Fractal Dimension as a discriminative feature for high accuracy classification in motor imagery EEG-based...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Brain-computer interfaces (BCIs) decode brain signals for device control, aiding individuals with motor impairments.
• Accurate classification of motor intent signals is crucial for effective BCI communication.
• Event-related desynchronization (ERD) is a known EEG feature for motor task discrimination.

Purpose of the Study:

• Introduce Fractal Dimension (FD) as a novel feature for subject-independent BCI event classification.
• Evaluate the efficacy of FD in machine learning models for motor imagery (MI) and motor execution (ME) tasks.
• Compare FD performance against traditional features like ERD.

Main Methods:

• Implemented machine learning models utilizing FD as a feature for EEG signal classification.
• Tested models on behavioral tasks involving both motor imagery and motor execution.
• Analyzed classification accuracy differences between unilateral and bilateral movements.

Main Results:

• FD significantly improved machine learning classification accuracy compared to ERD.
• Subject-independent classification performance was enhanced using FD.
• Unilateral hand movements demonstrated higher classification accuracy than bilateral movements in both MI and ME tasks.

Conclusions:

• FD shows significant potential as a discriminative feature for EEG signals in BCI applications.
• The findings contribute to improved subject-independent event classification in BCI systems.
• FD offers complementary information to frequency-based features for enhanced BCI performance.