Ambulatory ECG recordings are frequently corrupted by artifacts caused by, e.g., muscle activity or moving electrodes, which complicates the analysis of f-waves and motivates signal quality assessment to improve the reliability of f-wave analysis. Although many methods have been developed for assessing the quality of ECG signals in general, no method deals specifically with f-waves. This study proposes a novel signal quality index (SQI), using a modelbased approach for assessment of f-wave signal quality. To evaluate the performance of the SQI, 189 5-s recordings of f-waves from AF patients are studied, as is the same number of recordings with motion artifacts and electrode movements taken from the MIT-BIH Noise Stress Test Database. The signal quality index is capable of discriminating between f-waves and noisy recordings with an accuracy of 98%. The results suggest that the proposed signal quality index correctly identifies noisy recordings, and can be used to improve the reliability of f-wave analysis.