Sensor selection for tidal volume determination via regression – proof of methodology

Abstract

Measurement of respiratory volumes based on breath-related upper body movements continues to be a subject of interest in science and research. In general, smart garments are becoming more common in medical diagnostics and therapy monitoring, and improved, miniaturized and more accurate sensors are opening up new opportunities. A crucial issue in the development of smart clothing is how many sensors to use and where to place them in the clothes. Using data from a motion capture system, two different regression methods (Lasso and Ridge) were evaluated that can be used to select appropriate sensor subsets. The performance of the subsets, obtained by the regression methods, were compared with the best set of sensors obtained by analysing all possible subsets. The Lasso method showed clear performance advantages over Ridge regression in this field of application, but both methods can be employed as they significantly reduce time and computational effort.