Design of Electronic Nose Based on MOS Gas Sensors and Its Application in Juice Identification

Abstract

Due to its advantages of fast response, low cost, low power consumption, and easy integration, Metal Oxide Semiconductor (MOS) gas sensor is widely used in the electronic nose system (E-nose). However, the MOS sensor has cross-sensitivity to different gases, which can impair the performance of the E-nose. Another key factor affecting the E-nose performance is the extraction method of gas features. In order to overcome the above shortcomings, an E-nose system that can modulate the operating temperature of gas sensors during the gas detection was designed in this paper, and a new gas feature extraction algorithm named Boruta-Recursive Feature Elimination (Boruta-RFE) was proposed based on the designed system. In order to verify the effectiveness of the designed system and the gas feature extraction algorithm, they were applied to the identification of different categories of apple juice. The experimental results show that more gas features can be obtained by modulating the operating temperature of the gas sensors, and the Boruta-RFE algorithm can effectively reduce the dimensionality of the original gas feature dataset, and can quickly select the key gas features, so as to effectively improve the identification accuracy of the E-nose system.