A portable electronic nose based on convolutional neural network for beer identification

Food quality monitoring plays a crucial role in safeguarding product integrity and ensuring the well-being of consumers, serving as a cornerstone for both public health and industry standards. As devices that detect food volatiles, electronic noses (E-noses) can be applied to monitor the odors of beers, helping to verify their authenticity. In this study, an E-nose with vertically circular sensors array and neural network was designed for beer identification. To overcome the limitations of conventional planar E-noses that rely on regular testing chambers, gas sensors are vertically arranged. Furthermore, a two-dimensional convolutional neural network based on channel attention and cosine annealing warm restarts, denoted as CC-2DCNN, was adopted as pattern recognition algorithm. The channel attention mechanism enhances the abilities of learning key features and weight differentiation, while the cosine annealing warm restarts strategy dynamically adjusts the learning rate. Rapid identification was also studied to accelerate response speed while maintaining accuracy for beer identification. Then, CC-2DCNN was successfully deployed into the E-nose, and the accuracy of 99.3 % indicating a novel and promising approach for food recognition.