Improving pre-movement patterns detection with multi-dimensional EEG features for readiness potential decrease.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• The readiness potential (RP) is crucial for motor preparation in brain-computer interfaces (BCIs).
• A decrease in RP amplitude can severely impair pre-movement pattern detection accuracy.
• Existing methods struggle with detecting patterns under conditions of reduced RP.

Purpose of the Study:

• To enhance the accuracy of pre-movement pattern detection in BCIs, particularly when RP amplitude is decreased.
• To develop and validate a novel algorithm for improved BCI performance under challenging neural signal conditions.

Main Methods:

• Analysis of multi-dimensional electroencephalogram (EEG) features, including time-frequency, brain networks, and cross-frequency coupling (CFC).
• Proposal of a multi-dimensional EEG feature combination (MEFC) algorithm.
• Utilized features: RP waveforms, alpha/beta band energy and brain networks, and 2-10 Hz CFC values.
• Employed support vector machines for pattern recognition.

Main Results:

• The MEFC algorithm achieved an average recognition rate of 88.9% under normal conditions and 85.5% under RP decrease conditions.
• Compared to classical algorithms, MEFC demonstrated average accuracy improvements of 7.8% and 8.8% for the tasks.
• The proposed method effectively addresses the challenge of reduced RP amplitude in BCI applications.

Conclusions:

• The MEFC algorithm significantly improves the accuracy of pre-movement pattern decoding in BCIs.
• This approach offers a robust solution for BCI applications facing diminished readiness potential.
• The findings highlight the potential of multi-dimensional EEG feature analysis for advancing BCI technology.