Gas Sensor Drift Compensation by an Optimal Linear Transformation

Drift compensation plays an important role in electronic nose systems. Traditional methods compensate drift often by using a reference gas. However, practically it is very expensive to use an extra reference gas. Motivated by the fact that the goal of drift compensation is to improve the classification performance, in this paper, we propose a novel linear approach for drift compensation, which incorporates drift compensation into the classification process. In our method, the data drift is modeled by a linear transformation and a k-nearest neighbor classifier is used on the compensated data. An optimal linear transformation will be obtained by optimizing the classification performance. Our main contributions are: 1) the extra reference gas is not needed any more except some labeled gas data since the data drift is compensated by the optimal linear transformation; 2) our approach integrates drift compensation into the classification task which can guarantee the classification performance when compensating the drift data. The results of our experiments show that our approach significantly outperforms other methods based on both synthesized and real data.