Ventral and Dorsal Stream EEG Channels: Key Features for EEG-Based Object Recognition and Identification

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Object recognition and identification involve complex cognitive processes supported by visual and temporal cortices.
• These tasks, while similar, are distinct cognitive functions requiring integrated brain activity.

Purpose of the Study:

• To investigate if ventral and dorsal stream pathways contain information crucial for object recognition and identification model learning.
• To determine the optimal EEG channel configuration for accurate and efficient brain-computer interface (BCI) systems.

Main Methods:

• Utilized EEGNet models trained on data from object recognition and identification experiments.
• Compared models trained on specific pathways (ventral/dorsal streams) against a model using all channels.
• Employed Grad-CAM for visualizing channel contributions to model training.

Main Results:

• A model using only temporal region channels achieved 89% (recognition) and 85% (identification) accuracy.
• Incorporating ventral stream channels boosted accuracy to 95% (recognition) and 94% (identification).
• Grad-CAM confirmed significant contributions from ventral and dorsal stream channels.

Conclusions:

• The ventral stream channels are highly informative for object recognition and identification tasks.
• Optimizing channel selection based on brain pathways can enhance BCI performance.
• This research paves the way for developing more effective BCI systems for cognitive tasks.