Automatic Detection of Modulation Scheme Using Convolutional Neural Networks

Abstract

Automatic Modulation Classification (AMC) assumes a pivotal role in wireless communication systems. This paper presents an innovative intelligent modulation scheme detection approach utilizing a Convolutional Neural Network (CNN). The primary objective is to accurately identify the modulation schemes in the incoming signals without the need for selective feature extraction. The methodology involves transforming raw modulated signals into a 2D format and training a specialized CNN architecture. This novel approach eliminates the manual feature extraction process, simplifying the overall procedure and reducing computational complexity. The CNN learns intricate patterns and variations within the signals, enabling precise classification. Rigorous testing and validation demonstrate the high effectiveness of the CNN, achieving a remarkable prediction accuracy of 87.39%. The simulation results unequivocally substantiate the exceptional performance of the proposed. Furthermore, the system's robustness against noise is extensively evaluated and modelled, ensuring its reliability in real-world scenarios where signals are frequently corrupted by various forms of interference. The unique training methodology, well-designed CNN architecture and comprehensive evaluation of noise performance contribute to the novelty and efficacy of the proposed system.