针对低信噪比条件下一般调制识别算法识别率低的问题,对2ASK、2FSK等6种典型的低阶数字调制信号进行时域特征分析,提取出一组能够明显区分各调制方式的时域特征参数组成特征向量,辅助以随机森林算法,对6种典型的低阶数字调制信号进行自动分类识别.所提算法克服了决策树过拟合问题,具有特征参数提取简单、计算量小、易于实现、对噪声具有较好容忍性的优点,在低信噪比环境下有良好的识别效果.实验验证表明,在信噪比不小于-5 dB的条件下,所提算法对2FSK、BPSK、4FSK、QPSK的识别正确率可达78％以上;在信噪比不小于3 dB的条件下,所提算法的调制识别正确率达到100％.可见,所提算法对低信噪比条件下的识别性能具有极大的改善.%Aiming at the problem of low recognition rate for general modulation recognition algorithm under low SNR conditions, 6 typical low-order digital modulation signals such as 2ASK and 2FSK are analyzed in time domain. A set of feature vectors which can clearly distinguish the time domain characteristic parameters of each modulation mode is extracted, and aided by random forest algorithm, while the 6 typical low-order digital modulation signals are automatically classified and identified. The proposed algorithm could overcome the over-fitting problem of decision tree, and possesses the advantages of simple feature-parameter extraction, small computation, easy implementation, and good tolerance to noise, and again has good recognition effect in low SNR environment. The experiment indicates that the recognition accuracy of the proposed algorithm for 2FSK, BPSK, 4FSK and QPSK can reach more than 78％ under the condition that the SNR is not less than -5 dB; under the condition that the SNR is not less than 3 dB, the correct modulation recognition rate can reach 100％. It can be seen that the proposed algorithm can greatly improve the recognition performance under low SNR.
展开▼
