A new LeNet-5 gas identification convolutional neural network structure for electronic noses is proposed and developed in this paper. Inspired by the tremendous achievements made by convolutional neural networks in the field of computer vision, the LeNet-5 was adopted and improved for a 12-sensor array based electronic nose system. Response data of the electronic nose to different concentrations of CO, CH₄ and their mixtures were acquired by an automated gas distribution and test system. By adjusting the parameters of the CNN structure, the gas LeNet-5 was improved to recognize the three categories of CO, CH₄ and their mixtures omitting the concentration influences. The final gas identification accuracy rate reached 98.67% with the unused data as test set by the improved gas LeNet-5. Comparison with results of Multiple Layer Perceptron neural networks and Probabilistic Neural Network verifies the improvement of recognition rate while with the same level of time cost, which proved the effectiveness of the proposed approach.

The evaluation of the upkeep and freshness of aquatic products within the cold chain is crucial due to their perishable nature, which can significantly impact both quality and safety. Conventional methods for assessing freshness in the cold chain have inherent limitations regarding specificity and accuracy, often requiring substantial time and effort. Recently, advanced sensor technologies have been developed for freshness assessment, enabling real-time and non-invasive monitoring via the detection of volatile organic compounds, biochemical markers, and physical properties. The integration of sensor technologies into cold chain logistics enhances the ability to maintain the quality and safety of aquatic products. This review examines the advancements made in multifunctional sensor devices for the freshness assessment of aquatic products in cold chain logistics, as well as the application of pattern recognition algorithms for identification and classification. It begins by outlining the categories of freshness criteria, followed by an exploration of the development of four key sensor devices: electronic noses, electronic tongues, biosensors, and flexible sensors. Furthermore, the review discusses the implementation of advanced pattern recognition algorithms in sensor devices for freshness detection and evaluation. It highlights the current status and future potential of sensor technologies for aquatic products within the cold chain, while also addressing the significant challenges that remain to be overcome.

Discriminative dictionary learning has been successfully applied in pattern recognition field. In most of dictionary learning methods, ℓ ₀ -norm or ℓ ₁ -norm is used to regularize the sparse representation coefficients, which makes the computing time consuming. In this paper, we present a novel dictionary learning method to improve the gas identification performance of the electronic nose. It has significantly less complexity but leads to very competitive classification results. An analysis dictionary is trained to generate discriminative code by a simple linear projection, while a synthesis dictionary is trained to obtain discriminative reconstruction. Moreover, class label information is utilized to promote the discriminative power of the coding coefficients. The analysis dictionary and synthesis dictionary are trained jointly by an iterative method, which makes the learned projection dictionaries better fit with each other so that the more effective gas identification can be obtained. The proposed algorithm is evaluated on the analysis of different concentration of carbon monoxide, methane, hydrogen, benzene, formaldehyde, ethylene, propane, and ethanol. Experimental results show that the proposed method is not only effective in the signal analysis, but also useful and applicable to the performance enhancement of the current electronic noses.

The quality of gas sensor array is directly related to the performance of the electronic nose, which makes the optimization of sensor array a key issue in the study of electronic noses. A new sensor array optimization method is proposed based on Random Forest by using the Gini importance as the new measure of sensor contributions. An optimal sensor array of two sensors is built up targeting to classify CO, CH4 and their mixtures from an initial array composed of six sensors based on the method. Recognition results with the selected and other sensors by Random Forest, Back Propagation Neural Network and Support Vector Machine prove the effectiveness of the proposed array optimization algorithm.