Abstract: Applying a steep pulse voltage of appropriate amplitude to a cell membrane can induce transient and reversible breakdown of the membrane, which has broad application prospects in biomedicine and clinical fields. However, the noise generated by the steep pulse seriously interferes with a patient’s electrocardiogram (ECG) signal resulting in decrease in the accuracy of the ECG feature point detection algorithm. Thus, doctors are unable to understand the state of the patient during treatment, thus limiting complete benefits of the therapy. To eliminate the interference caused by the steep pulse, we analyzed the characteristics of steep pulse interference and established the mathematical model of steep pulse noise. Moreover, we proposed an ECG signal filtering algorithm based on variational mode decomposition (VMD) to extract the steep pulse interference component superimposed on the ECG signal. The proposed algorithm could identify and eliminate the steep pulse interference component. We also designed an ECG signal preprocessing algorithm to reduce the decomposition layer of the VMD algorithm, which improved the real-time performance and reduced the memory consumption. To identify the random noise in the medical environment accompanied by the occurrence of steep pulses, we analyzed the characteristics of random noise in the sub-signal after VMD. Further, we proposed a threshold denoising algorithm based on VMD for sub-signal energy estimation. On the basis of the characteristics of a band-pass filter bank with VMD, we proposed a QRS complex detection algorithm based on VMD sub-signal recombination. Combined with the filtering algorithm, the proposed algorithm was able to improve the accuracy of ECG signal detection. By conducting experiments on ECG signals from the MIT–BIH database with Gaussian white noise and simulated steep pulse interference and those collected in the medical environment, we compared and analyzed the filtering algorithm and QRS complex detection algorithm.