Abstract : Atrial fibrillation dominant frequency (AFDF) has been demonstrated to provide useful information on the characteristics of atrial fibrillation. The present work points forward a new method for AFDF estimation directly from a single-lead ECG recording, exploiting the concept of compressed spectrum as spectral estimator. The main purpose is to verify if ventricular activity subtraction might be avoided when the detection of the AFDF is the main goal. The greatest advantage of the proposed method is that it does not require any detection of the QT interval, making it very low cost and simple. Comparison of the estimated AFDFs with those obtained from the analysis of the corresponding atrial signals extracted using a Bayesian spatio-temporal QRST-cancellation technique showed that the two estimates were very similar. We conclude that the estimation of the AFDF can be pursued directly from a single-lead ECG recording, without any kind of QRSTsubtraction or QT-detection.