Classifying Faces: Can an Off-The-Shelf System be Effective?

Objective: Our overall research goal is to use electroencephalography (EEG) to identify recognition of faces by patients with dementia to improve reminiscence therapy. The objective of this paper was to evaluate the performance of the consumer grade Emotiv Insight using a facial recognition event related potential (ERP) task. Good classification accuracy would enable the use of commercial systems such as the Insight to be used for reminiscence therapy. Methods: EEG was recorded with the Insight while obscure and famous facial stimuli were presented; participants confirmed recognition with a button press and oral confirmation. ERPs were averaged across five presentations of each facial stimuli to improve the signal to noise ratio. Classifiers were trained on one session, tested on another session, and the accuracy of each system was compared. Results: The Insight did produce ERPs typical of recognition memory however, it did not perform much better than chance. Event integration and artefacts were a significant problem. Conclusion: The Insight has practical advantages; however, the recording quality is not high enough to achieve adequate classification accuracy.