Multi-channel Convolution Neural Network for Gas Mixture Classification

Mixed gas classification is a challenging problem because gas sensor arrays for mixed gases must process complex high-dimensional data. Additionally, it is expensive to obtain sufficient training datasets. To overcome these challenges, this paper proposes a novel method for mixed gas classification based on analog image representations with multiple sensor-specific channels and a convolutional neural network classifier. The proposed method maps a gas sensor array into a multi-channel image, adopts the CNN for feature extraction from images. The methodology is validated using a public gas sensor dataset. As a result, the proposed algorithm outperform the existing classification approaches in terms of the balanced accuracy score.