It has been established that neural response is time-locked to stimulus; however, the latencyin between may vary because of stimulus strength, subject fatigue, distraction, etc.;Instead of assuming perfect time-locking between stimulus and its neural response, we proposed here a statistical model that admits latency variation.;We tested the approach on an EEG data set from an image Rapid Serial Visual Presentation (RSVP) experiment. Results show that the proposed approach consistently outperforms those relying on perfect time-locking.;In addition, our approach can predict the stimulus' onset time when this information is not available. |