Unveiling the Smell Inspector and Machine Learning Methods for Smell Recognition

This paper presents the implementation of an odor classifier that utilizes various machine learning algorithms, including MLP models, LSTM models, Random Forest and XGBoost. These algorithms are applied to a small training dataset obtained from the Smell Inspector sensor, developed by SmartNanotubes Technologies. Study focuses primarily on the classification of five distinct smell substances: air, chlorinated water, vinegar, rum, and coffee, into their respective classes. The best proposed approach achieves a maximum accuracy of 92 percent in this classification task. To further enhance the classification task, binary classifiers are specifically tested to distinguish between air and the remaining substances, representing normal versus abnormal smell conditions. The best binary classifier achieves an accuracy of 97 percent.