Signal Quality Indices Evaluation for Optimized Radar-Based Blood Pressure Monitoring

This paper proposes a signal quality indices determination for blood pressure estimation based on pulse waveform detected by a SIL radar. 6 selected features are extracted from the pre-processed signal. After that, the proposed signal quality estimation analysis is conducted to assess whether reliable blood pressure can be obtained from the collected signal. Unlike traditional signal quality algorithms, this algorithm does not evaluate the signal as a whole and discards it if the results are poor. Each pulse is evaluated separately, and just the low-quality pulses are removed. The advantage of this method is that it reduces sampling time while improving blood pressure prediction accuracy. The algorithms are validated on manually labelled data. Compared with the reference blood pressure obtained by a sphygmomanometer, the model’s mean absolute error (MAE) performance can be decreased by 10% to 40%, proving that this method can effectively measure blood pressure using a single radar in a non-contact state.