L波段数字航空通信系统（LDACS）是民航新一代地空宽带数据链，为了动态占用相邻测距仪（DME）的空闲频谱，实现频谱资源受限下的宽带通信，提出了一种联合小波变换与深度卷积神经网络的频谱感知方法。首先通过小波变换将接收到的信号转换到小波域；然后利用信号差别提取出DME信号的小波系数，对系数进行低频置零和硬阈值处理，以形成数据集，同时将多组信噪比下的数据集进行融合处理；最后使用预处理过的数据对卷积神经网络Vgg19进行训练，得到针对DME信号的频谱感知模型。该方法通过结合小波变换的多尺度特征提取能力和Vgg19的深层特征学习能力，可以在低信噪比环境下提取DME信号的有效特征生成数据集，所训练出来的感知模型，显著提高了感知的准确性和稳定性，在低信噪比环境下表现优异，当信噪比大于-13dB时，检测概率可达100%。

The L-band digital aviation communication system (LDACS) is a new generation of ground to air broadband data link for civil aviation. In order to dynamically occupy the idle spectrum of adjacent distance meters (DME) and achieve broadband communication under limited spectrum resources, a frequency spectrum sensing method combining wavelet transform and deep convolutional neural network is proposed. Firstly, the received signal is transformed into the wavelet domain through wavelet transform; Then, the wavelet coefficients of the DME signal are extracted using signal differences, and the coeffi-cients are subjected to low-frequency zeroing and hard thresholding to form a dataset. At the same time, multiple sets of da-tasets under signal-to-noise ratios are fused for processing; Finally, the preprocessed data is used to train the convolutional neural network Vgg19, resulting in a spectrum sensing model for DME signals. This method combines the multi-scale feature extraction ability of wavelet transform with the deep feature learning ability of Vgg19 to extract effective features of DME signals in low signal-to-noise ratio environments and generate a dataset. The trained perception model significantly improves the accuracy and stability of perception, and performs well in low signal-to-noise ratio environments. When the signal-to-noise ratio is greater than -13dB, the detection probability can reach 100%.