[B-19] Healthcare and Medical Information Communication Technology

10:00 AM - 10:15 AM JST(1:00 AM - 1:15 AM UTC)Candidate for Academic Incentive AwardCareer Explorer

[B-19-02] Non-Contact Blood Pressure Estimation Based on PPG Signal Generation
from Mm-wave Sensor Signals Using CycleGAN: Accuracy Improvement via
Optimization of Features and Training Data

◎△Kohei Kikuchi¹, Takahiro Ishimoto², Takeshi Toda¹ (1. College of Science & Technology, NIHON UNIVERSITY, 2. Graduate School of Science & Technology, NIHON UNIVERSITY)

Keywords:

Blood Pressure,Photoplethysmogram,Deep Learning,Vital signs,Millimeter-wave sensor

In our previous work, we extracted pulse wave components from IF signals acquired via a millimeter-wave sensor, generated PPG signals through the adversarial learning of a generator and discriminator using a GAN (Generative Adversarial Network), and constructed a blood pressure estimation model based on features extracted from these signals. In the present study, we revisited the features and training data in an effort to improve accuracy.