摘要
由于水声信道中随机、复杂的时、频扩展特性的影响,非合作水声通信信号调制方式的自动识别极具挑战性.考虑到载频等调制参数提取较为困难,本研究基于信号功率谱、平方谱进行无需先验知识的水声通信信号特征参数提取,设计了一种基于多类别最小二乘支持向量机(LS-SVM)的水声通信信号调制方式分类器,该分类器具有泛化性能好、小样本学习能力强的特点,同时可避免传统神经网络分类器存在的过学习、欠学习以及局部最小化等问题.对海上实录信号数据的识别实验结果表明,本方法具有优于神经网络分类器的识别性能和信道稳健性.
Due to random, complex time and frequency spread characteristics of underwater acoustic channels, modulation classifica- tion of the non-cooperation underwater acoustic communication signal is extremely challenging. Considering the difficulty in prior knowledge extraction (such as carrier frequency) ,the spectra and square spectrum features that do not need any prior knowledge are adopted to incorporate with the least-squares support vector machine (LS-SVM) classifier to derive a recognition method for under- water acoustic communication modulation classification. The proposed method is capable of avoiding the drawbacks of the artificial neural network (ANN) classifier such as overfitting,underfitting and local minimum. The experimental modulation classification re- sults obtained with field signals at 4 different underwater acoustic channels show that the performance and the channel robustness of the proposed modulation recognition algorithm are superior to that of the ANN classifier.
出处
《厦门大学学报(自然科学版)》
CAS
CSCD
北大核心
2015年第4期534-539,共6页
Journal of Xiamen University:Natural Science
基金
国家自然科学基金(11274259)
关键词
水声通信信号
谱特征
调制识别
最小二乘支持向量机
underwater acoustic digital modulated signal
spectrum feature
modulation recognition
least squares support vector ma-chine