Abstract: A method based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) combined with differential, threshold iterative processing and signal segmentation is proposed to preprocess the impedance cardiogram (ICG) signal and identify multiple feature points. First, CEEMDAN decomposes the ICG signal to obtain multiple modal function components (IMF). According to the highfrequency and low-frequency noise frequencies existing in the ICG, the correlation coefficient method is used to remove the interference noise, and then the ICG signal after noise removal is differentiated and segmented. Looking for feature points B, C, X, and perform the above processing on the signals of 20 volunteers collected clinically to evaluate the accuracy of the algorithm. The final results show that the method can locate the feature points with an accuracy rate of 95.8%, the feature positioning effect is good.