Identification Techniques For Signal Modulation In Underwater Acoustic Communication Based On GAN And CNN

Compared with traditional methods,deep learning demonstrates more significant advantages in communication signal modulation recognition technology,including channel environment adaptability,recognition accuracy,and computational complexity.Due to its outstanding architectural performance,deep learning will become the research focus of the next-generation communication signal modulation recognition technology,making the investigation of deep learning-based communication signal modulation recognition technology of great scientific value.With the continuous advancement of underwater acoustic communication technology,research on underwater acoustic communication signal modulation recognition technology is gaining attention.This paper delves into the underwater acoustic communication signal modulation recognition technology and proposes a method combining Generative Adversarial Networks(GAN)and Convolutional Neural Networks(CNN)for underwater acoustic communication signal modulation recognition.This paper studies six types of underwater acoustic communication signals,including CW signals,LFM signals,BPSK signals,2FSK signals,4FSK signals,and OFDM signals,and conducts an in-depth discussion on underwater acoustic communication signal modulation recognition technology through a deep learning method combining GAN+CNN neural networks and multi-model joint preprocessing of CGAN and pix2 pix GAN.Finally,the algorithm performance of the proposed underwater acoustic communication modulation recognition technology in this paper is verified by simulation data and Weihai sea trial data,proving the effectiveness and robustness of the proposed method.The main research contents of this paper are as follows: First,in the data preprocessing stage,this paper proposes a GAN-based underwater acoustic communication signal modulation denoising and preprocessing scheme.Based on the GAN model,by combining the CGAN model and the pix2 pix GAN model,multi-model preprocessing of power spectrum diagrams based on the Welch transformation is achieved.The dataset expansion method based on CGAN effectively expands the underwater acoustic communication signal dataset,improves the model’s generalization ability,and reduces the risk of overfitting.Secondly,the pix2 pix GAN network model is used to denoise the power spectrum diagrams based on the Welch transformation,significantly improving signal quality,effectively eliminating noise while retaining the main features of the signal,achieving a high signal-to-noise ratio and accuracy,and providing strong support for subsequent signal recognition.Finally,this paper designs a CNN-based underwater acoustic communication signal modulation classification recognition model and verifies the algorithm through simulation experiments and sea trial experiments.Through the analysis of four types of graphs(t-SNE visualization of original image data,CNN model training accuracy and loss curve,algorithm simulation experiment confusion matrix,and t-SNE visualization of CNN network detection features),the effectiveness of the proposed GAN+CNN-based underwater acoustic communication signal modulation classification recognition technology is fully demonstrated,and the excellent performance of the model in processing different types of modulation signals is verified.