Abstract
An artificial electronic nose (e-nose) is developed, that mimics the human olfactory system, as an alternative to the human nose. In this work, we aim to develop a mini prototype of e-nose and use it for the detection of various types of gases present in the atmosphere. We then use existing machine learning models to carry out the classification task. Our study shows that the proposed e-nose system can find its potential application in various fields such as medical health care to detect chemical gas leakage, industries to detect hazardous gases, a substitute to the human nose when people are suffering from anosmia disorder, etc. We use k-Nearest Neighbours (kNN), Support Vector Classifier (SVC), Linear Regression (LR), Decision Tree (DT) and Random Forest (RF) algorithms to test the classification accuracy. Through the experimentation results we found that random forest model performs better with 97.77% classification accuracy compared to other models such as kNN, SVC, logistic regression and decision tree, whose classification accuracy are 93.33%, 62.22%, 71.11%, and 91.11% respectively. In future, we intend to extend this pilot work to automate the entire task where detected gaseous information by the e-nose is sent directly to the user to its mobile phone via Internet, instantly in real time fashion. We also aim to study using Deep Learning Techniques.
