Transfer Learning Method for Cuffless Blood Pressure Estimation Based on Measured PPG Data

Abstract

In this paper, we focus on cuffless blood pressure (BP) estimation based on measured PPG data. First, we design a Convolutional Neural Networks and Gated Recurrent Unit (CNN-GRU) network model to estimate BP. Furthermore, a transfer learning scheme is proposed to improve the training efficiency of CNN-GRU model. In detailed, a base model trained on one source user's data is transferred to other target users by freezing parameters of partial layers. The results based on measured data show that the proposed method can save training times while achieving superior performance.