ULTRA WIDE-BAND RADAR FOR THE REAL-TIME MONITORING OF HEART RATE USING CONVOLUTIONAL NEURAL NETWORK

Abstract

Ultra-wideband (UWB) radars are getting much attention for maritime applications of smart and luxury ships in which UWB radar could be integrated into Bridge Navigational Watch & Alarm System - BNWAS. One of the interesting applications of UWB radar is vital signs measurement, which is a contactless method. UWB radar measures respiration and heartbeat rate by the motion of thorax for detecting and checking the state of people on the bridge. However, the motion of the thorax caused by the heartbeat is usually low intensity and easily gets noisy and perturbed by a non-stationary signal. Due to this, an architecture built by a convolutional neural network is developed and modified to monitor heart rate using a contactless ultra wide-band (UWB) radar. The preprocessing part including many steps is necessary to clean raw signals from UWB radar. In this study, the evaluation metrics included a root mean square error of 11.34, a mean absolute error of 8.98, a standard deviation of the estimated signal of 4.05, and a percentage error of average HR at 5.77%. The proposed model could capture HR and is expected to be used for monitoring health and psychological status.