Bioradar Signal Processing Algorithms based on Machine Learning

Since human physiological signals (e.g., respiration and heart rate) generate very small Doppler shifts, they are highly susceptible to being masked by strong clutter backgrounds. The main research focus of this project is to accurately extract weak targets (i.e., vital signs) from complex environments. The signals received by BioRadar contain vital signs signals and clutter signals. Based on the analysis of the collected data, it is known that vital signals have low signal-to-noise ratio and are characterized by quasi-periodicity and multiple harmonics; whereas clutter can be manifested as Gaussian colored noise. For this reason, this study translates the bio-radar signal processing into the problems of harmonic modeling, clutter suppression, and signal-to-noise ratio enhancement in the presence of Gaussian colored noise. The research objective is to extract the weak vital sign signals of human body from the complex background by machine learning techniques. In order to effectively extract physiological signals such as respiration and heart rate, this paper employ an ultra-wideband bio-radar system and advanced signal processing algorithms, and the experimental results show that the respiratory signal of participant 1 at 0 s is 2.1 mV, the heartbeat signal is 1.0 mV, and the suppressed clutter signal is 0.1 mV, which effectively suppresses the clutter interference with high accuracy and reliability.