Abstract: Aiming at the problem of small UAVs threatening the low-altitude field and its difficulty in recognition under complex electromagnetic environment, a signal recognition algorithm based on convolutional neural network (CNN) is proposed.The characteristics of the direct sequence spread spectrum (DSSS) signal under the application of UAVs are analyzed,and the signal time-frequency characteristics are obtained by short-time Fourier transform (STFT).The energy threshold noise reduction method proposed is used to reduce the noise in the time-frequency features.The CNN is designed for classification and recognition.Simulation results show that the algorithm has strong robustness and low signal-to-noise ratio environmental adaptability,and its performance is significantly better than traditional algorithms.When the signal-to-noise ratio (SNR) is -6 dB,the recognition rate reaches 0.97.

Key words: small UAV, direct sequence spread spectrum (DSSS), modulation recognition, convolutional neural network (CNN), short-time Fourier transform (STFT)

摘要： 针对小型无人机威胁低空领域且在复杂电磁环境下难识别的问题,提出了一种基于卷积神经网络的信号识别算法。首先,分析了无人机应用背景下直接序列扩频信号的特性,并采用短时傅里叶变换获取信号时频特征;然后,利用提出的能量阈值降噪法降低时频特征中噪声成分;最后,设计了卷积神经网络用于分类识别。仿真结果表明,该算法在信噪比为-6 dB时可达到0.97的识别率,具备较强的鲁棒性和低信噪比环境适应性,其性能显著优于传统算法。

关键词: 小型无人机, 直接序列扩频, 调制样式识别, 卷积神经网络, 短时傅里叶变换