Open AccessData processing from electrical signals acquired by an E-nose system used for quality control of cocoa
Author(s) -
Ana María Álvarez Florez,
Cristhian Manuel Durán Acevedo,
Jose Luis Asesor Reyes Carrillo
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1704/1/012013
Subject(s) - electronic nose , data acquisition , normalization (sociology) , arduino , principal component analysis , python (programming language) , computer science , data processing , graphical user interface , signal processing , computer hardware , analog signal , pattern recognition (psychology) , artificial intelligence , digital signal processing , embedded system , database , sociology , anthropology , programming language , operating system
This research consists of the implementation of different pattern recognition methods applied to discriminate and classify the physical signals (electrical) acquired by an artificial electronic nose system composed of a gas sensor array of 10 units coupled with a data acquisition board that were used to perform a sensor chamber that receives the electrical information from volatile organic compounds generated by cacao beans. A concentration chamber for samples conditioning of fermented beans was used and the samples were fermented around 72 and 144 hours while the cacao samples infected with monilia were over-fermented. For obtaining the temperature inside of the sensor chamber, a digital temperature sensor was implemented by using a Peltier Cell mechanism which was controlled through a classical algorithm. At the design stage, a data acquisition system composed of an Arduino card and a graphical interface made in LabView was developed for data storing, controlling, and signals monitoring. For the electrical signals treatment and data analysis, two pattern recognition models were applied by using Python software where two signals pre-processing methods such as Euclidean normalization and Roboust Scaling were used afterward with data processing techniques as principal component analysis and clusters analysis, obtaining a 96.51% of variance in the two first components.