A multilayer deep neural network framework for hemodynamic assessment of cognitive load management during...

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Cognitive load significantly impacts learning and task performance.
• Accurate classification of cognitive load is crucial for optimizing educational strategies and understanding cognitive processes.
• Existing methods for cognitive load assessment have limitations in real-time, non-invasive monitoring.

Purpose of the Study:

• To develop and validate a novel hybrid model for classifying cognitive load.
• To investigate the utility of functional Near-Infrared Spectroscopy (fNIRS) for real-time cognitive load monitoring.
• To enhance the accuracy of cognitive load classification by integrating Long Short-Term Memory (LSTM) networks with the Block Attention Module (BAM).

Main Methods:

• Collected fNIRS data from 50 participants performing problem-solving tasks under varying cognitive load conditions (high, medium, low).
• Preprocessed fNIRS data using normalization and wavelet transform for feature extraction.
• Developed a hybrid LSTM-BAM model to analyze temporal dependencies and refine feature representation for improved classification.

Main Results:

• The proposed LSTM-BAM model achieved significant performance improvements in classifying cognitive load levels.
• The integration of BAM effectively enhanced feature representation, aiding the LSTM in capturing crucial temporal dynamics.
• Demonstrated the efficacy of the hybrid architecture in distinguishing between different cognitive load states.

Conclusions:

• The hybrid LSTM-BAM model offers a promising approach for accurate cognitive load classification.
• fNIRS proves to be a valuable non-invasive tool for real-time cognitive performance monitoring.
• This research has implications for advancing instructional design and cognitive research through improved understanding of cognitive load.