!! This repository is currently under construction. More code will be added over the next few weeks !!

SparseMambaNet

SparseMambaNet: A Novel Architecture Integrating Bi-Mamba and Mixture of Experts for Efficient EEG-Based Lie Detection

Hanbeot Park†1, Yunjeong Cho†1, Hunhee Kim*

• *Correspondence
• †These authors contributed equally to this work.

Overview

This repository contains the code for the offline implementation of the EEG-based lie detection methods described in the paper "SparseMambaNet: A Novel Architecture Integrating Bi-Mamba and Mixture of Experts for Efficient EEG-Based Lie Detection". A demo to classify deception from scalp EEG signals during the CQT (Comparison Question Technique) task is provided.

Installation

• Python environment: Set up a new venv environment using the provided requirements.txt as follows:

pip install -r requirments.txt

This will install all the required Python packages to run this code (installation will take around 10-15 minutes).

• System requirements: This code runs on a GPU for optimal performance. It has been tested on Linux (22.04 LTS) with RTX 4090 GPU.

Data

The EEG dataset is not publicly available due to privacy concerns.

Demo for detect lie

The README files within each folder provide further information.