In the hospital environment, the variations in respiratory information (RI) are typically used for indicating the health condition of the patients. With RI known, corresponding parameters in both time domain (respiratory rate) and frequency domain (respiratory power) can be extracted to indicate the health condition. However, the monitoring technologies available in the clinical environment commonly work obtrusively, which can result in discomfort for patients during a long-term monitoring. Meanwhile, these technologies also possess
high false alarm rates, especially during movement of the subjects. Thus, designing an unobtrusive equipment suitable for moving patients that could collect as well as the algorithm to analyze such signals to provide
an accurate and robust estimate of respiratory information obviously becomes very interesting.
The base signals of interest for estimating respiratory information are two commonly used bio-medical signals, Impedance pneumography (IP) and electrocardiography (ECG). Impedance pneumography is a commonly
used method for respiratory rate monitoring. Electrocardiography is a measurement of the electrical activity of heart and respiration modulates the ECG in both time domain and frequency domain so that it is feasible to estimate the RI also from ECG. These two signals can be utilized with the same measurement electrodes attached to the surface of the body from the same device which provides a potential wearable method for real-time RI monitoring.
This thesis investigates the feasibility of using these two unobtrusive bio-medical signals, IP and ECG, to monitor respiratory rate (RR) in home environments especially during various movements. Algorithm development focused on creating accurate and robust respiratory rate estimates both from impedance pneumography and ECG-derived methods based on the modulation of respiration. Furthermore, the RI estimate from individual signals will be further fused by the way of information fusion and investigate the feasibility of giving more accurate respiratory information than that obtained using one signal alone. Finally, we also
investigate the feasibility of using acceleration recordings (Acc) to attenuate the motion artefact distorting the underlying signals.
Signal quality indicators (SQI) and respiration quality indicators (RQI) together play a key role in providing a more accurate RR estimate. Performance of all methods is evaluated against spirometer recordings, as a golden reference, based on a set of typical statisticalmetrics.
The main results are that IP signal can provide a more accurate RR estimate compared with the ECG signal. Furthermore, the derived signal quality indicators, SQI and RQI, could distinguish data of good quality from bad quality and the RR smoothing process could improve the result both of IP and ECG-derived methods
with the help of signal quality indicator. Another main finding is that the correlation between the motion artefacts and Acc recordings is not high enough, resulting in that Acc is not a proper reference signal for motion artefacts attenuation. Finally, it’s also found that information fusion methods could improve the overall estimate from multiple signal sources.
For the future work, a better reference signal for motion artefacts attenuation should be explored. Furthermore, more respiratory information, like tidal volume, should also be explored.