Drift compensation algorithm based on Time-Wasserstein dynamic distribution alignment

The electronic nose (E-nose) is mainly used to detect different types and concentrations of gases. At present, the average life of E-nose is relatively short, mainly due to the drift of the sensor resulting in a decrease in the effect. Therefore, it is the focus of research in this field to propose an effective solution to sensor drift. This paper proposes Time-Wasserstein dynamic distribution alignment (TWDDA) to solve drift compensation according to the domain adaptive thought. This method simultaneously adapts the dynamic distribution according to the data distribution and the collection time span. Finally, this paper uses the public data set provided by the University of California San Diego (UCSD) to conduct experiments. The experimental results show that TWDDA is superior to other comparison algorithms and achieves better results.