The main goal of this thesis is to improve the cardiovascular health monitoring by developing a novel model-based blind system identification approach. This research lies on the core idea that the central aortic blood pressure (BP) waveform can be estimated from as few as two non-invasive circulatory signals.

To achieve this goal, first, we formulated a physiological model for the class of multi-channel systems with non-invasive BP measurements and expressed it as a blind system identification problem. We verified this model for estimating the central blood pressure waveform from pulse volume records (PVR) signals from arm and leg, collected from 10 human subjects. The results showed that the proposed approach could estimate central aortic blood pressure waveform accurately. The average root-mean-squared error associated with the central aortic blood pressure waveform was 4.1 mmHg while the average errors associated with central aortic systolic and pulse pressures were 2.4 mmHg and 2.0 mmHg respectively.

Afterward, we compared this method with a population-based technique to calculate cardiovascular risk predictors. First, we used the same approach to estimate the central blood pressure waveform from two non-invasive peripheral waveforms and then, calculated cardiovascular risk predictors. Experimental results obtained from 164 human subjects with a wide blood pressure range showed that this approach could estimate cardiovascular risk predictors accurately. Further analysis showed that the suggested approach outperformed a generalized transfer function regardless of the degree of pulse pressure amplification, but especially in high and low amplification ranges.

Finally, a new closed-loop approach to input de-convolution in coprime multi-channel systems based on state estimation techniques is proposed. This approach is based on the idea that the unknown input signal in a multi-channel system may be regarded as a state variable to be estimated from multiple output signals of the system. The validity and potential of the approach were illustrated using the clinically significant case study of estimating central aortic BP waveform from two non-invasively peripheral arterial pulse waveforms. The proposed algorithm could reduce the root-mean-squared error associated with the central aortic blood pressure by up to 27.5% and 28.8% relative to two conventional central aortic blood pressure estimation techniques: open-loop inverse filtering and peripheral arterial pulse waveforms scaled to central aortic diastolic and mean pressures.