Cardiotocography is the most commonly used noninvasive diagnostic technique that provides physicians information about fetal development (in particular about development of autonomous nervous system - ANS) and wellbeing. It allows the simultaneous recording of Fetal Heart Rate (FHR), by means of a Doppler probe, and Uterine Contractions (UC), by means of an indirect pressure transducer. Currently, in cardiotocographic devices, Doppler methodology involves autocorrelation techniques to recognize heart beats, so evaluation of inter-beats time-interval is very improved. However, recorded FHR signals may contain artifacts, because of the possible degradation, or even less, of the Doppler signal due to relative motion between probe and fetal heart, maternal movements, muscle contractions and other causes. Moreover, fetal cardiac arrhythmias can have an effect on FHR signals. These arrhythmias do not represent an expression of the physiological behavior of the ANS. Both, artifacts and cardiac arrhythmias represent outliers of the FHR signals, so they affect both time domain and time frequency signal analysis. Their detection and correction is therefore necessary before carrying on signal processing. In this work, an algorithm for detection and successive correction of outliers (signal artifacts and fetal cardiac arrhythmias) was developed and tested, both on simulated FHR series and real FHR series.