Contactless and continuous blood pressure measurement with wrist pulse signals obtained from millimeter wave frequency-modulated continuous wave radar

Abstract Blood pressure (BP) measurement is essential for assessing the health status of the cardiovascular system. Compared with the invasive approach of inserting a catheter into an artery, non-invasive solutions usually rely on wearable devices, which may be frustrated due to discomfort and inconvenience. In this paper we develop a novel continuous and contactless BP measurement system (CCBP) to measure BP according from the reflective pulse transit time (RPTT), which is acquired from a wrist pulse signal with a single-frequency-modulated continuous-wave millimeter-wave radar. CCBP consists of three functional modules. In the pulse motion extraction module, range-gating and beamforming techniques are utilized to isolate the signal from surrounding clutter and the direction of the reflection coming from pulse motion is identified to extract the phase signal with a high signal-to-noise ratio. In the pulse signal translation module, a convolutional neural network (CNN)-based translation filter is devised to transform the extracted phase signal to a fine-grained pulse signal. To prevent overfitting, three data augmentation approaches are designed to train the CNN-based filters, including overlapping, stretching and noising. In the BP estimation module, the RPTT is calculated from the pulse signal to estimate BP: the systolic BP (SBP) and diastolic BP (DBP) can be calculated from the RPTT using the BP estimation algorithm. Results of the experiment satisfy both the British Hypertension Society and the Association for the Advancement of Medical Instrumentation BP measurement standards, with 67.7% of errors being less than 5 mmHg and −1.30 ± 6.17 mmHg (mean error ± standard deviation) accuracy for SBP, and 66.2% and −3.10 ± 4.93 mmHg for DBP.