Information Integration for Neuroscience Team Seminar (Dr. Fadilla Zennifa）

This is an online seminar. Registration is required.
Registered participants will receive details for online access.

Title: Biometric Labeling for Supervised Machine Learning in attention level evaluation using hybrid EEG-ECG-NIRS.

Abstract:
Attention is a condition when someone concentrates on a specific task while ignoring other perceivable information. Numerous methods of attention level evaluation such as observation, self-assessment, and objective performance have been applied especially in supervised machine learning. But those methods tend to be delayed, sporadic, not at the moment in time, and based on participant cognitive ability. This study proposed a new labeling method for attention level evaluation by using biometric information.

To find suitable biometric information for data labeling, this study tried to investigate the response of blink rates and pupillometry toward task load during 10 seconds. Pupillometry showed the difference between the attention level in the late 4 seconds (P<0.05). On the other hand, blink rates did not show differences in attention level (P>0.05). After that, pupillometry data was converted into z- score and plot the data into a histogram to find parameter settings range in pupillometry in different attention levels. This study showed z-score within a specific range (-0.965 ≤ pupil ≤ 1.014) as high attention, other that range, as low attention showed the best parameter settings range.

Furthermore, two-electrode wireless EEG, a wireless ECG, and two wireless channels NIRS which pupillometry label has been used to evaluate attention levels during task load. This study calculated the classification accuracy by using the combination of correlation-based feature selection and k-nearest neighbor algorithm. This algorithm helped to find the most optimum features to be used for classifying the data. The CFS+kNN shown the highest performance (83.33± 5.95%) compared with other methods such as CFS+SVM (55.49± 27.89%), kNN (80.84± 3.88%), and SVM (55.88± 13.14%). Overall, these results demonstrate that the proposed method can be used to improve performance in attention level evaluation.

Keyword: Supervised machine learning, Electroencephalograph, Electrocardiograph, Near-Infrared spectroscopy, Pupillometry, Hybrid system