In the field of electrocardiography, accurate detection of fetal heart signals is vital during infancy which provides specific information to diagnose fetal arrhythmias. Noninvasive fetal electrocardiogram (FECG) method is mostly preferred to know about the status of fetus. However the FECG extraction method is till challenging for clinical technicians and engineers due to FECG taken from mother's abdomen has very low power relative to the maternal electrocardiogram (MECG). Complexity of extraction is increased by more sources of interferences among which the dominant source of interference is maternal component present in the abdomen which is nonlinearly transformed while travelling from chest to abdomen. In order to determine the nonlinearity of maternal component this study considered ANFIS (Adaptive Network based Fuzzy Inference System) algorithm. Then the maternal component is eliminated from the abdomen electrocardiogram (AECG), which can be further used to identify the arrhythmias easily, present in the FECG. The results demonstrate the effectiveness of the proposed technique which improves the quality signal-to-noise ratio (qSNR) more than double in all cases of arrhythmias as compared with LMS adaptive filter. © Copyright 2016 American Scientific Publishers All rights reserved.