Integration of Bionic Olfactory Model With MEMS Sensor Array Enhances Odor Classification

This letter presents a solution that integrates a microelectromechanical systems sensor array with a bionic olfactory model (BOM) to simplify data processing and enhance odor classification accuracy. The integrated sensor array adopts a quadrilateral cantilever beam structure with four resistive sensors, each sputtered with a different sensitive material, including indium oxide ($\mathrm{In_{2}O_{3}}$) doped with Au, Ag, Pt, and Pd. The BOM consists of a bionic olfactory receptor layer and a bionic olfactory bulb layer, capable of encoding sensor signals and efficiently extracting odor features without manual feature engineering. This system focuses on the classification of food types based on odor characteristics. To verify the performance of the system, data collection and performance analysis were performed on seven kinds of fruits (apple, banana, orange, mango, strawberry, pear, kiwi). The proposed model can directly extract odor features from sensor signals without feature engineering. Compared with traditional method, the system achieves an improvement in classification accuracy from 78.1% to 91.9% when using the k-nearest neighbors classifier.