Odor filtering and sensing system based artificial nose for chemical vapor class recognition

Objective of present study is to design an efficient odor sensor array based artificial electronic nose (E-nose) system for the detection of volatile organic compounds (VOCs) in medical and biometric applications. The E-nose system is designed with suitable unification of adsorbent materials, microceramic heaters, metal oxide semiconductor (MOX) gas sensors and pattern recognition unit. Polymers and carbon molecular sieves are used as adsorbent materials over the micro-ceramic heater to design an efficient odor filtering unit before analyte exposure to the sensor array. Sensor array response is measured for 75 different concentrations (in between 3-500 parts per million (ppm)) of five target VOCs including acetone, benzene, ethanol, pentanal, and propenoic acid. The influence of separating adsorbent material on sensor resistance is studied with objective to search the optimal material for VOCs filtering. Polydimethylsiloxane (PDMS) results highest sensitivity while carboxen- 1021 and carboxen-1012 result the lowest sensitivity as adsorbent materials for selected target VOCs. Sensor resistance (Ra) due to adsorption of VOCs by separating adsorbent material could be extracted from the sensor resistance (Rd) due to solvent desorption from the same. Fusion of these two sensor resistances is studied for data mining with objective to get improved class identification of target VOCs. Principal component analysis (PCA) as feature extraction method in combination with support vector machine (SVM) based classification is used to design pattern recognition unit for the class recognition of VOCs. Analysis results improved clustering of target VOCs in principal component (PC) space and average class recognition rate 82.66-100% in 3-fold cross validation of SVM classifier.