A Supervised Learning Approach for Real Time Vital Sign Radar Harmonics Cancellation

摘要

Vital signs radar has proven to be an interesting and useful tool; however it is still limited by a few key problems. One of these is the generation of harmonics due to nonlinearities arising from the large signal amplitude of respiration when compared to that of heartbeat. As a result, harmonics arise in the spectrum which confound accurate measurement of either. The gamma filter is a supervised machine learning based approach that offers a calibration-free and computationally efficient solution for many nonlinear filtering applications. Here, it is demonstrated for the first time as a tool for real-time heart rate estimation using the baseband signal from a non-contact vital sign signal measured from a 5.8-GHz quadrature Doppler radar. Experimental results show that the proposed filter for removing respiration harmonics can accurately measure heart rate even if it is weak or overwhelmed by the respiratory movement.