A Fetal Heart Rate Signal Source Recognition Model Based on Fast Fourier Transform and Ensemble Learning

Abstract

Fetal heart rate monitoring is a necessary routine examination item in obstetric clinic, which has important significance in the health examination of the perinatal fetus. Accurate extraction of fetal heart rate is a key technology in electronic fetal monitoring technology. There are still some difficulties and challenges in extracting the fetal heart rate from the ultrasound Doppler signals. The ultrasound Doppler fetal monitoring probe is difficult to maintain in the correct position, therefore, the Doppler ultrasound signals obtained may be the abdominal aorta signals which will cause fetal heart rate extraction error. In this paper, a signal source recognition model based on fast Fourier transform(FFT) and ensemble learning for ultrasound Doppler signals source recognition is proposed. The spectral features of the signals are extracted by FFT, and the spectral features are used as the input of ensemble learning model to decide whether the mother’s abdominal aorta signals are detected. The experimental results show that the proposed model can achieve the best recognition effect with the rule that the signals are regarded as from abdominal aorta if more than 93% of the signals get the negative output by the model within the time window of more than 13 seconds.