The algorithm empirical mode decomposition(EMD) has the issue of mode mixing, and the algorithm ensemble empirical mode decomposition(EEMD) has poor realtime performance. Aiming at these defects, the method of empirical wavelet transform (EWT) combining the EMD with wavelet analysis is proposed. In this paper, the method is taken to segment ECG signal spectrum adaptively, and constructs wavelet filter banks in segmental intervals to extract the single component equipped with tight support, then eliminates direct component and noise residual. After that the signal from other component can be reconstructed. The experimental data comes from the real ECG signal in the MITBIH database, and the simulation result shows that the algorithm removes baseline wander and powerline interference. The parameters of signal noise ratio, meanrootsquare error and autocorrelation coefficient are better than other two adaptive algorithms: EMD and EEMD. The whole running time of algorithm is less than 1 s, which can satisfy the requirements of realtime and accurate ECG monitor.