Online Gas Diagnosis by a Capillary-attached Gas Sensor Coupled to a Pattern Recognition System

We have applied a pattern recognition technique to the transient responses of a Capillary-attached Gas Sensor (CGS) for the first time. Two different feature extraction techniques were utilized to extract the discriminant features from the recorded data. Feature extraction based on the first derivative of the responses demanded a nonlinear classifier; however, the use of principal component analysis for feature extraction made the classification possible by a linear classifier. A systematic assessment of the trained system revealed that it was able to identify an unknown gas and determine its concentration online during the response recording process.