Skip to content
This repository was archived by the owner on Jul 9, 2025. It is now read-only.

cam-mind/yellow-team

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 

Repository files navigation

MIND Hackathon - Team Yellow

Project Overview

We explore the relationship between brain activity and music perception by building a cross-modal learning framework that fuses EEG, fMRI, and musical features.

Pipeline

  1. Pre-Processing
  • fMRI: Activation masking based on intensity.
  • EEG: Filtering using Notch and Bandpass filters.
  1. Audio Pre-Processing
  • Audio-to-Notes model based on a Transformer architecture (Transkun, Yan, 2024).
  • Audio-to-Chords translation.
  • Notes features extracted: Average pitch, Variance in pitch
  1. Model Architecture
  • Temporal Encoding: Using S4 layers to capture temporal dependencies and reduce information loss.
  • Cross-Modal Fusion: Supervised and Self-Supervised learning, Separate towers for EEG and fMRI signals, Resampling for temporal alignment.
  • Joint Embedding Space: Enables downstream tasks and interpretability through full linearity.

Key Ideas

  • Temporal Alignment of EEG and fMRI signals with musical features.
  • Supervised and Self-Supervised Cross-Modal Fusion for robust representation learning.
  • Full Linearity for better feature interpretability.
  • Brain Activity Captures Musical Semantics.

Run

In the src folder, we have provided two easy-to-use notebooks for the supervised and self-supervised fusion framework. Please change your data paths accordingly.

Link to the trained supervised model

The supervised model weights can be found at https://drive.google.com/file/d/1wIArSbeqgtdVjGCMpT1Mw2-9OcB-GvdD/view?usp=sharing

Team Yellow – MIND Hackathon Presentation Download Link

https://docs.google.com/presentation/d/1MtFk89UXxywOB0Fw8p8BR2spbh8IJdgr/edit?usp=sharing&ouid=108632635720761470695&rtpof=true&sd=true

About

MULTIMODAL INTERPRETABILITY-DRIVEN FUSION AND EMBEDDING

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors