Non-linear source separation under the Langmuir model for chemical sensors

Pierre Maho (1, 2) , Simon Barthelme (2) , Pierre Comon (1)

1 GIPSA-CICS - GIPSA - Communication Information and Complex Systems
2 GIPSA-VIBS - GIPSA - Vision and Brain Signal Processing

Electronic nose is a promising bio-inspired instrument for the detection of Volatil Organic Compounds (VOCs), meaning a compound containing carbon which easily evaporates. One of the most important parts of these devices is a set of non-specific chemical sensors, which will interact with the VOC and output valuable information for its identification. The non-specificity of these chemical sensors ensures the universality of the instrument. The main task achieved by this instrument is the detection of individual VOC. However, in many real-life applications, mixtures of VOCs are observed. The recovery of the mixture composition, meaning the individual signatures and their relative contribution, is a challenging task which can be studied in a Blind Source Separation framework. In this paper, we propose a non-linear mixture model for a particular type of chemical sensors. This model is based on the Langmuir isotherm for a multi-component gas. We study the joint identifiability of signatures and concentrations, and propose a necessary identification condition. Finally, we propose an algorithm for the blind estimation of the parameters and assess its performance through simulations.

Mots clés: gas mixture; non-linear source separation; electronic nose; Langmuir; two-block coordinate descent

Domaines: Traitement du signal et de l'image [eess.SP]

Pierre Maho, Simon Barthelme, Pierre Comon. Non-linear source separation under the Langmuir model for chemical sensors. SAM 2018 - 10th IEEE Workshop on Sensor Array and Multichannel Signal Processing, Jul 2018, Sheffield, United Kingdom. pp.380-384, ⟨10.1109/SAM.2018.8448636⟩. ⟨hal-01802358⟩