Measuring Human Auditory Attention with EEG

Human attention has been a complex topic of study for decades to understand cognitive ability. The electroencephalogram (EEG) signals act as an essential tool for understanding human cognitive ability and remain an active part of the study in clinical psychology. This paper attempts to improve upon the current research on selective auditory attention while considering the task of speech-to-text annotation along with the EEG signals. It proposes a new model to measure auditory attention using the 14 channel EEG signals from the Physiology of Auditory Attention (PhyAat) dataset. Using EEGNet and other classical machine learning techniques, a thorough study shows the improvement in the attention prediction task for test MAE of 29.65 to 22.47 on a single subject of the PhyAat dataset. This paper also demonstrates the effectiveness of EEG for auditory attention measurement across different subjects while achieving a Mean Absolute Error (MAE) of 31.47 overall and 28.83 for Subject 1 using the same model on the same test splits defined by the PhyAat dataset.