Extraction of fetal heart activity by using bispectrum-based electromyography signal processing

This paper proposes the use of third-order spectrum named bispectrum that can be applied to biomedical diagnostics as a useful tool for detection of fetal heart activity. Fetal heart activity detection, recognition and analysis are proposed by extraction of informative features contained in bispectrum estimates computed for the electromyography signals measured on the abdominal surface of pregnant woman. Experimental investigations have been performed by using the recorded and accumulated abdominal electromyography signals. Contributions extracted in bispectral domain by computation of bimagnitude estimates demonstrate detection of novel class of information features indicating the phase coupling relationships between the frequencies contained in the abdominal electromyography signals. This phase coupling content can serve as discriminative feature for extraction fetal heart activity contribution in the intricate interference environment. Obtained results demonstrate effective interference suppression provided by using bispectrum-based signal processing.