Aimed at the problem of detecting electrical shock current fast and correctly, this paper proposes an N-LMS adaptive algorithm based electrical shock detection method. According to the principle of adaptive filter, it establishes adaptive electrical shock current detection model, and then applies adaptive filters to eliminate the influence of noise and normal residual current in turns from the measured total residual current detected and separates the electrical shock current out. Finally, the proposed method is testified by the electrical shock experimental data which proves the validity of the method. Compared with the increment method, BP neural network based method, LS-SVM based method, and present current-amplitude based method for RCDs, the results conclude that the proposed method has short response time and great noise robusticity and eliminates protection dead-zone. In short, the method could help develop novel RCDs. This work is supported by Science and Technology Department of Sichuan Province, China (No. 2015RZ0055).