A Novel Contactless Approach to Continuous Blood Pressure Monitoring Using Complex CEEMDAN and Neural Networks

Abstract—Contactless blood pressure (BP) monitoring is essential for daily healthcare. Typically, existing methods based on radar or remote photoplethysmography (rPPG) estimate only systolic blood pressure (SBP) and diastolic blood pressure (DBP). However, continuous monitoring of the complete blood pressure waveform provides deeper insights into cardiovascular health. This paper proposes a novel contactless continuous blood pressure monitoring algorithm termed CCNN-BP (Complex CEEMDAN and neural networks for blood pressure monitoring). Radar signals reflected from the chest are decomposed using Complex CEEMDAN (complete ensemble empirical mode decomposition with adaptive noise), to extract heartbeat-related intrinsic mode functions (IMFs). These IMFs are fed into a specialized neural network trained to estimate continuous blood pressure waveforms with accurate morphology. The final blood pressure waveform is obtained by adjusting the amplitude through calibration factors in combination with waveform characteristics. The proposed algorithm was trained and validated using an open-source dataset from Hamburg University of Technology. Experimental results show that the method achieves a mean absolute error (MAE) of 4.79 mmHg and a root mean square error (RMSE) of 7.13 mmHg for continuous blood pressure amplitude estimation, with a Temporal Pearson correlation coefficient of 0.86. For peak delay evaluation, the method attains aMAEof 25.98 ms andRMSEof 34.79 ms, while achieving a Spectral Pearson correlation coefficient of 0.94, indicating robust preservation of key frequency components. These results demonstrate the feasibility of CCNN-BP for accurate, continuous, and contactless blood pressure waveform monitoring.