Robust radar wrist vital signs estimation exploiting phase correlation characteristics

Abstract

In this paper, we propose a robust radar wrist vital signs estimation method to solve the wrist signal location and overcome the signal distortion caused by motion artifacts. The core idea lies in the selection and extraction of vital signs by exploiting the phase correlation characteristics. Specifically, we first utilize range fast Fourier transform and multi-angle signal gain to obtain phase signals, and then select potential vital sign’s signals by using the correlation spatial distributions of phase signals. After this, we extract the intrinsic vital sign’s signals from potential signals using the independent component analysis and the phase periodicity. We estimate pulse rate and respiration rate, and the root mean square error (RMSE) for pulse rate is only 2.98 beats per minute (bpm) with the mean absolute error (MAE) of 0.97 bpm, and the RMSE for respiration rate is only 2.88 bpm with the MAE of 1.00 bpm.