Quantifying mental workload of operators performing n-back working memory task: Toward fNIRS based passive BCI system

摘要

Functional near infrared spectroscopy (fNIRS) as a relatively new brain imaging technique is increasingly being used for functional brain assessment. This however, needs robust hemodynamic features to be extracted in an appropriate time interval from proper brain's location. In this research, we performed a statistical analysis to evaluate the effect of such parameters as chromophore type, hemisphere, position of the recording channel and workload level on hemodynamic derived features during an n-back working memory task. In continue, the performance of three different classifiers, together with three feature ranking algorithms in classifying user's current cognitive state from single trial of hemodynamic signals were evaluated. The results revealed that a reliable change in brain oxygenation as a function of workload level was observed in dorsolateral prefrontal recording sites. In addition, maximum classification accuracy of 63.7% was achieved in classification between three workload levels. Finally, the time interval of 4 to 34 seconds after the task onset was found to be the most effective time window for feature extraction.