Enhancing Long Range E-nose Based Pineapple Ripeness Classification through Odor Transfer and Wavelet Transform

Pineapple ripeness detection using an electronic nose (e-nose) is often constrained by the distance between the sensor and the object, which causes a decrease in classification accuracy at a greater distance. This study aims to analyze the effect of sensor distance (15 cm and 65 cm). The dataset used was collected from 105 ‘Honey’ pineapple samples, balanced across three classes (35 young, 35 half-ripe, 35 ripe). Time-series data were acquired using an array of nine MQ-type gas sensors. Initial results showed a significant decrease in accuracy at a distance of 65 cm. To overcome this, the study proposes a physics-based signal compensation method called Odor Transfer, which uses a physics-inspired heuristic approach based on the Inverse-Square Law to reconstruct the aroma signal intensity at a long distance. Sensor data was processed using Discrete Wavelet Transform (DWT) for feature extraction and noise reduction, then classified using four algorithms: Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Naive Bayes (NB). Without Odor Transfer, the accuracy at 65 cm decreased significantly (SVM: 69.05%; KNN: 61.91%; RF: 64.29%; NB: 66.67%). After implementing Odor Transfer, all models showed an increase in accuracy, with the highest improvement seen in KNN (+9.52%). The main contribution of this research is the application of the Inverse-Square Law Heuristic-based Odor Transfer technique as a novel solution to the challenge of accuracy degradation due to distance in e-nose systems. Furthermore, the combination of the Odor Transfer method with DWT creates a more stable and accurate classification system, which is relevant for applications in non-destructive sensor systems in smart agriculture and post-harvest automation.