[Olfactory electroencephalogram signal recognition based on wavelet energy moment]

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Olfactory dysfunction assessment relies on understanding the brain's odor recognition capabilities.
• Electroencephalogram (EEG) signals offer a potential avenue for objective olfactory evaluation.
• Novel signal features are needed to improve the accuracy of odor classification from EEG.

Purpose of the Study:

• To propose and evaluate the Wavelet Energy Moment (WEM) as a feature for olfactory EEG signal classification.
• To compare WEM performance against other established signal features for odor identification.
• To determine the optimal classifier and EEG frequency band for accurate odor classification.

Main Methods:

• Collected olfactory evoked EEG data from 13 distinct odors.
• Extracted Wavelet Energy Moment (WEM) features, alongside Power Spectrum Density (PSD), approximate entropy, sample entropy, and wavelet entropy.
• Employed k-nearest neighbor (k-NN), Support Vector Machine (SVM), Random Forest (RF), and Decision Tree classifiers for odor identification.

Main Results:

• WEM feature consistently outperformed other features across all tested classifiers.
• The k-NN classifier combined with WEM achieved the highest overall classification accuracy of 91.07%.
• Analysis of EEG frequency bands revealed superior performance using the gamma (γ) band, with WEM and k-NN reaching 93.89% accuracy.

Conclusions:

• Wavelet Energy Moment (WEM) is a highly effective feature for classifying odors from EEG signals.
• The combination of WEM, the gamma (γ) band, and the k-NN classifier provides a robust method for olfactory dysfunction assessment.
• This research offers a novel objective basis for evaluating olfactory function and insights into olfactory-induced emotions.