Study of a Signal Classification Method in Electronic Noses Based on Suprathreshold Stochastic Resonance

Abstract

The research on stochastic resonance (SR) in threshold systems has received much attention recently, for multithreshold networks, SR is also observed in suprathreshold system. Generally suprathreshold SR (SSR) has been shown to exist by the mutual information and input-output cross-correlation. In this project, a novel method of "maximum cross-correlation coefficient" based on SSR was proposed to identify five gases gathered by the electronic nose. In the experiment, six carbon nanotubes gas sensors were chosen to compose a sensor array of the electronic nose, which were all sensitive to formaldehyde, benzene, toluene, xylene and ammonia. The data gathered from the sensor array were passed through the SSR system, which was quantified by the cross-correlation coefficient. Form the SSR curves, "maximum cross-correlation coefficient" of different gas classes was found to be completely different, and the "maximum cross-correlation coefficient" was a constant for each gas. So it can be used to accurately represent the different classes of gases. Compared with the classified results of the BP(back propagation) network, the method of "maximum cross-correlation coefficient" based on SSR has high accuracy in identifying five kinds of gases. So the method of "maximum cross-correlation coefficient" can be used as a new signal classificaiton method.

One-Class Drift Compensation for an Electronic Nose

“…Accordingly, semisupervised learning [13,14] and active learning [15,16] methods have been introduced to use a relatively small size of full-category drift correction samples selected from massive unlabeled drift data. Once the size of labeled drift correction samples reduces to zero, that is, all drift correction samples become unlabeled data, some dimension reduction methods can be used as long as the drift disturbance is regarded as an abnormal component [17][18][19][20]. Moreover, domain adaptation has been utilized, projecting the drift data and initial training samples for a shorter distance.…”