Sparse EEG compressive sensing for web-enabled person identification

Electroencephalogram (EEG) person identification, a relatively new biometric method, aims to distinguish subjects by measuring, extracting and comparing features of EEG signals. This paper introduces sparse EEG compressive sensing to web-enabled EEG person identification and demonstrates the feasibility. Specifically, in order to adjust the person identification system to ubiquitous web-enabled scenarios, a wearable consumer-grade EEG headset with sparsely distributed dry electrodes is used to record EEG signals on the motor cortex and a nondirective short-time mental task is designed to simplify the measuring process. Moreover, the EEG data compression module reduces the volume of data for web transmission dramatically. Then feature vectors are extracted from reconstructed EEG data by analyzing power spectral density (PSD), concentration and meditation index. Lastly, a data segmentation based support vector machine classification method is proposed, which eliminates the classification error caused by unreliable EEG data segments. The final identification accuracy reaches 93.73% and the experimental results also indicate that each of the 16 subjects spends only 60 s and 10 s on training and identification respectively while the transmitted data are cut by 50%. It proves that the person identification based on sparse EEG compressive sensing is viable in consumer-grade and web-enabled applications.