An Effective Approach to Handling Noise and Drift in Electronic Noses

Sensor drift and noise handling in electronic noses (E-noses) are two different challenging problems. Sensor noise is caused by many factors such as temperature, pressure, humidity and cross interference. Noise can occur at any time, producing irrelevant or meaningless data. On the other hand, drift is appeared as a long term signal variation caused by unknown dynamic physical and chemical complex processes. Because sensor drift is not purely deterministic, it is very hard, if not impossible, to distinguish it from noise and vice versa. With respect to this property of E-nose, we propose a new approach to handle noise and sensor drift simultaneously. Our approach is based on kernel Fuzzy C-Mean clustering and fuzzy SVM (K-FSVM). The proposed method is compared to other currently used approaches, SVM, F-FSVM and KNN. Experiments are conducted on publicly available datasets. As the experimental results demonstrate, the performance of our proposed K-FSVM is superior than all other baseline methods in handling sensor drift and noise.