Enhanced FastICA Algorithm with Overrelaxation Techniques for Improved Biomedical Signal Processing

Abstract

With growing concerns about maternal and infant health, accurate and non-invasive monitoring of fetal heart rate (FHR) is critical for early detection of potential complications such as fetal distress and congenital heart defects. Current monitoring methods, including invasive scalp electrodes and ultrasound Doppler, face limitations related to signal contamination and maternal-fetal interference. This paper proposes an improved FastICA algorithm that introduces an overrelaxation factor to optimize the initial weight selection, addressing the sensitivity and slow convergence issues of the traditional FastICA. The proposed algorithm relaxes constraints on initial weights, reduces the number of iterations, and improves the signal-to-noise ratio, making it suitable for real-time FHR extraction. Experimental results demonstrate that the improved FastICA algorithm not only effectively separates clean FHR signals but also achieves balanced and efficient convergence, offering significant practical value for continuous fetal monitoring applications.