Development of an Integrated E-Nose System with Controlled Sample Heating Chamber for Coffee Aroma Discrimination

Abstract

Coffee aroma classification is critical for quality assurance in the coffee industry, yet conventional sensory and analytical approaches remain costly, labor-intensive, and often subjective. This study introduces a low-cost electronic nose (E-nose) system, integrating a nine-MQ sensor array and a precision-controlled sample heating chamber, to provide objective, reproducible, and efficient evaluation of coffee aroma. Forty Arabica and forty Robusta coffee samples were analyzed at three controlled temperatures (35°C, 40°C, and 45°C). Raw sensor signals underwent baseline correction and Savitzky-Golay filtering before extraction of statistical features (mean, maximum, and area under the curve) from each sensor. Feature analysis using Principal Component Analysis (PCA) demonstrated enhanced class separability at higher temperatures. Classification models were developed using both Linear SVM and SVM with an RBF kernel, with hyperparameter optimization via grid search and evaluation by leave-one-out validation. The SVM-RBF model achieved the highest accuracy and ROC AUC at 45°C, confirming that optimal sample temperature significantly improves aroma discrimination. Overall, the findings highlight the importance of integrated sample heating and machine learning in E-nose systems, offering an affordable and scalable approach for robust coffee aroma classification with strong potential for broader food quality applications.