Multidimensional adaptive method for cancelling EMG signal from the ECG signal

Electrocardiographic signals (ECG) are corrupted with a certain number of noisy signals like baseline drift, mains interference (50 Hz or 60 Hz) due to electric sources, and electromyographic signals (EMG). This interference usually disturbs long periods of the useful signal and normally overlaps its spectrum. Till now, an adaptive approach remains difficult, because requires the acquisition of a noise reference channel (Xu, 1987; Thakor, 1991), or an "a priori" information based on detected QRS complexes and RR intervals (Provaznik 1993, 1994). Here, the authors show a new multidimensional adaptive method without an electromyographic noise reference, and without any "a priori" information. This technique was evaluated by using real cardiac signals recorded on three channels.