review the literature on neural networks, deep learning, and different neural networks. Secondly, the process of generating the 10 types of modulation signal datasets by using MATLAB will be discussed. Finally, the different neural networks were built, trained, and tested. Through the testing, relationships were discovered, and conclusions were drawn. The results indicate that, in general, the RNN-based models have better performances compared to the CNN-based model.