Portable E-nose for Enhanced Pizza Toppings Recognition using MQ Gas Sensors

In this paper, a portable electronic nose system was developed and evaluated for its performance in rating pizza toppings, as compared to subjective evaluation of quality. In this study, four pizza-topping types were prepared: P1: 100% minced beef with Edam cheese, P2: 50% minced beef and 50% minced Kadid (air-dried salted meat) with Edam cheese, P3: 100% minced Kadid with Edam cheese, and P4: 100% minced beef with parmesan cheese. Kadid was similar to plain meat with respect to perception and preference. The experiment was performed on 101 prepared pizza-topping samples. Our study objective was to differentiate between various pizza toppings using the developed portable E-nose. Additionally, we aimed to highlight the impact of lemon smell as an olfactory disturbance in this differentiation. For this purpose, several procedures for feature selection, machine learning techniques were evaluated. Firstly, a Principal Component Analysis (PCA) showed a modest grouping of pizza toppings except for P4 samples (based on parmesan cheese) which were more distinct from others. By applying One-way ANOVA feature selection before performing PCA, Cluster Analysis (CA) and Support Vector Machines (SVMs), a significant improvement was observed in the identification of the four pizza toppings. Finally, the results from CA reveal that the presence of an olfactory disturbance caused by lemon scent significantly alters the order in which toppings are identified by the portable E-nose, particularly affecting cheese recognition.