Research On Ensemble Classification Algorithm For Gas Sensor Array

Owing to the cross-sensitivity of gas sensors, the single sensor cannot accomplish the qualitative classification and quantitative detection of multiple gases. Compared with single sensor, electronic nose(E-nose) has a better performance in accuracy and real time, as a result of that, E-nose has been widely applied in gas sense of various fields. The E-nose system at present consists of gas sensor array units and signal processing algorithms, so this article is intended to carry a research on signal processing algorithms, aiming at improving the accuracy of E-nose.The research of this article on signal processing algorithms focuses on signal preprocessing and classification algorithm two parts. Firstly, get the high dimensional information with kernel principal component analysis (KPCA), and remove redundant elements in accordance with orthogonal signal correction. Combining the above two algorithms, obtain kernel orthogonal signal correction algorithm, which can improve the classification performance in gas sense preprocessing. Secondly, make a study on the theories of different classification algorithms and come up with the classification ensemble algorithm for gas sensor array units to modify the classifier, in order to further enhance the sense accuracy.This article takes a multiple gas as the sample, whose components are 6 volatile organic gases and measurement period is 3 years. After processing the sample with the classification algorithm designed in this article and comparing the results with relevant papers, the accuracy of the new classification algorithm rises by an average of 20%, indicating the new classification algorithm has an improvement in gas sense.