Many research groups in academia and industry are focusing on the performance improvement of electronic nose (E-nose) systems mainly involving three optimizations, which are sensitive material selection and sensor array optimization, enhanced feature extraction methods and pattern recognition method selection. For a specific application, the feature extraction method is a basic part of these three optimizations and a key point in E-nose system performance improvement. The aim of a feature extraction method is to extract robust information from the sensor response with less redundancy to ensure the effectiveness of the subsequent pattern recognition algorithm. Many kinds of feature extraction methods have been used in E-nose applications, such as extraction from the original response curves, curve fitting parameters, transform domains, phase space (PS) and dynamic moments (DM), parallel factor analysis (PARAFAC), energy vector (EV), power density spectrum (PSD), window time slicing (WTS) and moving window time slicing (MWTS), moving window function capture (MWFC), etc. The object of this review is to provide a summary of the various feature extraction methods used in E-noses in recent years, as well as to give some suggestions and new inspiration to propose more effective feature extraction methods for the development of E-nose technology.

Electronic noses (E-noses) have been successfully applied in various fields. However, as a result of the inherent variability of chemical sensors, a signal processing algorithm well trained with the data from the existing domain often cannot be directly applied to the domain of interest. This severely limits the large-scale use of E-noses. In this paper, an unsupervised discriminative domain reconstruction based extreme learning machine (DDR-ELM) is proposed to compensate for such drifts/shifts and address the domain adaptation problem. Specifically, the method learns a domain-invariant space to minimize the distribution difference between different domains by discriminatively handling the different domain data using an extreme learning machine (ELM) framework. This method retains as many of the useful spatial characteristics of the source domain as possible and reduces the divergence between domains without any labeled target domain data. It avoids the cost and labor of obtaining access to the labels of data from the domain of interest. In addition, both the domain reconstruction and classification processes utilize the ELM, which is solved by pseudoinverse operations without error back-propagation iterations, consequently keeping computational complexity low. Experiments on different sensor datasets demonstrate that the proposed method is superior to several state-of-the-art drift/shift compensation methods not only in classification accuracy but also maintaining higher efficiency.