Noncontact Accurate Measurement of Cardiopulmonary Activity Using a Compact Quadrature Doppler Radar Sensor

• Published in: IEEE transactions on biomedical engineering Vol. 61; no. 3; pp. 725 - 735
• Main Authors: Hu, Wei, Zhao, Zhangyan, Wang, Yunfeng, Zhang, Haiying, Lin, Fujiang
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.03.2014; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adult; Algorithms; Arctangent demodulation; Baseband; cardiopulmonary; Computer Simulation; Demodulation; Doppler Effect; Doppler radar; Electrocardiography; Female; Heart beat; Heart rate; Heart Rate - physiology; Heart rate variability; heart rate variability (HRV); Humans; Male; Monitoring, Physiologic - instrumentation; Monitoring, Physiologic - methods; noncontact; Radar; Receivers; Signal Processing, Computer-Assisted - instrumentation; Signal-To-Noise Ratio; Young Adult
• ISSN: 0018-9294; 1558-2531; 1558-2531
• DOI: 10.1109/TBME.2013.2288319

The designed sensor enables accurate reconstruction of chest-wall movement caused by cardiopulmonary activities, and the algorithm enables estimation of respiration, heartbeat rate, and some indicators of heart rate variability (HRV). In particular, quadrature receiver and arctangent demodulation with calibration are introduced for high linearity representation of chest displacement; 24-bit ADCs with oversampling are adopted for radar baseband acquisition to achieve a high signal resolution; continuous-wavelet filter and ensemble empirical mode decomposition (EEMD) based algorithm are applied for cardio/pulmonary signal recovery and separation so that accurate beat-to-beat interval can be acquired in time domain for HRV analysis. In addition, the wireless sensor is realized and integrated on a printed circuit board compactly. The developed sensor system is successfully tested on both simulated target and human subjects. In simulated target experiments, the baseband signal-to-noise ratio (SNR) is 73.27 dB, high enough for heartbeat detection. The demodulated signal has 0.35% mean squared error, indicating high demodulation linearity. In human subject experiments, the relative error of extracted beat-to-beat intervals ranges from 2.53% to 4.83% compared with electrocardiography (ECG) R-R peak intervals. The sensor provides an accurate analysis for heart rate with the accuracy of 100% for p = 2% and higher than 97% for p = 1%.