Blood pressure prediction method based on multi-factor clue network

Abstract

The invention discloses a blood pressure prediction method based on a multi-factor clue network. The blood pressure prediction method includes the steps: applying personal basic information of a userand time sequence measurement data associated with blood pressure as cofactors to blood pressure prediction; form a blood pressure prediction channel and a time sequence prediction channel by means ofan existing dual-channel predictive LSTM model, wherein the dual channel works in the same way; predicting a blood pressure prediction value y1 by means of the recent blood pressure observation values through the blood pressure prediction channel through the applied multi-task learning; and predicting a time sequence data prediction value y2 through the time sequence measurement data associated with the blood pressure by means of the time sequence prediction channel. For the blood pressure prediction method based on a multi-factor clue network, the user can change the living habit through theprediction value, and can improve the blood pressure situation by means of certain means and measures so as to achieve the aim of health body.

Description

technical field [0001] The invention relates to a method for predicting blood pressure based on a multi-factor clue network. Background technique [0002] In blood pressure detection, researchers often use photoplethysmography (PPG) [12] , Pulse transit time (PTT) [13] , electrocardiogram (ECG) [14] , sphygmomanometer oscilloscope [15] Wait for these human physiological signals to estimate the real-time blood pressure value. The most commonly used model for this type of blood pressure estimation is the linear regression model [16] , support vector machine (SVM) [17] , Support Vector Regression (SVR) [18] , a recurrent neural network [19] , an improved Gaussian mixture regression (IGMR) method [20] , a multi-model mixture [21，22] etc. Literature [12] developed a high-precision model for estimating continuous arterial blood pressure based on recurrent neural network (LSTM) using only photoplethysmography (PPG). Literature [13] adopted a new data processing method Two...

AI Technical Summary

Problems solved by technology

The above blood pressure detection research uses oscillometric signals to estimate the blood pressure data, and uses the estimated results to evaluate the user's blood pressure status. When the user's blood pressure is in an abnormal state, the user is reminded and warned. This method requires equipment to It is impossible to predict blood pressure changes in advance, and it is very difficult to achieve timely and effective blood pressure early warning

Images