Modeling and Alignment Algorithms of Multiple Sensors for the Wearable Human Respiration Monitoring System

The health condition can be reflected by human respiration monitoring, which requires the cooperation of various sensors including flow and concentration sensors. Based on the signal characteristics of the respiratory process, several modeling methods were used to reduce measurement error and improve response speed. The autoregressive exogenous (ARX) model resulted in smoother data from the turbine flowmeter and enabled 96.2% of tests to have an error of less than 3%. Wiener filtering significantly reduced the response time of the gas concentration sensors. The response time was shortened from 140 to 100 ms for the CO2 sensor, and from 220 to 100 ms for the O2 sensor. The end-tidal gas concentration characteristics were used to perform an alignment criterion between the different sensors to calculate end-tidal oxygen (FETO2) and end-tidal carbon dioxide (FETCO2) in comparison to Cosmed K5, which shows clinically insignificant differences according to the Bland-Altman analysis. This article provides a comprehensive modeling approach for the breath-by-breath (BBB) respiratory measurement method, enhancing the system's performance through sensor modeling and sensor signal alignment. Results indicate potential for practical application and scalability, offering an effective reference for similar systems.