Committee Machine with Over 95% Classification Accuracy for Combustible Gas Identification

Gas identification represents a big challenge for pattern recognition systems due to several particular problems such as non-selectivity and drift. This paper proposes a gas identification committee machine (CM), which combines various gas identification algorithms, to obtain a unified decision with improved accuracy. The CM combines 5 different classifiers: K nearest neighbors (KNN), multi-layer perceptron (MLP), radial basis function (RFB), Gaussian mixture model (GMM) and probabilistic PCA (PPCA). A data acquisition system using tin-oxide gas sensor array has been designed in order to create a real gas data set. The committee machine is implemented by assembling the outputs of these gas identification algorithms based on weighted combination rule. Experiments on real sensors' data proved the effectiveness of our system with an improved accuracy 95.9% over the individual classifiers.