Abstract : The dominant frequency (DF) of the atrial activity signal is arguably one of the most relevant features characterizing atrial fibrillation (AF), the most common cardiac ar-rhythmia. Its accurate estimation from noninvasive acquisition modalities such as the electrocardiogram (ECG) can avoid risks of potential complications to patients in a cost-effective manner. However, the approximation of the underlying intracardiac atrial activity by noninvasive techniques such as average beat subtraction or blind source separation has not always been satisfactory. In the present work, a new approach based on the ensemble empirical mode decomposition (EEMD) is proposed for AF DF estimation. Our results suggest that EEMD provides more accurate estimates of intracardiac AF DF than alternative noninvasive methods. In addition, the empirical nature of EEMD overcomes important drawbacks of other techniques, simplifying its implementation in automatic tools for diagnosis aid.