Aiming at the problem of poor signal recognition effect in shortwave signal reconnaissance, a shortwave signal recognition method based on signal time-frequency image and WT-YOLO is proposed. This method first converts the shortwave signal into a signal time-frequency image form through short-time Fourier transform (STFT, short time fourier transform). To address the issue of mutual interference between low-frequency semantic information and high-frequency details of the signal, a wavelet transform upsampling WFU module is introduced into the model to enhance the feature fusion ability of the model. A triple receptive field (TRF, triple receptive field) module is designed to solve the problem that a single receptive field cannot extract multiple features. The PIoUv2 module is introduced to improve the model"s positioning accuracy, thereby enhancing the detection and recognition accuracy of the model. Experimental results show that the improved YOLOv11 has a higher recognition accuracy of 96.4% compared to the original network model, with a relative decrease in recognition error rate of 55.6%.