Sensor Drift Compensation Based on the Improved LSTM and SVM Multi-Class Ensemble Learning Models.

Zhao Zhao,Meng Meng,Li Li,Han Han,Xiao Xiao,Yu Yu

doi:10.3390/s19183844

Abstract

Drift is an important issue that impairs the reliability of sensors, especially in gas sensors. The conventional method usually adopts the reference gas to compensate for the drift. However, its classification accuracy is not high. We propose a supervised learning algorithm that is based on multi-classifier integration for drift compensation in this paper, which incorporates drift compensation into the classification process, motivated by the fact that the goal of drift compensation is to improve the classification performance. In our method, with the obtained characteristics of sensors and the advantage of Support Vector Machine (SVM) in few-shot classification, the improved Long Shot Term Memory (LSTM) is integrated to build the multi-class classifier model. We tested the proposed approach on the publicly available time series dataset that was collected over three years by the metal-oxide gas sensors. The results clearly indicate the superiority of multiple classifier approach, which achieves higher classification accuracy as compared with different approaches during testing period with an ensemble of classifiers in the presence of sensor drift over time.

Highlights

In recent years, with the rapid development of the machine olfactory technology, the gas identification systems have been widely applied in many fields, such as food testing, medical diagnosis, and environmental monitoring [1,2,3]
Sensor drift implies the interference of some factors, such as the temperature of the surrounding environment, humidity, pressure, as well as the aging and poisoning effects of the sensor material, which results in the sensor input signal that is involved in the interference signals
With the objective to improve the calculation accuracy, we have developed a multi-classifier integrated with a new loss function and Support Vector Machine (SVM) for the base classifier Long Shot Term Memory (LSTM), which greatly combines the advantages of SVM for small samples with the advantages of LSTM in time series

Summary

Introduction

With the rapid development of the machine olfactory technology, the gas identification systems have been widely applied in many fields, such as food testing, medical diagnosis, and environmental monitoring [1,2,3]. In the gas identification systems, the gas sensors are often used as the core function for sensing, identifying, and measuring different gases. Zero drift means that the reference deviates from a fixed value due to the influence of the external environment when the input signal of the amplifying circuit is zero. Span drift refers to a change of the coefficient and conversion factor of the value amplifier with the changes of time and temperature. Sensor drift implies the interference of some factors, such as the temperature of the surrounding environment, humidity, pressure, as well as the aging and poisoning effects of the sensor material (including external pollution, irreversible combination), which results in the sensor input signal that is involved in the interference signals.

Objectives

Results

Conclusion