Blood Pressure Measurement Based On Face Video

Blood pressure measurement is used for monitoring cardiovascular health,it plays an important role in the prevention and treatment of hypertension.the PPG signal obtained from the surface of the human body can be used for blood pressure measurement.But the quality of the PPG signal Obtained from face videois not good.At present,blood pressure modeling is mainly based on the artificially designed features of the PPG signal.The established blood pressure models generally have the problems of poor model generalization ability and poor applicability in practical use.This article do some research in the method of using face video to detect blood pressure,the main work and innovations of this article include:This article do some research in the method of detecting PPG signal.The PPG signal obtained from the face multi-region acquisition is constructed into a Multiple Measurement Vector matrix,combined with the motion mask obtained from the face motion detection,and the joint sparse reconstruction method is adopted to complete the reconstruction of the PPG signal.Designed a blood pressure deep network based on attention.It use a bidirectional longshort-term memory neural network to catch the features of PPG signal automatically,and a self-attention mechanism is adopted.Integrate it with personal characteristics to achieve blood pressure measurement of the human body.Different from traditional blood pressure model,the model designed in this article does not require perform complex artificial feature.The accuracy of the blood pressure model designed in this article on the Facevideo PPG-BP test set is systolic blood pressure 7.56 ± 9.61 mm Hg,and diastolic blood pressure 5.34 ± 6.71 mm Hg.The accuracy achieved on the selected MIMIC1.0 data set is 4.54 ± 5.88 mm Hg,and diastolic blood pressure 3.36 ± 4.21 mm Hg,indicating that the blood pressure model designed in this article has a good effect on facial video blood pressure measurement,and it’s adaptability is good.