Contactless breathing estimation using radars has been explored since the 1960s and an accurate system with the ability to continuously monitor the health of non-critical patients without obstructing their day to day lives could significantly improve their well being. The current state of the art in this area does not have the accuracy required to work in a real-world environment and many of the existing methods have been tested only under very controlled situations. Low performance of breathing estimation algorithms under different scenarios inspired us to improve breathing estimation algorithms and develop a system for automated analysis of large number of algorithms against data from the reference sensors. A novel accurate breathing rate estimation method and a system to use multiple algorithms on the same set of data in real-time and identify the best performing algorithm dynamically to report breathing rate have been proposed in this thesis. In addition, automated data-collection and processing frameworks were developed to collect a large amount of data and process them and generate reports automatically. The proposed system has been tested under multiple test-cases involving multiple subjects and the accuracy of both new and existing algorithms have been evaluated by comparing the results with reference data collected using a respiration belt. The mean absolute error rate of breathing rate estimation after conducting experiments for a total of 9 subjects was found to be 0.79 breaths per minute for the novel CEEMD based method presented in this thesis. The mean absolute error rate after applying the scoring algorithm to select the best performing algorithm is 0.78 breaths/minute.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/37049 |
| Date | January 2017 |
| Creators | Gunasekara, A. K. Isuru Udayanga W. |
| Contributors | Bolic, Miodrag |
| Publisher | Université d'Ottawa / University of Ottawa |
| Source Sets | Université d’Ottawa |
| Language | English |
| Detected Language | English |
| Type | Thesis |
Page generated in 0.0025 seconds