Electronic Nose System implemented on ZYNQ Platform for Fruits Freshness Classification

The electronic nose (e-nose) is a novel bionic detection system that is widely used in the food safety industry. Currently, most e-nose systems deploy the recognition algorithm on a PC, which limits the portability of the e-nose. This work implements an electronic nose based on the ZYNQ7000 hardware platform for fruit freshness classification. Software simulations were performed before the hardware implementation. According to the response characteristics of the sensor array, a transient feature extraction method is proposed to reduce the time required for recognition. Meanwhile, the principal component analysis-kernel fisher discriminant analysis (PCA-KFDA) model is proposed to reduce the dimensionality of the extracted features. Then, A combination of three-point descent and the Mann-Kendall trend test was designed to enable the hardware circuit to automatically detect the response onset point. The results show that based on the support vector machine classification algorithm, the PCA-KFDA reduction model has higher classification accuracy than the traditional principal component analysis-linear discriminant analysis model (PCA-LDA). Finally, we achieved 92.9% accuracy in fruit freshness on the ZYNQ7000 platform.