Analysis of Simple Algorithms for Motion Detection in Wearable Devices

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-Computer Interfaces (BCIs) leverage electroencephalogram (EEG) signals for specific applications.
• Imagery movement detection from EEG is a growing area of interest for non-invasive applications.
• Wearable devices require algorithms with minimal resource demands.

Purpose of the Study:

• To develop simple algorithms for detecting imagery upper-limb movement.
• To ensure algorithms are suitable for non-invasive, low-resource wearable devices.
• To identify optimal signal characteristics (electrodes, frequency bands) for imagery movement detection.

Main Methods:

• Developed two algorithms (FBA and BLA) based on signal correlation, wavelet energy per segment, and wavelet energy per electrode.
• Utilized a public EEG database from 105 subjects performing imagery upper-limb movements.
• Tested algorithm performance with varying numbers of electrodes and frequency bands.

Main Results:

• The best performance was achieved using 6 electrodes.
• The frequency range of 25 Hz - 30 Hz yielded the best results, outperforming the beta band in this study.
• Algorithm performance was found to be dependent on the number of electrodes and frequency band, but not always linearly.

Conclusions:

• The developed simple algorithms do not fully meet wearable requirements, indicating a need for adaptive approaches.
• Individual subject differences necessitate adaptive algorithms for robust imagery movement detection.
• Optimal information for imagery movement detection is not solely dependent on a high number of electrodes or specific frequency bands like beta.