Accurate single-trial detection of movement intention made possible using adaptive wavelet transform

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) require accurate real-time motor movement detection.
• Electroencephalography (EEG) signals suffer from low signal-to-noise ratio (SNR) and artifact ambiguity, complicating detection.
• Event-related-desynchronization (ERD) patterns are key indicators of motor activity.

Purpose of the Study:

• To develop a novel algorithm for precise single-trial detection of motor movement onset.
• To enhance BCI performance by improving the accuracy and temporal precision of ERD detection.
• To address challenges posed by low SNR and artifacts in EEG signals.

Main Methods:

• Utilized an adaptive matched filter with an optimized continuous Wavelet Transform for single-trial ERD detection.
• Implemented a maximum-likelihood (ML) electrooculography (EOG) artifact removal method.
• Applied the technique to a dataset of 6 healthy subjects recorded using Emotiv®.

Main Results:

• Achieved an average detection selectivity of 85 ± 6% and sensitivity of 88 ± 7.7%.
• Demonstrated high temporal precision in onset detection, ranging from -1250 to 367 ms.
• Significantly improved detection performance through effective artifact removal.

Conclusions:

• The novel algorithm enables precise detection of single-trial ERDs for motor movement onset.
• The combination of optimized Wavelet Transform and ML-based EOG artifact removal enhances BCI system performance.
• This approach offers a promising solution for reliable motor imagery detection in BCI applications.