异常性检测算法在引信干扰信号识别中的应用

目的 解决传统分类引信抗干扰算法因干扰信号难获取、特征信号样本少、正负样本数不平衡而导致计算精度低的问题，克服引信抗干扰算法对样本的依赖性，并提高引信信号识别准确率。方法 通过WVD时频变换的方法，将拆分后的真实含扰引信信号切片进行重组，使其由一维时序信号向二维图片信息进行扩展，基于数据倍增策略，提升算法泛化性，并降低其对真实数据样本的依赖。融合GANomaly与EfficientNet网络，在扩充的引信数据集上进行线下干扰信号特征学习，并对含扰引信图像数据进行线上异常性判断与干扰信号识别。结果 GE-FS网络能够在真实引信小样本信号的基础上进行有效数据扩充，基于扩充数据训练后，引信含扰识别准确率达到98.4%。结论 GE-FS网络能有效针对引信异常信号进行精确检测与识别，可以增强引信系统的抗干扰能力与作战自适应性。

Application of Anomaly Detection Algorithm in Fuze Interference Signal Recognition

The paper intends to solve the problem of low calculation accuracycaused bythe difficult acquisition of jamming signals, the small number of characteristic signal samples and the imbalance of positive and negative samples in traditional classification fuze anti-jamming algorithm, overcome the dependenceon fuze anti-jamming algorithm on samplesand improve the accuracy of fuze signal recognition. Through the method of WVD time-frequency transformation, the split real jamming Fuze Signal slices are reorganized to expand fromaone-dimensional time-series signal to two-dimensional picture information. The data multiplication strategyis used to improvethe generalization of the algorithm and reduce its dependence on real data samples. Through thecombination of GANomaly and EfficientNet, the offline jamming signal feature learning is carried out on the expanded fuze data set, and the online abnormality judgment and jamming signal recognition are carried out on the image data of the disturbed fuze. The experiment proves that the GE-FS network can effectively augment the data on the basis of the real fuze small sample signal. After training based on the augmented data, the accuracy of fuze disturbance identification reaches 98.4%. The GE-FS algorithm can accurately detect and identify the abnormal signals of the fuze, and enhance the anti-jamming ability and operational adaptability of the fuze system.