基于压缩感知的水声通信网络流量预测

Traffic Prediction for Underwater Acoustic Communication Network Based on Compressive Sensing

水声通信中过多的流量数据给采样和网络传输带来了困难,而压缩感知是一种可行的低速采样理论。提出基于贝叶斯压缩感知理论的水声通信网络流量预测方法。将网络流量预测转化为贝叶斯压缩感知重构问题,为了将需要重构的向量稀疏化,将其设置为受超参数控制的后验概率密度函数。通过该方法可以自适应地找出含有重要信息的网络流量,并用回归算法来进行重构。实验结果显示该方法具有较高的预测精度。

It is difficult to sample it directly and transport network as its too much data in underwater acoustic communica- tion, however, compressed sensing （CS） is a theory that provides a feasible way with lower sampling speed. A traffic pre- diction method for underwater acoustic communication nework based on Bayesian compressed sensing （BCS） is proposed in this paper. Traffic prediction is translated into BCS reconstruction problem, in order to make parse for the vector that demand reconstruction, it sets posterior probability density function that is controlled by hyperparameters to the vector. This method can find the critical network data and reconstruct them with regression algorithm. The experiment results show that this method can reconstruct original underwater acoustic network traffic effectively.