While most medical applications of the EEG BMI require clean data, recorded artifacts from facial movements and eye-blinks have been shown to be exceedingly accurate in classifying individuals.
Historically, when a subject undergoing an EEG recording session blinks or twitches their face, a sharp spike appears on the recording trace that renders that epoch unusable. While most medical applications of the EEG BMI require clean data, recorded artifacts from facial movements and eye-blinks have been shown to be exceedingly accurate in classifying individuals.
In addition to facial movements, eye-blinks are now being integrated with EEG systems to achieve a higher classification rate than using brainwaves alone. In , it was discovered that using a cheap, consumer-version EEG headset (the Neurosky Mindwave) achieved a 99.4% classification rate by discriminating eye-blinks in addition to EEG brainwaves. This represents the highest classification rate for a significant subject size (31 persons) to date using an EEG instrument. Eyeblinks may have some disadvantages relative to brainwave patterns (such as forced-volition and lack of continuous verification), and are likely to be more prone to mimicry, but may be easier to train. Integrated with the continuous verification protocols offered by EEGs, eyeblinks may provide a high-level of security when requested in discretized intervals.
An exciting new modality analyzes differences in neural connectomes derived from functional magnetic resonance imaging (fMRI) to identify individuals. Unlike EEGs, fMRI relies on imaging deep neuronal structures, often in a clinical setting. The neural connectome is a diverse, complicated web of linkages among neurons inside the brain, which varies both with experience and genetics. In , connectomes constructed in the Human Connectome Project identified persons in a group of 126 persons with high accuracy (approximately 93%). It was noted that analyzing just the frontoparietal network was more distinctive across subjects compared to analyzing the whole brain (p < 10-9), and that these ‘neural fingerprints’ did not change when a person was at rest or performing a task. A person’s fMRI data from one session was used to create a connectivity matrix of their neural connections. This connectivity matrix was then compared to all other connectivity matrices from the remaining sessions, and a correlation coefficient for each session was calculated.
Disadvantages of using fMRI as a biometric is that the subject must remain stationary, it is contraindicated in subjects with implanted metallic devices, and it is significantly more expensive than other methods of biometric verification. However, conclusions drawn on fMRI data are likely to hold true on developing mobile technologies, such as DOT (Diffuse Optical Tomography) . DOT has a greater temporal resolution than fMRI, allowing for continuous verification. However, it may be more dependent on detector placement than EEG .
- Facial movements, eeg authentication using artifacts http://link.springer.com/chapter/10.1007/978-3-319-07995-0_34#page-1
- Multi-level approach based on eye-blinking http://www.sciencedirect.com/science/article/pii/S0167865515002433
- DOI is good http://www.nature.com/nphoton/journal/v8/n6/full/nphoton.2014.107.html
- DOI vs fMRI – http://www.ncbi.nlm.nih.gov/pubmed/23578579