Passive Fetal Movement Signal Detection System Based On Intelligent Perception

The fetal movement signal is an important clinical indicator to assess the health of fetal growth and development in the uterus.The fetal movement with regular changes and normal count indicates that the placenta is functioning well,the fetus is getting enough oxygen in the uterus and the growth and development is sound.The frequency of fetal movements reaches a plateau at 32 weeks of gestation and remains at this level until birth.At this stage,two main active methods are used to assess fetal movement.One is the mother’s self-counting of fetal movements,which is often used for prenatal fetal health monitoring.Another is the Doppler ultrasound imaging technique used in hospital obstetrics and gynecology departments,which requires the involvement of highly qualified medical personnel,does not allow for long-term monitoring,and has a high economic cost.In recent years,a non-invasive intelligent sensing fetal movement detection system that can monitor high-risk pregnancies at home has received much attention in the field of wearable health monitoring.However,it is still challenging to recover the fetal movement signal in the continuous low-amplitude background contaminated by noise and detect real fetal movements.In response to these problems,this topic has been studied in the following areas.1)This paper uses flexible circuit board technology to design a wearable device based on the NRF52840 microprocessor as the core for acquisition fetal movement signals from acceleration sensors.Dual acceleration sensors are used to acquire the feature signals of fetal movement on the abdominal wall surface of pregnant women,and the data and recognition results are transmitted to the smartphone APP terminal for visual observation through the Bluetooth protocol of wireless communication integrated on the microprocessor chip.Realize the acquisition,detecting and storage of fetal movement signals.The proposed wearable fetal movement signal acquisition and detection system has the advantages of flexible and convenient operation,comfortable to wear and fast response.2)In this paper,we study the time domain features of the fetal movement,the noise and artifact interference signals in the original data.The Kalman filter is used to recover the fetal motion signal in a continuous low-noise background,the amplitude threshold is used to determine the artifact interference signal,the K-singular value decomposition algorithm is used to establish a complete dictionary of fetal motion features,the sparse identification algorithm and the adaptive filtering algorithm are used to identify the fetal motion signal,and the mask fusion algorithm is used to determine the final output.The experimental results show that the proposed two core recognition algorithms provide more satisfactory results for fetal movement signal detection,with advantages in positive prediction value and time complexity for each.3)In this paper,we study the nonlinear characteristics of fetal movement signal,and combine Kalman filter,time-frequency domain and wavelet domain feature extraction,Bayesian optimization ensemble machine learning Light GBM model with hyperparameters to achieve the prediction and identification of fetal movement.The accuracy,recall,precision,F-score and area under the ROC curve are evaluated and compared with existing preprocessing algorithms,feature extraction methods,hyperparameter optimization algorithms and fetal movement recognition methods,respectively.The results show that the method of this paper has improved in all evaluation metrics.