Efficient combination of DWT and ICA to localize and remove ECG from surface electromyography measurement

Interpretations of surface electromyography (sEMG) to study thoracic back muscles responses to spinal manipulation are strongly limited by electrocardiogram (ECG) artifacts. In order to reduce the deterioration of sEMG information, the present paper proposes an efficient ECG removal method based first on ECG localization and secondly on ECG cancellation only where cardiac pulses have been localized. The main contribution of the proposed method consists in employing a combination of discrete wavelet transforms (DWT) and independent component analysis (ICA) to detect ECG position even if sEMG is strongly overlapped with ECG. Simulations realized from real data demonstrate that the proposed structure is significantly better than others simulated methods to remove ECG in term of relative error and coherence.