Fetal electrocardiogram modeling using hybrid evolutionary firefly algorithm and extreme learning machine

Extraction of fetal electrocardiogram (FECG) from the abdominal region of the mother's skin is challenge task due to the high overlapping of maternal and fetal signals in this area. To overcome the problem, this paper proposes the utilization of extreme learning model (ELM) as the prediction algorithm to train on the FECG signal extracted by least mean square approach from the input abdominal and thoracic signals. The trained ELM model is used to model the FECG signal for the testing samples. Also, this paper investigates the firefly algorithm (FA) to tune the parameters of ELM and improve its performance. Due to the high complexity and too many parameters of FA, this paper embeds the evolutionary scheme into the FA algorithm which benefits from adaptive crossover and mutation probabilities. Evaluation of the proposed method on both synthetic and actual datasets proves its qualification in FECG extraction and modeling.