Abstract: In response to the long processing time and low accuracy of feature extraction in current relay protection signal processing methods, this study proposes a new signal processing and feature extraction method for relay protection from the perspective of integrated automation. Noise reduction of relay protection signals is achieved through multi-scale decomposition and reconstruction of conjugate matrices. Based on STFT and wavelet energy spectrum analysis, relay protection signals are processed automatically. Within this automated framework, Mel Frequency Cepstral Coefficients (MFCC) are introduced to extract features of electrical signals, enabling high-precision characterization of continuous signals such as relay operating voltage and current, thereby improving the stability and anti-interference capability of fault identification. Experimental results demonstrate that the proposed method processes signals in less than 20 seconds under complex mixed fault conditions, exhibiting low feature extraction errors, robust stability, and high practical value.