Identifikasi Sinyal Modulasi ASK Menggunakan VGGNet dengan Batch Normalization

Abstract

Automatic signal modulation classification is a crucial component in digital communication systems, particularly for the development of intelligent signal recognition technologies. This research aims to identify Amplitude Shift Keying (ASK) signals as friendly signals and distinguish them from enemy signals, namely Binary Phase Shift Keying (BPSK) and Quadrature Phase Shift Keying (QPSK). The classification process is carried out using a modified Convolutional Neural Network (CNN) architecture based on VGGNet, enhanced with Batch Normalization layers. The signal dataset was generated using GNU Radio and converted into time–frequency spectrum, which serve as input to the model. A custom VGGNet-6 architecture was employed, with Batch Normalization inserted after each convolutional layer to stabilize activation distributions and accelerate the training process. The experimental results show that the model with Batch Normalization achieved a validation accuracy of 90.77%, outperforming the model without normalization, which only reached 85.22%. These findings indicate that Batch Normalization significantly enhances model performance in distinguishing between friendly and enemy signals. This study demonstrates the potential of efficient CNN-based architectures for automatic signal recognition in wireless communication systems.