Rapid Identification of Multiple Gases

Abstract

Rapid identification of low-leveled toxic and harmful gases is of a challenge in current environmental monitoring. In this paper, we combined convolutional neural networks and bidirectional long short-term memory neural network, and proposed a method for fast identifying gases existing in trace amount in the environment. The attention mechanism was introduced to extract the key features of the input, and the Bayesian optimization method was applied to optimize the hyper-parameters. In order to evaluate the proposed method, we ran experiments using the low-concentration gas sensing data employing several existing predictive methods and the proposed one, and eventually compared their performances with recall and F1-score metrics. The results demonstrate that the performance of the proposed method exceeds that of the other methods, and also gives better performance on classifying gas components, given the gas concentration is below 125 ppm and the response time is limited to 0.5s.