.png){kind=icon}
.png){kind=icon}
Return
Metal oxide semiconductor based electronic nose data pre-processing, review
.png){kind=icon}
PRE
AI
DOI：10.1016/j.engappai.2025.110540.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Metal Oxide Semiconductor (MOS) based Electronic Nose (E-nose) mimics Human Olfaction mechanisms in Odor Detection by creating a Unique Odor-Print that is processed by Pattern Recognition (PARC) algorithms. Odor Metal Oxide Semiconductor sensors do not behave as independent sensors. Each gas sensor in E-nose responds more selectively to a certain group of Volatile Organic Compounds (VOCs) but also shows a broad, overlapping response and sensitivity to the other Gas Compounds. Like in Natural Olfaction, a key role is played by Data Analysis. In this comprehensive literature review, the current state of different design stages in Metal Oxide Semiconductor (MOS)-based Electronic Nose Raw Data Pre-processing to condition input data prior to array processing and pattern recognition were thoroughly discussed. In depth technical application of various data preprocessing techniques to both static and dynamic sensor responses, emphasizing their importance in improving the sensors' performance and pattern recognition accuracy were provided. A use case has been attempted, discussing data samples, feature extraction processes, and dimensionality reduction. A sample MATLAB code was provided to show cluster formations. Transient feature extraction improved classification accuracy by up to 30 %, while piecemeal signal feature extraction proved highly effective. Multivariate analysis and chemometric methods reduced data dimensionality by approximately 50-80 %, for creating Parsimonious Odor Classification Models.
Keywords:
Metal oxide semiconductor
Electronic nose
Human olfaction
Pattern recognition
Transient feature extraction
Early-stage disease diagnosis
Journal
IF:
8
Papers:
5.4K
Citations:
3.5W