Fetal heart rate (FHR) is the main feature for monitoring fetal well-being as recommended by the International Federation of Gynecology and Obstetrics (FIGO). The ultrasound based cardiotocography (CTG) is nowadays the world-wide non-invasive clinical reference technique for intrapartum FHR monitoring, as stated by FIGO. However, some limitations of CTG have justified the interest for alternative solutions, based on abdominal electrocardiography (ECG). Their aim is to extract a fetal ECG signal (fECG), from which R peaks are detected to deduce FHR. Nevertheless, most published methods require a large number of abdominal sensors, which is not suitable for an expected use in real clinics. A novel methodology using a single abdominal ECG derivation is then investigated for FHR estimation in real clinical conditions. After maternal ECG attenuation, a source-filter model is considered to characterize the resulting fetal signal. Based on a non-negative matrix factorization of its spectrogram, the fundamental frequency of the source part is estimated, corresponding to FHR. Validation of the proposed FHR estimation is carried out on a real clinical database. Compared to a selection of other FHR estimation methodologies, our proposition presents the best performance in terms of reliability and similarity to the reference CTG. Evaluated for 8 subjects, corresponding to a total of 167 min of recordings, the mean ratio of outliers is lower than . These good results in real clinical conditions makes the proposed methodology a promising solution for use in routine practice by hospital departments.