Deep Learning-Based Modulation Classification for Underwater Acoustic Communication: A Convolutional Neural Network Approach

Underwater acoustic communication (UWAC) is crucial for maritime applications but confronts obstacles such multipath fading, Doppler shifts, and noise interference. This research provides a deep learning-based modulation classification system employing CNNs and time-frequency spectrograms. The model achieves 95.6% accuracy at SNR = 10 dB, surpassing VGG-16 (92.3%), ResNet-18 (94.1%), and MobileNetV2 (90.8%). Convolutional layers extract features, while batch normalization and dropout boost robustness. Results exhibit enhanced classification across various SNRs, solving critical challenges in maritime and military networks, boosting UWAC dependability.