基于CNN-BiLSTM混合神经网络的雷达信号调制方式识别

Radar Signal Modulation Recognition Based on CNN-BiLSTM Hybrid Neural Network

针对具有时频特性的雷达信号,传统的雷达信号识别方法已经无法满足对信号类型精准识别的需求,因此需要通过采集并分析雷达信号脉内的时频特征实现对目标雷达的具体信息进行有效评估。设计了一种卷积-双向长短时记忆(Convolution-Bidirectional Long Short-Term Memory,CNN-BiLSTM)混合神经网络模型,主要通过BiLSTM的时序记忆特性深度挖掘雷达信号的时域特征,结合权值共享特性和CNN层捕获雷达信号的时频特征,再利用二者信号特征联合完成对雷达信号调制方式的识别。通过对比实验验证,所提方法对若干种雷达信号的识别具有较高的准确度,平均值达到95.349%;优于只使用单一特征的网络和传统算法,具有良好的抗噪声能力。

For radar signals with time-frequency characteristics,traditional radar signal recognition methods are unable to meet the needs of accurate recognition of signal types.Therefore,it is necessary to evaluate the specific information of target radar effectively by collecting and analyzing the time-frequency characteristics in the radar signal pulse.A Convolution-Bidirectional Long Short-Term Memory(CNN-BiLSTM)hybrid neural network model is designed.The network mainly uses the time series memory characteristics of BiLSTM to deeply mine the time domain characteristics of radar signals.The weight sharing feature and the time-frequency characteristics of radar signals captured by CNN layer are combined.Then the signal characteristics of the above are used to jointly complete the identification of the radar signal modulation mode.The experimental results show that the method has high accuracy in the recognition of several radar signals,with the average value reaching 95.349%.It is better than the network using only a single feature and the traditional algorithm,and has good anti-noise ability.