SHINE: Sequential Hierarchical Integration Network for EEG and MEG

TL;DR

This paper introduces SHINE, a neural network model that reconstructs speech-silence sequences from MEG signals, achieving high accuracy in a speech detection challenge using extensive brain signal data.

Contribution

We propose SHINE, a novel hierarchical neural network for EEG/MEG data that improves speech detection accuracy and incorporates auxiliary tasks for enhanced training.

Findings

SHINE achieved F1-macro scores above 0.91 on the leaderboard.

Auxiliary reconstructions improved model performance.

Ensemble methods further boosted detection accuracy.

Abstract

How natural speech is represented in the brain constitutes a major challenge for cognitive neuroscience, with cortical envelope-following responses playing a central role in speech decoding. This paper presents our approach to the Speech Detection task in the LibriBrain Competition 2025, utilizing over 50 hours of magnetoencephalography (MEG) signals from a single participant listening to LibriVox audiobooks. We introduce the proposed Sequential Hierarchical Integration Network for EEG and MEG (SHINE) to reconstruct the binary speech-silence sequences from MEG signals. In the Extended Track, we further incorporated auxiliary reconstructions of speech envelopes and Mel spectrograms to enhance training. Ensemble methods combining SHINE with baselines (BrainMagic, AWavNet, ConvConcatNet) achieved F1-macro scores of 0.9155 (Standard Track) and 0.9184 (Extended Track) on the leaderboard test set.