EEG_Shape_Reconstruction

Title: Multi-task Generative Adversarial Learning on Geometrical Shape Reconstruction from EEG Brain Signals

PDF: ICONIP2019, arXiv

Authors: Xiang Zhang ([email protected]), Xiaocong Chen, Manqing Dong, Huan Liu, Chang Ge, Lina Yao ([email protected])

Overview

This repository contains reproducible codes for the proposed EEG_Shape_Reconstruction model. In this work, we propose a novel approach to reconstruct the geometrical shape based on the brain signals. We first develop a framework learning the latent discriminative representation of the raw EEG signals, and then, based on the learned representation, we propose an adversarial reconstruction framework to recover the geometric shapes which are visualizing by the human. In particular, we propose a semantic alignment method to improve the realism of the generated samples and force the framework to generate more realistic geometric shapes. The proposed approach is evaluated over a local dataset and the experiments show that our model outperforms the competitive state-of-the-art methods both quantitatively and qualitatively. Please check our paper for more details on the algorithm.

Qualitative comparison. Our model can reconstruct all the shapes correctly which have the highest similarity with the ground truth.

Code

EEG feature learning

In EEG_featurelearning.py, we put our CNN classifier which used to extract EEG features.

Multi-task GAN

In GAN_shape.py, we build a multi-task GAN along with the semantic alignment to reconstruct the shapes.

Citing

If you find our work useful for your research, please consider citing this paper:

@article{zhang2019multi,
  title={Multi-task Generative Adversarial Learning on Geometrical Shape Reconstruction from EEG Brain Signals},
  author={Zhang, Xiang and Chen, Xiaocong and Dong, Manqing and Liu, Huan and Ge, Chang and Yao, Lina},
  journal={arXiv preprint arXiv:1907.13351},
  year={2019}
}

Miscellaneous

Please send any questions you might have about the code and/or the algorithm to [email protected].

License

This repository is licensed under the MIT License.