XML Persian Abstract Print

Abstract:   (357 Views)
Introduction: In spite of more than a century research on how the brain perceives time, research on EEG pattern recognition related to time perception has been rare. Deficit in time perception has been demonstrated in many mental and neurological disorders such as Parkinson, ADHD, depression, autism and schizophrenia. Classification of event-related potentials in time discrimination tasks could be used as screening tool for such diseases. The goal of this study is recognizing whether a duration is judged as short or long by a person based on his electroencephalography (EEG) signal. Method: for this purpose the oddball paradigm was used. 24 male and female students of Tabriz University took part in the experiment. From which 18 were selected after artifact rejection. Written informed consent was obtained from all the participants. EEG of the participants were recorded simultaneous with the task presentation. in an auditory oddball task, after two or four presentation of a pure tone of frequency 1000Hz each one lasting for 500ms (standard stimuli) at a constant inter-stimulus interval of 300ms, a tone of a different frequency (500hz,700hz,1300hz or 1500hz) and with duration of 400ms or 600ms was presented (oddball stimulus).Participants were asked to compare the duration of the short (400ms) or long (600ms) oddball stimulus to the duration of its preceding standard stimuli (500ms). After the required preprocessing steps, averaging was performed to obtain the ERP segments at each of the experimental conditions. In order to reduce further analysis, Cz electrode was selected among the 64 channels used for recording since most of the factors were significant at Cz and the highest amplitude of ERP components were seen over there. Then, the labeled target samples (ERP segments related to either 400ms or 600ms oddball stimuli) and labeled non-target samples (ERP segments related to the standard stimuli) were fed to the classifiers as the training set.  Subsequent to training the classifiers with a subset of this dataset, test was executed on the reminder. Results: Behavioral results revealed that in 4 repetition condition, participants significantly perceived the oddball stimulus as longer than in the 2 repetition condition which implies that repetition causes time overestimation.  Electroencephalography results indicated that repetition suppression effect was evident throughout the ERP waveform. That is, repeating a standard stimulus causes reduction in the amplitude of ERP components elicited by it. Whereas, presentation of a novel stimulus –oddball- leads to an increase in the corresponding ERP components amplitude. Finally classification results showed that SVM (RBF) classifier could recognize the target samples (400ms interval) from non-target samples (500ms and 600ms intervals) with the highest accuracy of 94.25% among decision tree and multilayer perceptron (MLP) classifiers. Also, SVM (RBF) yielded the highest accuracy of 93.98% when classifying 600ms related samples (target) versus 500ms and 600ms related samples (non-target). Finally, it achieved the best accuracy of 87.95% when classifying the test samples into multiclass of 400ms, 600ms and 500ms classes separately. Conclusion: Results of this study, demonstrates that machine learning could accurately detect patterns related to short and long perceived intervals based on the peoples’ electroencephalography (EEG) signal.
Type of Study: Research |
Received: 2020/09/5 | Accepted: 2020/12/10

Add your comments about this article : Your username or Email:

Send email to the article author

Designed & Developed by : Yektaweb