Rapid face recognition based on single-trial event-related potential detection over multiple brains

The automatic machine face recognition has achieved great performance but is still far from satisfactory in the uncontrolled environments. Human brain has a powerful ability to recognize faces across various conditions, which makes it possible to introduce brain-computer interface (BCI) technology to face recognition. However, the performance of single-participant based BCI suffers from the low signal-to-noise ratio of electroencephalography (EEG) signals, especially detecting the event-related potential (ERP) in single trial. Here, we propose a rapid face recognition approach based on single-trial ERP detection, but the EEG signals are integrated from multiple participants. After the first-layer classifier to detect the ERP of each individual, the support vector machine scores of all individuals are concatenated and inputted to a second-layer classifier by the voting method to obtain the final classification results. The results show that our approach significantly outperforms the single-participant based BCI, and the voting method is superior to the ERP averaging method. In addition, when the area under the receiver operating characteristic curve is required to be greater than 0.9, our approach can recognize target faces about 300 ms ahead of button-press responses when integrating EEG signals from 9 participants. These results indicate that our approach can integrate decisions from multiple individuals to achieve rapid and accurate face recognition.