Study on Cardiopulmonary Activity Monitoring Using Doppler Radar with Hardware Imperfection

By observing a person's chest wall movement using Doppler radar, we can remotely monitor cardiopulmonary activity. However, imperfect hardware may significantly distort the received signals, causing errors in the estimation of vital parameters like respiration rate and heart rate. To improve the estimation accuracy, we can estimate the hardware imperfection then calibrate the received signal using Gram-Schmidt process. A series of literatures have shown that many ellipse fitting algorithms perform well in estimating hardware imperfection for cardiopulmonary activity monitoring application. In this work, we consider algebraic distance based and geometric distance based ellipse fitting algorithms, examine their performance in estimating hardware imperfection parameters including DC-offsets, amplitude error, and phase error. We present numerical results that compare the vital parameter estimation accuracy using each ellipse fitting algorithm over a wide range of scenarios. These results suggest that the performance of ellipse fitting algorithms is very sensitive to SNR and chest wall movement amplitude, yet relatively insensitive to the composition of hardware imperfection unless it is very severe. Furthermore, while geometric distance based fitting is preferable when the SNR and chest wall movement are relatively large, it may be outperformed by algebraic distance based fitting in other cases.