As science and technology continue to advance, radar systems have expanded their utility beyond everyday applications, playing an increasingly pivotal role in modern warfare [1]. The precise and efficient identification of radar signals carries profound implications, impacting areas such as missile warning, atmospheric detection, and target tracking. Consequently, this paper delves into the challenge of radar signal identification and classification. Numerous studies have explored the theme of "identification and classification of radar signals in complex environments." Li YJ introduced an integral rotation factor and devised a radial integration method, which, while creative, exhibits sensitivity to noise and reduced recognition accuracy at low signal-to-noise ratios. HU H took a distinct approach, employing spectral correlation analysis and the Gaussian kernel-support vector machine as a classifier for radar signal recognition, yielding enhanced robustness [4]. Dudul ventured into the realm of neural networks for the classification of radar echo signals in ionospheric studies, offering innovative ideas. However, the method's limited generalizability and susceptibility to interference remain drawbacks [5]. Iglesias adopted an automatic modulation classifier characterized by low-complexity signal features and hierarchical decision trees, boasting efficiency and immediacy as advantages. WANG Y C leveraged the Morlet method, employing wavelet ridges as features for radar signal classification after signal wavelet transformation. Yet, this technique presents computational intensity and reduced recognition efficiency [7]. WANG Q integrated a convolutional neural network with a bidirectional long and short-term memory network, achieving higher recognition success rates. However, the introduction of radar signal types led to a significant decrease in classification accuracy. In sum, while existing methods yield promising outcomes in signal classification, they grapple with challenges related to poor immunity against interference and limited classification accuracy under conditions characterized by low signal-to-noise ratios.