A novel technique for automatic hybrid modulation classification in UWA communications using 3D constellation diagrams

Abstract

Underwater acoustic (UWA) communication systems operate in severe channel impairments, including strong multipath propagation, long delay spreads, frequency selectivity, Doppler effects, and high ambient noise. These challenges significantly complicate automatic modulation classification (AMC), especially in dense underwater networks where interference further degrades performance. In this paper, we investigate AMC for frequency-indexed three-dimensional hybrid modulation schemes, namely frequency-phase keying (FPK) and frequency quadrature amplitude modulation (FQAM), in UWA environments. While AMC has been extensively studied for conventional modulation formats, its application to frequency-indexed hybrid modulation schemes in UWA communication systems has received comparatively limited attention in the existing literature. The proposed approach depends on a three-dimensional (3D) constellation representation for robust feature extraction and modulation discrimination. The resulting 3D signal representations are converted into image-based inputs and processed with deep convolutional neural network (CNN) models, including AlexNet, VGG-19, and ResNet50, to automatically extract discriminative features and enable reliable classification. The proposed framework is evaluated under both single-carrier (SC) and orthogonal frequency-division multiplexing (OFDM) transmission schemes to comprehensively assess its robustness and adaptability. Simulation results demonstrate accurate modulation recognition under severe UWA conditions and low signal-to-noise ratio (SNR), confirming the effectiveness of combining hybrid modulation with deep-learning-based AMC for next-generation UWA communication systems.