Discrimination Enhancement with Transient Feature Analysis of a Graphene Chemical Sensor

A graphene chemical sensor is subjected to a set of structurally and chemically similar hydrocarbon compounds consisting of toluene, o-xylene, p-xylene, and mesitylene. The fractional change in resistance of the sensor upon exposure to these compounds exhibits a similar response magnitude among compounds, whereas large variation is observed within repetitions for each compound, causing a response overlap. Therefore, traditional features depending on maximum response change will cause confusion during further discrimination and classification analysis. More robust features that are less sensitive to concentration, sampling, and drift variability would provide higher quality information. In this work, we have explored the advantage of using transient-based exponential fitting coefficients to enhance the discrimination of similar compounds. The advantages of such feature analysis to discriminate each compound is evaluated using principle component analysis (PCA). In addition, machine learning-based classification algorithms were used to compare the prediction accuracies when using fitting coefficients as features. The additional features greatly enhanced the discrimination between compounds while performing PCA and also improved the prediction accuracy by 34% when using linear discrimination analysis.