水下
计算机科学
编码(社会科学)
接头(建筑物)
无线
光通信
水声通信
电子工程
信道编码
光无线
解码方法
调制(音乐)
光无线通信
自由空间光通信
信号处理
频率调制
光学滤波器
遥控水下航行器
编码(内存)
光纤
通信系统
相位调制
传输(电信)
工程类
人工智能
数据传输
链路自适应
作者
Meiwei Kong,F Zhang,Hui Zhou,Han Yu,Qunhui Yang,Chen Chen,Harald Haas
标识
DOI:10.1109/jlt.2026.3706971
摘要
Joint modulation and coding recognition (JMCR) can provide the receiver with a complete perception of the transmission format, thereby directly supporting adaptive demodulation and decoding. However, under the combined influence of turbidity and bubble disturbance, the received signals will suffer from amplitude attenuation, signal distortion, and time-varying envelope fluctuation, making the reliable recognition of modulation and coding schemes (MCSs) extremely challenging. To address this issue, we propose a JMCR method based on a parallel heterogeneous multi-task learning framework for turbid and dynamic underwater environments. This framework is composed of an in-phase and quadrature sequence representation module, a parallel convolutional neural network-residual network feature extraction module, a bidirectional gated recurrent unit-based temporal dependency modeling module, and a correlation-aware multi-task recognition module, which are used to extract complementary features, capture temporal dependencies, and improve joint recognition consistency. For 24 MCSs composed of 4/16/64 quadrature amplitude modulation and 4/8/16 phase shift keying modulation formats, along with low-density parity-check codes (with code rates of 0.50 and 0.75) and Polar codes (with code rates of 0.25 and 0.50) coding schemes, we conduct feasibility verification of the proposed method in composite underwater environments with turbidity and bubble disturbance in the laboratory. The results show that the average JMCR accuracy of the proposed method reaches 92.7% under 16 underwater conditions, which includes the 0.6% performance improvement brought about by the correlation-aware supervision. This indicates that the proposed method has great application potential in composite underwater environments with turbidity and bubble disturbance.
科研通智能强力驱动
Strongly Powered by AbleSci AI