Sensor Evaluation in a Breath Analysis System

Breath analysis systems contain arrays of correlated chemical sensors. For such systems, sensor selection is needed. From the process of sensor selection, some insight behind the performance of different sensor arrays can be gotten. Thus, we can know more about the sensors, which could help us with the selection work in turn. In this paper, a breath analysis system for diabetes diagnosis with 16 sensors is described. Based on this system, several methods are proposed to evaluate the importance, unique discriminant information and redundancy of each sensor. They are based on the results of exhaustive sensor selection. These methods are made convenient to observe and draw intuitive conclusions. They are applied to the breath analysis system and some useful discoveries about the sensors in the system are made accordingly.