Abstract: The ECG signal is easily interfered by the acquisition equipment and the state of the subject. For this reason, a denoising method combining normalized minimum mean square error (NLMS) and adaptive noise complete ensemble mode decomposition (CEEMDAN) is proposed. Among them, the optimized NLMS algorithm reduced the amount of computation by simplifying the relationship between the step factor and the input signal, and optimized the step factor in combination with the number of iterations to improve the convergence performance of the algorithm; the improved CEEMDAN algorithm combined the statistical properties of white Gaussian noise for all IMFs. The components were tested for significance to identify and filter the components containing noise, so that the clean signal could be separated from the noise signal. The experimental results show that the method has a better denoising effect than CEEMDAN under different noise intensities, and alleviates the contradiction between the traditional NLMS convergence speed and the amount of computation.