Classifying Awareness with a Lightweight CNN in an Olfactory Oddball Passive BCI

Area of Science:

• Neuroscience
• Human-Computer Interaction
• Sensory Systems

Background:

• Olfaction offers potential for brain-computer interface (BCI) enhancement and passive cognitive state monitoring.
• Current challenges include precise odor delivery and robust brain activity interpretation algorithms.
• Reactive BCIs can use odor cues for intuitive commands, while passive BCIs can monitor cognitive states.

Purpose of the Study:

• To propose and evaluate a novel approach for predicting user awareness levels using olfactory stimuli.
• To demonstrate the feasibility of inferring user awareness from neurophysiological data.
• To develop and validate a lightweight convolutional neural network (CNN) for passive olfactory BCI.

Main Methods:

• Utilized an olfactory modality oddball paradigm.
• Combined electroencephalography (EEG) and electrobulbogram (EBG) recordings.
• Developed a multiclass, lightweight convolutional neural network (CNN) for data analysis.

Main Results:

• Achieved improved user awareness classification performance in a pilot study.
• The developed CNN surpassed previously reported results for this passive olfactory BCI modality.
• Demonstrated successful inference of user awareness levels from concurrent EEG and EBG data.

Conclusions:

• The proposed method combining EEG, EBG, and a CNN is feasible for inferring user awareness.
• This approach enhances the potential of olfactory BCI for cognitive state monitoring.
• The findings suggest a promising direction for developing more intuitive and informative BCIs.