The aim of this study was to propose an electrocardiogram (ECG) de-noising framework based on ensemble empirical mode decomposition (EEMD) to eliminate electromyography (EMG) interference without signal distortion. ECG signals are easily corrupted by EMG, especially in Holter monitor recordings. The frequency component overlapping between EMG and ECG is a challenge in signal processing that remains to be solved. The aim of the present study, therefore, was to resolve ECG signals from recorded segments with EMG noise. Two units were put into our proposed framework; first, modified moving average filter for signal preprocessing to cancel baseline wandering, and second, EEMD to cancel EMG. In order to enhance the de-noising capability (such as signal distortion in traditional EEMD), we developed a novel EEMD signal reconstruction algorithm using a statistical ECG model. We tested the proposed framework using MIT-BIH database, artificial and single-lead recorded real-world noisy signals. Correlation coefficients and ECG morphological features were used to evaluate the performance of the proposed algorithm. Our results showed that the proposed de-noising algorithm successfully resolved ECG signals from baseline wandering and EMG interference without distorting the signal waveform.