AUTOMATIC DETECTION AND RECOGNITION OF DIGITAL MODULATED SIGNALS USING A NEURAL NETWORK

Abstract

Issues on automatic detection and recognition of digital modulated signals are becoming increasingly relevant in modern communication systems. These processes play a key role in ensuring reliable and efficient communication in a wide range of areas, from mobile communication to broadcasting and satellite systems. The importance of these processes continues to grow with the increase in the amount of data transmitted through these systems and with the development of new communication technologies. Accurate detection and classification of signals allow communication systems to correctly interpret the transmitted information, which, in turn, ensures the reliability and efficiency of data transmission. This is especially important in conditions when signals are subject to various types of interference and distortions. Therefore, the development and improvement of the methods for automatic detection and recognition are becoming increasingly important for ensuring high-quality communication. The present article presents a neural network-based system capable of automatically detecting signals with a single carrier frequency in the range up to 6 GHz and classifying four types of digitally modulated signals: BPSK, QPSK, 16-QAM, and 64-QAM. The system is capable of operating in the presence of additive white Gaussian noise, phase and frequency offset in the received signal. The algorithms for detection and recognition of signals, developed in the LabVIEW software environment are presented. The architecture and hyperparameters of the neural network allowing quick and efficient training of the network are described. These algorithms and training methods can be applied and adapted for other systems and applications, making them very useful for a wide range of tasks. The system was tested both on simulated signals and in a real communication channel. The results obtained confirm the effectiveness of the proposed methods and emphasise their potential for further application and research.