Harmonic model-based feature representation for temperature-modulated electronic noses

Electronic noses that employ periodic temperature-modulated metal oxide semiconductor (MOS) gas sensors typically extract features for odor detection and identification in the transformed domain, using either the fast Fourier transform (FFT) or the discrete wavelet transform (DWT). However, these conventional methods exhibit limitations in effectively discerning between various volatile organic compounds (VOCs). To address this issue, a novel harmonic model-based feature extraction technique is proposed to enhance odor detection and identification through the exploitation of previously overlooked phase information combined with a phase unwrapping method. To assess the efficacy of our proposed approach, we collect two extensive datasets ¹ 1 The code of the proposed method is available on https://github.com/guolisuccess-web/HarmonicModelOdorDiscr , and the dataset is available on https://ieee-dataport.org/documents/odor-discrimination-temperature-modulated-mos-gas-sensors . comprising four distinct types of MOS gas sensors and six different VOCs with varying concentrations, utilizing our recently developed gas-sensing platform. Subsequently, a comparative analysis is conducted between our proposed features and other commonly employed ones using these datasets. The experimental results unequivocally showcase the superior performance of the proposed feature extraction method and demonstrate its promising potential for integration into versatile electronic nose systems, augmenting their capacity to detect and identify multiple VOCs across different concentration levels.