Detection and Classification of Fetal Heart Rate (FHR)

Fetal heart rate (FHR) is an important parameter for long-term prenatal monitoring of intrauterine fetal health. FHR, if measured correctly, can help reduce incidences of miscarriage and infant mortality and detect potential heart problems prior to delivery. We propose a novel technique to predict whether the fetal heart rate is normal/abnormal using raw audio signals acquired from an electronic stethoscope. They undergo adaptive bandpass filtering based on extracted continuous wavelet transform (CWT) coefficients. From the filtered signal, the fetal heart rate is detected using a Shannon energy envelope-based beat localization algorithm. The detected BPM and frequency-based features extracted from the signal are compiled and undergo data preprocessing techniques to generate a suitable dataset that trains a support vector machine (SVM) classifier that is capable of classifying any new data samples that are fed to the system. The proposed method presents good performance with an 84% recall and 100% precision.