E-Nose System Based on Fourier Series for Gases Identification and Concentration Estimation From Food Spoilage

Abstract

This work presents an electronic nose (EN)-based gas identification and concentration estimation method for the detection of food spoilage. The response data of sensors were acquired through a commercial gas sensor array and data acquisition circuit board and transformed into pictures with the form of the Fourier series. A convolutional neural network (CNN) model was used to identify the pictures from the conversion of sensor data, thus achieving the purpose of identifying the gases (C 2 H 5 OH, NH 3 , and H 2 S). In order to solve the problem of sample imbalance and to improve the generalization performance of classification models, the synthetic minority oversampling technique (SMOTE) and dropout technique were employed. Fivefold cross-validation was used to evaluate the performance of the model, of which the gas identification accuracy rate reached 96.67%. Moreover, a gas concentration regression model with the advantages of simplicity and strong interpretability was further proposed to estimate the concentrations of C 2 H 5 OH, NH 3 , and H 2 S. The mean absolute errors and coefficients of determination for the concentration estimation of C 2 H 5 OH, NH 3 , and H 2 S are (3.71 ppm, 0.968), (0.50 ppm, 0.968), and (0.13 ppm, 0.99), respectively. Furthermore, we used our model to evaluate the freshness of kiwifruit, pork, and beef, and it showed satisfactory predictive performance. The method proposed in this work realizes high-precision detection of gases from food spoilage and has a good application prospect in the rapid judgment of food freshness on the EN system.

A food quality detection method based on electronic nose technology

“…Ren et al [26] proposed an E-nose system based on CNN, which can effectively extract the time-series features of odor information of food with different freshness and achieve the classification of the freshness of 20 food items. Luo et al [27] achieved the freshness assessment of kiwifruit, pork, and beef by using the CNN model for feature extraction and recognition of converted images from sensor data. This suggests that the use of CNN methods on E-nose systems has good application prospects for rapid determination of food freshness.…”

Deep Learning for Gas Sensing via Infrared Spectroscopy

“…The emergence of powerful deep learning pattern recognition in gas speciation applications, as explored here, is expected to benefit IR absorption gas sensing by replacing manual feature selection and spectral interpretation with automated processes [ 19 , 20 ]. Using these learning methods, sensors can learn on large quantities of spectral information (features and their shapes) in frequency bands that are arbitrarily selected and/or imposed by hardware or other constraints and make useful predictions on constituent species and their concentrations [ 17 , 18 , 21 , 22 , 23 , 24 , 25 ]. Automated feature extraction, based on compound-distinct spectral fingerprints, followed by deep neural network classification provides a model architecture that can automate gas sensing and supplant the limitations of expert humans in spectral feature selection and spectral interpretation.…”

Multichannel Hierarchical Analysis of Time-Resolved Hyperspectral Data for Advanced Colorimetric E-Nose

“…Advanced detection of biological and chemical substances by sensor systems is required for numerous applications including health care, environmental monitoring, and security. − Especially electronic nose (e-nose)-based pattern recognition methods have substantially increased the classification performance of gas sensors. − Among the various types of sensors, colorimetric sensors are promising for utilizing the e-nose mechanism in a variety of gas detection applications due to their simple array manufacturing method. − Colorimetric sensor array-based e-nose systems have demonstrated discrimination abilities for various volatile organic compounds (VOCs) in exhaled breath diagnostics. − However, most colorimetric e-nose experiments rely on conventional RGB sensors which cannot provide the full information on the dynamic spectral features induced by the target gases on the sensor arrays. Meanwhile, a hyperspectral imaging system can provide 3D information (2D spatial and 1D spectral data), preserving both full spectral information and the shape of the sample. , While the complete 3D hyperspectral data provides a tremendous advantage in feature recognition and classification, the increased dimensionality of this data introduces complexity in the analysis of colorimetric measurements with a simple algorithm. , The deep-learning (DL) approach enables the analysis of complex data in order to extract features, and, in particular, convolution neural networks (CNNs) have demonstrated outstanding abilities to capture spatial patterns and hierarchical features in visual data. , The utilization of convolution filters enables the model to learn while maintaining the association between neighboring pixels, thereby conserving spatial information and features .…”