Quantitative analysis of gas mixtures for detecting low concentrations of gases like acetone or ethanol in exhaled breath is of high interest in medical diagnostics and patient monitoring. In this purpose, we intend to use source separation methods for separating the gas sources measured with MOX sensors. These unsupervised methods require a relevant mathematical model of responses of MOX sensors to gas mixtures, but only a few calibration measurements for recovering concentrations. Source separation methods also require a form of diversity, usually obtained by using a set of sensors. In this paper, we show that this diversity can be virtually obtained by using a unique MOX sensor, used sequentially at different temperatures. For a proof of concept, we propose an experimental investigation of metal-oxide sensor response model in a dual temperature mode for quantitative analysis of a mixture of acetone and ethanol in air. © 2017 IEEE.