Development of statistical regression models for ventilation estimation.

Estimation of ventilation volume from dimensional changes of the rib cage and abdomen is of interest to researchers interested in quantifying internal exposure to environmental pollutants in the atmosphere. In this paper, we present different statistical regression models for estimating ventilation volume during free-living activities. The movements of the rib cage and abdomen were measured by piezoelectric sensor belts. Multiple linear regression as the calibration method was applied. Five regression models with different combinations out of thirteen features were developed and the performance of these models was compared through experimental study of 11 subjects. The effect of training approaches - model trained for each subject and for all subjects, and the effect of time intervals for computing features were also investigated. The results indicate that Model 2, combining respiratory features and breathing frequency, with a longer time intervals will lead to a higher accuracy.