基于GANomaly的水声通信信号调制方式开集识别

Open-set identification of modulation modes for underwater acoustic communication signals based on GANomaly networks

摘要:

针对水声通信信号调制方式的开集识别问题,提出了一种基于GANomaly的新型开集识别方法。GANomaly网络模型结合了生成对抗网络(GAN)的生成能力和异常检测技术的判别能力,通过对抗训练,重构调制信号的时频特征并确定误差阈值,进而有效地区分已知和未知的调制方式。在信噪比＞8 dB的条件下进行的仿真实验显示,该方法对已知调制方式信号的识别率均超过了86.00%,对未知调制方式信号的识别率超过了80.00%。仿真实验结果验证了所提方法对未知调制方式的有效识别能力,为水声通信领域开集识别问题提供了新的解决思路。

Abstract:

For the open-set identification problem ofunderwater acoustic communication signal modulation modes, a new open-set identification method based on GANomaly was proposed. The GANomaly network model combines the generative ability of generative adversarial net works (GAN) and the discriminative ability of anomaly detection technology. Through adver sarial training, it reconstructs the time-frequency features of modulated signals and deter mines error thresholds, thereby effectively distinguishing between known and unknown mod ulation methods. Simulation experiments conducted under the condition of a signal-to-noise ratio greater than 8 dB show that the recognition rate of this method for signals with known modulation modes exceeds 86.00%, and for signals with unknown modulation modes, it exceeds 80.00%. The results of the simulation experiment verify the effective recognition ability of the proposed method for unknown modulation modes, providing a new solution for the open-set recognition problem in the field of underwater acoustic communication.