Non-Contact Heart Rate Measurement Based on Adaptive Notch Filter and Elimination of Respiration Harmonics

Abstract

Continuous-wave Doppler radar is an established method for non-contact heart rate estimation. The output signal of the Doppler radar includes the fundamental wave of a heartbeat, heartbeat harmonics, the fundamental wave of respiration, respiration harmonics, and artifacts due to body movement. The respiration harmonics and artifacts degrade the accuracy of heart rate estimation. To this end, a novel heart rate estimation method is proposed that uses adaptive notch filters (ANFs). First, the proposed system estimates respiration frequency. Next, cascaded notch filters, the notch frequencies of which are controlled by the estimated respiration frequency, eliminate respiration harmonics; thus, the accuracy of heart rate estimation improves. In addition, the proposed method uses an ANF based on all-pass filter and an adaptive algorithm to estimate heart rate. The ANF is robust against disturbances, allowing the proposed method to improve the estimation accuracy in the presence of moving artifacts. The simulation results for human subjects show that MAPE, MAE, MSE, and RMSE of the proposed method are 5.24%, 4.00, 28.38, and 5.26, respectively; thus, they can be reduced up to fourth-order respiration harmonics. In an experiment on a dog not restrained in a cage, the proposed method, handling up to the fifth-order respiration harmonic, yields MAPE of 7.89%, MAE of 5.34, MSE of 96.11, and RMSE of 9.80. Thus, the proposed method shows better estimation accuracy than any other conventional method.