eduzhai > Applied Sciences > Engineering >

Evidence of Task-Independent Person-Specific Signatures in EEG using Subspace Techniques

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 16 pages

Abstract: Electroencephalography (EEG) signals are promising as alternatives to otherbiometrics owing to their protection against spoofing. Previous studies havefocused on capturing individual variability by analyzingtask condition-specific EEG. This work attempts to model biometric signaturesindependent of task condition by normalizing the associated variance. Towardthis goal, the paper extends ideas from subspace-based text-independent speakerrecognition and proposes novel modifications for modeling multi-channel EEGdata. The proposed techniques assume that biometric information is present inthe entire EEG signal and accumulate statistics across time in a highdimensional space. These high dimensional statistics are then projected to alower dimensional space where the biometric information is preserved. The lowerdimensional embeddings obtained using the proposed approach are shown to betask-independent. The best subspace system identifies individuals withaccuracies of 86.4 and 35.9 on datasets with 30 and 920 subjects,respectively, using just nine EEG channels. The paper also provides insightsinto the subspace model s scalability to unseen tasks and individuals duringtraining and the number of channels needed for subspace modeling.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×