An Improved YOLO Algorithm for Identifying Civil Aviation Suppression Interference Signals

In view of the current situation that there are many types of civil aviation interference sources and the interference identification algorithm is relatively scarce in the field of interference detection, an improved YOLOv7-ESC interference identification algorithm is proposed. Firstly, four common suppression interferences in civil aviation signals are modeled and an interference data set is constructed; secondly, continuous wavelet transform (CWT) is introduced as a time-frequency domain processing and analysis method to highlight the time-frequency feature information of the signal; then, the ECA attention mechanism, SE attention mechanism, and CBAM attention mechanism are integrated into the YOLOv7 backbone network to enhance the signal feature extraction capability; finally, a YOLOv7-ESC algorithm that integrates three attention mechanisms is studied to accurately classify and identify different interference signals. Experimental results show that compared with the traditional YOLOv7, the recognition accuracy (P) of the YOLOv7-ESC model increased from 0.930 to 0.986, an increase of 6.0%; the mean average precision (mAP) increased from 0.975 to 0.982, an increase of 0.7%; and the recall rate (R) increased from 0.965 to 0.989, an increase of 2.5%. The YOLOv7-ESC model has obvious advantages in interference identification and anti-interference capabilities, and has broad application prospects in the field of accurate investigation and identification of civil aviation interference sources. 针对民航干扰源种类繁多以及干扰识别算法在在干扰检测领域较为匮乏的现状，提出了一种改进的YOLOv7-ESC干扰识别算法。首先，对民航信号中四种常见的压制式干扰进行建模，并构建干扰数据集；其次，引入连续小波变换（CWT）作为时-频域处理分析方法突出信号的时频特征信息；然后，在YOLOv7骨干网络中融合ECA注意力机制、SE注意力机制、CBAM注意力机制以增强信号特征提取能力；最后，研究出一种融合了三种注意力机制的YOLOv7-ESC算法来对不同干扰信号进行精准的分类识别。实验结果表明，与传统YOLOv7相比，YOLOv7-ESC模型识别精度（P）由0.930提高到0.986，增加了6.0%；均值平均精度（mAP）从0.975提升至0.982，增加了0.7%；召回率（R）则从0.965提升至0.989，增加了2.5%。YOLOv7-ESC模型在干扰识别和抗干扰能力方面具有明显优势，在民航干扰源精准排查与识别领域有广阔的应用前景。