Quantitative analysis of multiple kinds of volatile organic compounds using hierarchical models with an electronic nose

This paper studies hierarchical discrimination and quantification models in order to simultaneously quantify multiple kinds of odors with an improved electronic nose. Such tasks are first regard as multiple discrimination tasks and then as multiple quantification tasks, and implemented by the hierarchical models with the divide-and-conquer strategy. The discrimination models are the common classifiers, including nearest neighbor classifiers, local Euclidean distance templates, local Mahalanobis distance templates, multi-layer perceptrons (MLPs), support vector machines (SVMs) with Gaussian or polynomial kernels. Similarly, the quantification models are multivariate linear regressions, partial least squares regressions, multivariate quadratic regressions, MLPs, SVMs. We developed several types of hierarchical model and compared their capabilities for quantifying 12 kinds of volatile organic compounds with the improved electronic nose. The experimental results show that the hierarchical model composed of multiple single-output MLPs followed by multiple single-output MLPs with local decomposition, virtual balance and local generalization techniques, has advantages over the others in the aspects of time complexity, structure complexity and generalization performance.