Radar-Based Contactless Blood Pressure Estimation System Using Signal Decomposition and Deep Neural Network

Noncontact blood pressure monitoring holds significant importance in terms of convenience, real-time monitoring, and long-term surveillance, and it plays a positive role in fields of blood pressure management and disease prevention. This article proposes a blood pressure monitoring method based on a novel signal processing method with frequency-modulated continuous wave (FMCW) radar. First, we employ the millimeter-wave radar to collect the phase signal of the neck. Then, we present an empirical mode decomposition method based on time-varying filters to decompose the phase signal and propose to select efficient intrinsic mode functions (IMFs) for blood pulse-wave signal reconstruction. Finally, this article proposes a dual-stream fusion neural network for pulse-wave feature extraction and blood pressure estimation. The validation of the proposed method’s feasibility and effectiveness underscores its comparative advantage over the most advanced techniques in the field, as evidenced by the results.