基于支持向量机分类器的数据链信号调制方式识别

摘要

In consideration of modulation recognition technology of communication signals playing an important role in non-cooperative communications, and for the modulation systems (MSK, BPSK, QPSK, OQPSK,π/4-DQPSK and 8PSK) used for US army data link including Link-11, Link-16, Link-22 and CDL, this paper analyzes the extraction of characteristic parameters on square spectrum, high-order cumulant, quartic spectrum, symbol rate and etc. of various modulation signals under the condition of aeronautical channel. And then it carries out the modulation recognition of six types of data link signals using the classifier of support vector machine (SVM), and optimizes the classifier parameters of SVM by utilizing the particle swarm optimization (PSO) algorithm. Simulation results show that SVM classifier can improve the whole recognition rate under the condition of low SNR, compared with the decision tree classifier, and decrease the blindness for selection of SVM classifier parameters by using the POS algorithm.%鉴于通信信号的调制方式识别技术在非合作通信中具有重要的地位，针对美军Link-11、Link-16、Link-22、CDL数据链信号的调制方式(MSK、BPSK、QPSK、OQPSK、π/4-DQPSK、8PSK)，研究了各调制信号在航空信道条件下的平方谱、高阶累积量、四次方谱、码元速率等特征参数的提取。然后，运用支持向量机(SVM)分类器对六种数据链信号的调制方式进行了识别，并利用粒子群优化(PSO)算法对支持向量机分类器参数进行优化。仿真结果表明，相比决策树分类器，SVM分类器在低信噪条降下提高了整体识别率；采用PSO算法则减少了SVM分类器参数选择的盲目性。