Signal Selection Technique based on Statistical Approach for Enhanced Detection of Single-trial Auditory Evoked...

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Volume conduction in the brain complicates the detection of low-amplitude evoked potentials in electroencephalography (EEG).
• Accurate identification of single-trial auditory evoked potentials (AEPs) is crucial for understanding brain responses to auditory stimuli.

Purpose of the Study:

• To develop and validate a statistical signal selection method for enhanced detection and classification of single-trial EEG AEPs.
• To improve the classification accuracy of AEPs elicited by a subject's own name audio stimulus compared to familiar names.

Main Methods:

• A statistical analysis-based signal selection stage was employed.
• A support vector machine (SVM) classifier was utilized for AEP classification.
• EEG signals from the Fp1 electrode of 24 subjects were analyzed to create classifier-dependent feature vectors.

Main Results:

• The proposed method successfully selected one-quarter of the AEP signals for analysis.
• A single-trial classification accuracy of 70.59% was achieved using the selected AEP signals.
• This study represents the first report on classifying single-trial AEPs evoked by own-name versus familiar-name audio stimuli.

Conclusions:

• The statistical signal selection method enhances the detection and classification of single-trial EEG AEPs.
• The SVM-based classifier, combined with signal selection, demonstrates significant potential for identifying specific auditory responses.
• This research provides a novel approach for analyzing brain responses to personalized auditory stimuli.