A deep feature mining method of electronic nose sensor data for identifying beer olfactory information. (December 2019)

• Record Type: Journal Article
• Authors: Shi, Yan; Gong, Furong; Wang, Mingyang; Liu, Jingjing; Wu, Yinong; Men, Hong
• Is Part Of: Journal of food engineering. Volume 263(2019)
• Page Start: 437
• Page End: 445
• Publication Date: 2019-12
• Subjects: Electronic nose -- Feature mining -- Convolutional neural network -- Support vector machine -- Beer; Food industry and trade -- Periodicals; Food -- Analysis -- Periodicals; Aliments -- Industrie et commerce -- Périodiques; Aliments -- Analyse -- Périodiques; Aliments -- Recherche -- Périodiques; 664.005
• DOI: 10.1016/j.jfoodeng.2019.07.023
• Languages: English
• ISSNs: 0260-8774

Abstract

Abstract: In this work, a deep feature mining method for electronic nose (E-nose) sensor data based on the convolutional neural network (CNN) was proposed in combination with a support vector machine (SVM) to identify beer olfactory information. According to the characteristics of E-nose sensor data, the structure and parameters of the CNN was designed. By means of convolution and pooling operations, the beer olfaction features were extracted automatically. Meanwhile, the SVM replaced the full connection layer of the CNN to enhance the generalization ability of the model, and two important parameters affecting the classification performance of the SVM were optimized based on an improved particle swarm optimization (PSO). The results indicated that the CNN-SVM model achieved deep feature automatic extraction of beer olfactory information, and a good classification performance of 96.67% was obtained in the testing set. This study shows that the CNN-SVM can be used as an effective tool for high precision intelligent identification of beer olfactory information. Highlights: A deep feature mining method was proposed based on the CNN for E-nose sensor information. According to the characteristics of E-nose sensor data, the structure and parameters of CNN were designed. The CAAPSO was proposed to optimize the parameters affecting the classification performance for SVM. CNN and SVM were integrated to identify beer olfactory information.