Data Processing for Multiple Electronic Noses Using Sensor Response Visualization

Traditional data processing methods for electronic noses (e-noses) generally rely on manually selected features with complex preprocessing steps, and it is difficult for traditional methods to deal with multiple e-noses' sensor data simultaneously. To overcome these shortcomings, a novel data processing method is proposed in this paper. According to the temporal correlations embedded in sensor responses, a visualizing approach encodes sensor responses as images to facilitate the subsequent feature learning. Considering the small sample size in e-noses, we design a small scale convolutional neural network model, which is capable of extracting useful features automatically rather than manually. The proposed method can be applied for multiple e-noses without data preprocessing and feature reduction. Experimental results on classification of six kinds of Chinese liquors show that the proposed method has a decent performance in data processing for multiple e-noses, and it also considerably outperforms traditional data processing methods for e-noses.