Drift Insensitive Features for Learning Artificial Olfaction in E-Nose System

Atiq ur Rehman, Amine Bermak

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Domain Features and Independence Maximization are proposed recently for learning domain invariant subspace to handle drift in gas sensors. Proposed domain features were acquisition time and a unique device label for collected gas samples. In real time applications of gas sensing, a sample is usually collected using a multi-sensor sensing approach, so a unique device label is not possible in that case, which results in performance degradation. Similarly, semi supervised learning algorithms are proposed to handle drift for gas sensing applications, but getting data from the target domain for calibration of the system is not always possible. To address these problems, this paper proposes a novel approach to handle drift in gas sensors, with following merits: 1) a new classification system based on cosine similarity is developed and features are exploited using a metaheuristic, the outcome is drift insensitive features which are capable of handling drift in gas sensors 2) proposed system is robust against drift without requiring any re-calibration, domain transformation or data from target domain 3) the classification system is an integration of two classifiers, this enables the system to outperform other base line methods 4) only median values of drift insensitive features are used for learning, so the system requires very few memory cells for storage. The proposed system is validated against a large-scale dataset of 13910 samples from six gases, with 36 months drift and has demonstrated 86.01% classification accuracy which is 2.76% improvement, when compared to other state-of-the-art methods.

Original languageEnglish
JournalIEEE Sensors Journal
DOIs
Publication statusAccepted/In press - 5 Jul 2018

Fingerprint

learning
Chemical sensors
Gases
Labels
gases
Calibration
sensors
Supervised learning
Learning algorithms
Classifiers
Data storage equipment
Degradation
Sensors
classifiers
acquisition
degradation
cells

Keywords

  • Classification algorithms
  • Cosine Similarity
  • Electronic Nose
  • Feature extraction
  • Gas detectors
  • Gas sensors
  • Gases
  • Particle Swarm Optimization
  • Sensor arrays
  • Sensor drift
  • Sensor systems

ASJC Scopus subject areas

  • Instrumentation
  • Electrical and Electronic Engineering

Cite this

Drift Insensitive Features for Learning Artificial Olfaction in E-Nose System. / Rehman, Atiq ur; Bermak, Amine.

In: IEEE Sensors Journal, 05.07.2018.

Research output: Contribution to journalArticle

@article{38c4b395968d48bab8142499df30444a,
title = "Drift Insensitive Features for Learning Artificial Olfaction in E-Nose System",
abstract = "Domain Features and Independence Maximization are proposed recently for learning domain invariant subspace to handle drift in gas sensors. Proposed domain features were acquisition time and a unique device label for collected gas samples. In real time applications of gas sensing, a sample is usually collected using a multi-sensor sensing approach, so a unique device label is not possible in that case, which results in performance degradation. Similarly, semi supervised learning algorithms are proposed to handle drift for gas sensing applications, but getting data from the target domain for calibration of the system is not always possible. To address these problems, this paper proposes a novel approach to handle drift in gas sensors, with following merits: 1) a new classification system based on cosine similarity is developed and features are exploited using a metaheuristic, the outcome is drift insensitive features which are capable of handling drift in gas sensors 2) proposed system is robust against drift without requiring any re-calibration, domain transformation or data from target domain 3) the classification system is an integration of two classifiers, this enables the system to outperform other base line methods 4) only median values of drift insensitive features are used for learning, so the system requires very few memory cells for storage. The proposed system is validated against a large-scale dataset of 13910 samples from six gases, with 36 months drift and has demonstrated 86.01{\%} classification accuracy which is 2.76{\%} improvement, when compared to other state-of-the-art methods.",
keywords = "Classification algorithms, Cosine Similarity, Electronic Nose, Feature extraction, Gas detectors, Gas sensors, Gases, Particle Swarm Optimization, Sensor arrays, Sensor drift, Sensor systems",
author = "Rehman, {Atiq ur} and Amine Bermak",
year = "2018",
month = "7",
day = "5",
doi = "10.1109/JSEN.2018.2853674",
language = "English",
journal = "IEEE Sensors Journal",
issn = "1530-437X",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - Drift Insensitive Features for Learning Artificial Olfaction in E-Nose System

AU - Rehman, Atiq ur

AU - Bermak, Amine

PY - 2018/7/5

Y1 - 2018/7/5

N2 - Domain Features and Independence Maximization are proposed recently for learning domain invariant subspace to handle drift in gas sensors. Proposed domain features were acquisition time and a unique device label for collected gas samples. In real time applications of gas sensing, a sample is usually collected using a multi-sensor sensing approach, so a unique device label is not possible in that case, which results in performance degradation. Similarly, semi supervised learning algorithms are proposed to handle drift for gas sensing applications, but getting data from the target domain for calibration of the system is not always possible. To address these problems, this paper proposes a novel approach to handle drift in gas sensors, with following merits: 1) a new classification system based on cosine similarity is developed and features are exploited using a metaheuristic, the outcome is drift insensitive features which are capable of handling drift in gas sensors 2) proposed system is robust against drift without requiring any re-calibration, domain transformation or data from target domain 3) the classification system is an integration of two classifiers, this enables the system to outperform other base line methods 4) only median values of drift insensitive features are used for learning, so the system requires very few memory cells for storage. The proposed system is validated against a large-scale dataset of 13910 samples from six gases, with 36 months drift and has demonstrated 86.01% classification accuracy which is 2.76% improvement, when compared to other state-of-the-art methods.

AB - Domain Features and Independence Maximization are proposed recently for learning domain invariant subspace to handle drift in gas sensors. Proposed domain features were acquisition time and a unique device label for collected gas samples. In real time applications of gas sensing, a sample is usually collected using a multi-sensor sensing approach, so a unique device label is not possible in that case, which results in performance degradation. Similarly, semi supervised learning algorithms are proposed to handle drift for gas sensing applications, but getting data from the target domain for calibration of the system is not always possible. To address these problems, this paper proposes a novel approach to handle drift in gas sensors, with following merits: 1) a new classification system based on cosine similarity is developed and features are exploited using a metaheuristic, the outcome is drift insensitive features which are capable of handling drift in gas sensors 2) proposed system is robust against drift without requiring any re-calibration, domain transformation or data from target domain 3) the classification system is an integration of two classifiers, this enables the system to outperform other base line methods 4) only median values of drift insensitive features are used for learning, so the system requires very few memory cells for storage. The proposed system is validated against a large-scale dataset of 13910 samples from six gases, with 36 months drift and has demonstrated 86.01% classification accuracy which is 2.76% improvement, when compared to other state-of-the-art methods.

KW - Classification algorithms

KW - Cosine Similarity

KW - Electronic Nose

KW - Feature extraction

KW - Gas detectors

KW - Gas sensors

KW - Gases

KW - Particle Swarm Optimization

KW - Sensor arrays

KW - Sensor drift

KW - Sensor systems

UR - http://www.scopus.com/inward/record.url?scp=85049675654&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85049675654&partnerID=8YFLogxK

U2 - 10.1109/JSEN.2018.2853674

DO - 10.1109/JSEN.2018.2853674

M3 - Article

JO - IEEE Sensors Journal

JF - IEEE Sensors Journal

SN - 1530-437X

ER -