Since the first observations of Cremer in 1906, fetal electrocardiogram (ECG) measurements via the maternal abdominal wall have remained a formidable challenge for clinical technicians and engineers in the cutting-edge field of information theory. Previous obstacles in extracting fetal ECG still complicate their acquisition at the present. These include three main difficulties for non-invasive measurement of fetal ECG: first, the low signal to noise ratio; second, the lack of a standard lead system for fetal ECG on the maternal abdomen; and third, the factor of fetal movement or non-stationarity during recording. A new extraction system based on blind source separation with reference signals (BSSR) was utilized and our detection rates, both off-line (91%) and on-line (60%), in pregnancies of 20 to 41 weeks of gestation have shown a marked improvement from earlier attempts. With this development, we discuss the potentials and limitations of this new system.