联合特征提取水声通信信号调制识别方法

A Joint Feature Extraction Method for Modulation Recognition in Underwater Acoustic Communication Signals

摘要:

为提高复杂水声信道环境下非合作数字通信信号的调制识别准确率，针对水声通信中常见的MPSK（M=2、4、8）、2FSK、4FSK和OOK等典型信号开展调制识别研究。提出一种联合特征提取方法，通过结合连续小波变换与频域分析提取多种时频域特征；针对水声通信中MPSK类内识别难题，创新性地提出基于K-means的似然类内识别方法，有效克服了水声环境下星座图未知带来的识别障碍。实验结果表明：在典型水声信道条件下，信噪比高于-5dB时，MPSK、2FSK、4FSK及OOK信号的总体识别率达到93.2%；信噪比高于5dB时，MPSK信号的类内识别率可达96.5% 。所提出的识别框架针对水声通信环境特点进行了优化，在低信噪比条件下展现出优异的鲁棒性和识别性能，为复杂水声非合作通信场景中的调制识别提供了有效解决方案。

Abstract:

To enhance the modulation recognition accuracy of non-cooperative digital communication signals in complex underwater acoustic channels, this study focuses on the modulation recognition of typical signals commonly encountered in underwater acoustic communication, including MPSK (M=2, 4, 8), 2FSK, 4FSK, and OOK. A joint feature extraction method tailored is proposed, which combines continuous wavelet transform with frequency domain analysis to extract multiple time-frequency domain features. Addressing the challenge of intra-class recognition for MPSK in underwater acoustic communication, an innovative likelihood-based intra-class recognition method utilizing K-means clustering is introduced, effectively overcoming the recognition obstacles caused by unknown constellation diagrams in underwater environments. The experimental results demonstrate that under typical underwater acoustic channel conditions, the overall recognition rate of MPSK, 2FSK, 4FSK, and OOK signals reaches 93.2% when the signal-to-noise ratio exceeds -5 dB. Furthermore, the intra-class recognition rate of MPSK signals achieves 96.5% when the signal-to-noise ratio surpasses 5 dB. The proposed recognition framework is optimized for the characteristics of underwater acoustic communication environments, exhibiting excellent robustness and recognition performance under low SNR conditions, thereby providing an effective solution for modulation recognition in complex non-cooperative underwater acoustic communication scenarios.