Skip to Main Content
This paper introduces an approach to extract an accurate fetal heart signal from noisy sensor measurements, which enables diagnosticians to make a reliable assessment of the fetus's health status. Therefore the primary objective in this work is to generate a fetal heart signal which reproduces all information required for analysis, while mitigating the noise effects. For this an Eigen filter based subspace separation technique is used for noise reduction in the sensor measurements, and an adaptive signal processing method based on the Wiener filter is used for fetal signal separation. The measured maternal abdomen signal contains various interfering signals in addition to the desired fetal signal. Among these interfering signals the most significant component is the maternal heart signal which is super imposed on the fetal heart signal. This obstacle was overcome in this research using a supervised adaptive signal processing method which utilizes a reference maternal heart signal measured from the mother's chest. The subspace separation technique we introduce in this work was applied prior to this to remove other external noises which corrupted the measurement signals. Furthermore, the noise reduction capability between the proposed subspace separation technique and standard IIR and FIR filters based methods used in practice were compared for various noise levels. This noise reduction enabled the Wiener filter to reduce the modeling mismatch, thereby improving the quality of the extracted signal as shown in the results of our work.