Introduction: ECG recordings obtained with mobile ECG devices are susceptible to different kinds of noise which may affect clinical interpretation. Among other noise components, the broadband electromyographic (EMG) noise is especially challenging for removal due to its spectral overlap with the QRS complex. Method: In this paper, we presented the iterative regeneration method (IRM) for the suppression of the EMG noise from ECG signals. The main part of the IRM method, the filter block, generates a tentative noise-free ECG signal which is based on a modified ensemble averaging approach and accounts for the inter-beat variations. It is thereafter subtracted from the noise-contaminated ECG to obtain EMG signal approximation. The quality of noise removal and morphology preservation were evaluated using the RMSE and cross-correlation between the denoised and the recorded noise-free signal (XCORR). Results: The IRM method was tested on the SimEMG database and compared with benchmark methods including Adaptive Wavelet Wiener Filter (AWWF), Wavelet Transform filter, and Finite Impulse Response filter. The IRM method shows superior denoising performance with an average RMSE of 19.9 mV compared to 21.9 mV of AWWF as the second best. It also yielded the highest correlation with the recorded noise-free signal XCORR=0.983. The morphology preservation at fiducial points P, R, J, and T is confirmed by Pearson coefficient >0.78 in these points. Conclusion: Essential characteristics of the IRM are the preservation of heartbeat morphology by retaining the low-frequency content and the preservation of inter-beat variation. The robustness of IRM to the changes in the EMG noise level and the low computational cost make it a good candidate for direct application in mobile ECG devices.