Cables are one of the most commonly used forms of electrical energy transmission corridors in distribution networks, and their weak points are the terminal joints. Once a fault occurs, it will affect the stability of power supply. Based on the time-domain and frequency-domain correlation analysis of partial discharge signals in cable terminal joints under different defects, this article proposes a cable terminal partial discharge pattern recognition algorithm based on Deep Residual Shrinkage Network (DRSN). Firstly, simulation experiments were conducted on different defects of cross-linked polyethylene cables to obtain partial discharge signal data under different working conditions and defects. The data collected underwent preprocessing utilizing the power frequency synchronization method and a Phase Resolved Partial Discharge (PRPD) graph, resulting in the construction of a dataset. The pulse signal dataset was trained using DRSN network, and a cable terminal joint defect classification model was constructed based on this, and compared and tested. The results show that the proposed model based on Deep Residual Shrinkage Network can achieve an accuracy of 99.7% on the pulse test set, and the classification results are better than the comparison models. This article has practical value in achieving accurate detection of cable terminal joint defects and improving the convenience of cable terminal joint defect detection.