Non-contact Blood Pressure Prediction Employing Microwave Reflection based on Machine Learning

Hinako Ochi, Jiang Liu, Shigeru Shimamoto

Abstract: The typical measurement method using a cuff is physiologically stressful for many, especially the elderly people. In order to eliminate the burden and realize comfortable measurements, a non-contact method for monitoring blood pressure is necessary. This paper proposes a non-contact method to detect the human pulse, acceleration pulse waveform and in turn, predict blood pressure. In the experiment, microwave signals were transmitted against, and reflected from the chest and wrist, upon which the time-varying Insertion loss of Scattering parameter(S21) was acquired. Pulse, respiration rate and Augmentation Index(AIx) are first estimated via postprocessing of acquired raw data. Further signal processing also detects the acceleration pulse waveform. Blood pressure is then predicted via Machine Learning (ML) methods with parameter values derived from the detected pulse waveform, respiration waveform and acceleration waveform. Our works indicate that our proposed non-contact method is practical and has great potential for future smart health solutions.