Enhancing Electronic Nose Performance Based on a Novel QPSO-KELM Model.

Area of Science:

• Biomedical Engineering
• Computational Intelligence
• Microbiology

Background:

• Accurate bacterial detection is crucial for wound management.
• Electronic nose (E-nose) technology offers a promising approach for non-invasive diagnostics.
• Existing classification methods may have limitations in complex bacterial identification.

Purpose of the Study:

• To propose a novel multi-class classification method, QPSO-KELM, for bacteria detection using E-nose data.
• To evaluate the performance of QPSO-KELM against established classification algorithms.
• To assess the impact of different optimization methods and kernel functions on KELM performance.

Main Methods:

• Extraction of time and frequency domain features from E-nose signals.
• Implementation and comparison of QPSO-KELM with LDA, QDA, ELM, KNN, and SVM.
• Evaluation of PSO, GA, and GS optimization algorithms for KELM.
• Testing of Gaussian, linear, polynomial, and wavelet kernel functions for KELM.

Main Results:

• QPSO-KELM demonstrated superior classification accuracy for four wound classes (uninfected, S. aureus, E. coli, P. aeruginosa).
• The proposed method outperformed traditional optimization techniques and other classification algorithms.
• QPSO-KELM showed robust performance on independent E-nose datasets.

Conclusions:

• QPSO-KELM is a highly effective and superior method for bacteria detection using E-nose technology.
• The developed model holds significant potential for various E-nose applications in diagnostics and research.
• This study highlights the advancement in computational intelligence for microbial detection.