The　non-invasive　fet　al　Electrocardiogram　(FECG)　can　provide　reliable　information　about　the　fetus＇state.　The　original　FECG　signal　is　nevertheless　very　complex　and　severely　contaminated　by　external　disturbances　or　noises.　It　is　very　hard,　even　not　impossible,　to　reliably　extract　the　FECG　from　the　abdominal　signal　using　traditional　techniques.　In　this　work　the　recursive　least　squares　(RLS)　based　adaptive　noise　canceling　(ANC)　approach　is　applied　to　eliminate　the　maternal　ECG　(MECG)　and　hence　to　extract　the　FECG.　The　experimental　results　have　demonstrated　that　the　developed　ANC　approach　can　speed　up　convergence　of　the　normalized　least　mean　squares　(NLMS)　algorithm　and　is　able　to　track　non-stationary　FECG　signal　in　an　adaptive　manner.