Discriminating
structurally similar volatile organic compounds
(VOCs) molecules, such as benzene, toluene, and three xylene isomers
(BTX), remains a significant challenge, especially, for metal oxide
semiconductor (MOS) sensors, in which selectivity is a long-standing
challenge. Recent progress indicates that temperature modulation of
a single MOS sensor offers a powerful route in extracting the features
of adsorbed gas analytes than conventional isothermal operation. Herein,
a rectangular heating waveform is applied on NiO-, WO3-,
and SnO2-based sensors to gradually activate the specific
gas/oxide interfacial redox reaction and generate rich (electrical)
features of adsorbed BTX molecules. Upon several signal preprocessing
steps, the intrinsic feature of BTX molecules can be extracted by
the linear discrimination analysis (LDA) or convolutional neural network
(CNN) analysis. The combination of three distinct MOS sensors noticeably
benefits the recognition accuracy (with a reduced number of training
iterations). Finally, a prototype of a smart BTX recognition system
(including sensing electronics, sensors, Wi-Fi module, UI, PC, etc.)
based on temperature modulation has been explored, which enables a
prompt, accurate, and stable identification of xylene isomers in the
ambient air background and raises the hope of innovating the future
advanced machine olfactory system.