Background: Cardiotocography and obstetric ultrasound imaging are the standard for fetal monitoring during pregnancy and labor. These technologies are often expensive and, with very few exceptions, can only be used by highly trained personnel. Medical care during gestation differs in low- to middle-income countries (LMIC) from high-income countries. Our research has previously demonstrated that a low-cost 1D Doppler ultrasound (DUS) can be used during pregnancy to assess maternal and fetal health. However, differentiation between DUS signals from the fetal heart (FH) and umbilical cord (UC) can be challenging for untrained users. Methods: We trained a random forest classifier to detect whether 1D DUS recordings originated from FH activity or UC blood flow. This classifier was trained using the relative energy in each 10 Hz interval of the power spectrum derived from a balanced set of recordings. We used leave-one-out cross-validation to test our results. Results: We achieved an area under the curve of 0.93 and an accuracy of 82.6% for identifying FH activity, and 84.7% for UC blood flow. Conclusions: It is possible to differentiate the source of the 1D Doppler ultrasound signal. Depending on the source, different clinical parameters can be analyzed, enabling more targeted assessments of maternal and fetal health.