PNN神经网络模型在雷达信号调制类型识别中的应用

Application of probabilistic neural network model in modulation type recognition for radar signal

针对雷达信号调制类型多样且识别成功率较低的问题,提出一种基于瞬时频率特征提取的雷达信号快速识别新算法。首先利用短时傅里叶变换(STFT)对信号进行特征提取,将提取到的特征进行排列组合,最后把重新组合的特征作为PNN神经网络的输入向量对其进行调制类型的识别。其中,通过改进PNN神经网络的输入输出部分使得最后的输出结果为各组合特征输入向量所得结果的最优值。通过计算机仿真对上述方法进行验证,并分析了PNN神经网络的结构和性能,与BP神经网络进行了比较。仿真结果表明,在雷达信号调制类型识别中应用PNN神经网络能大幅提高其识别效率,并拥有明显优于BP神经网络的分类性能。

A new algorithm for radar signal fast recognition based on instantaneous frequency feature extraction is proposed to recognize the modulation types of various radar signals and improve the recognition success rate.The short-time Fourier transform(STFT)is used to extract the features of the signal.The extracted features are arranged and combined,and the recombined features are taken as the input vectors of the probabilistic neural network(PNN)to recognize their modulation types.The input and output sections of the PNN are improved to make the final output result as the optimal value obtained by the input vectors of each combination feature.The proposed method was verified with computer simulation.The structure and performance of PNN are analyzed,and compared with the back propagation neural network(BPNN).The simulation results show that the PNN applied to modulation type recognition of radar signal can greatly improve the recognition efficiency,and has better classification performance than BPNN.