MEG-to-MEG Transfer Learning and Cross-Task Speech/Silence Detection with Limited Data

Xabier de Zuazo; Vincenzo Verbeni; Eva Navas; Ibon Saratxaga; Mathieu Bourguignon; Nicola Molinaro

cs.LGFeb 20, 2026

MEG-to-MEG Transfer Learning and Cross-Task Speech/Silence Detection with Limited Data

Authors: Xabier de Zuazo, Vincenzo Verbeni, Eva Navas, Ibon Saratxaga, Mathieu Bourguignon, Nicola Molinaro

Organizations: HiTZ Center, University of the Basque Country – UPV/EHU, Spain · Basque Center on Cognition, Brain and Language – BCBL, Spain · Laboratory of Functional Anatomy, Faculty of Human Motor Sciences, Université libre de Bruxelles (ULB), Belgium · Laboratoire de Neuroanatomie et Neuroimagerie translationnelles (LN2T), ULB Neuroscience Institute, Université libre de Bruxelles (ULB), Belgium · WEL Research Institute, Belgium · Ikerbasque, Basque Foundation for Science, Spain

Abstract

Data-efficient neural decoding is a central challenge for speech brain-computer interfaces. We present the first demonstration of transfer learning and cross-task decoding for MEG-based speech models spanning perception and production. We pre-train a Conformer-based model on 50 hours of single-subject listening data and fine-tune on just 5 minutes per subject across 18 participants. Transfer learning yields consistent improvements, with in-task accuracy gains of 1-4% and larger cross-task gains of up to 5-6%. Not only does pre-training improve performance within each task, but it also enables reliable cross-task decoding between perception and production. Critically, models trained on speech production decode passive listening above chance, confirming that learned representations reflect shared neural processes rather than task-specific motor activity.

Figures & tables

Scratch (baseline)
Task		Accu. (%)	F1 (%)	AUC (%)
	listen	76.2 $\pm$ 4.8	85.5 $\pm$ 3.2	64.0 $\pm$ 9.4
	playback	75.1 $\pm$ 6.1	84.0 $\pm$ 4.7	67.7 $\pm$ 4.2
	production	83.6 $\pm$ 5.4	89.7 $\pm$ 3.5	81.1 $\pm$ 8.6
Transfer Learning
	listen	79.0 $\pm$ 4.8	87.7 $\pm$ 3.2	68.7 $\pm$ 6.2

Table 1: In-task evaluation results (mean

\pm

std across subjects, in %). Improvements over baseline are marked in bold.

Train $\rightarrow$ Test	Accu. (%)	F1 (%)	AUC (%)
listen to play.	72.5 $\pm$ 3.6	83.0 $\pm$ 2.7	59.4 $\pm$ 5.3
listen to prod.	71.5 $\pm$ 3.5	82.3 $\pm$ 2.6	57.9 $\pm$ 7.4
play. to listen	73.4 $\pm$ 2.9	83.7 $\pm$ 2.2	59.4 $\pm$ 5.7
play. to prod.	71.4 $\pm$ 4.4	82.1 $\pm$ 3.6	57.4 $\pm$ 6.7
prod. to listen	66.1 $\pm$ 6.3	77.7 $\pm$ 5.8	54.0 $\pm$ 6.3
prod. to play.	65.0 $\pm$ 5.9	76.5 $\pm$ 5.5	54.6 $\pm$ 4.7

Table 2: Cross-task decoding results for the model trained from scratch (mean

\pm

std across subjects, in %). All reported cross-task results are significantly above chance.

Train $\rightarrow$ Test	Accu. (%)	F1 (%)	AUC (%)
listen to play.	76.9 $\pm$ 4.0	86.5 $\pm$ 2.8	61.1 $\pm$ 4.5
listen to prod.	75.3 $\pm$ 5.2	85.3 $\pm$ 3.9	57.3 $\pm$ 7.3
play. to listen	78.0 $\pm$ 3.9	87.1 $\pm$ 2.9	61.6 $\pm$ 5.8
play. to prod.	73.6 $\pm$ 7.1	83.7 $\pm$ 6.1	58.1 $\pm$ 9.1
prod. to listen	69.3 $\pm$ 8.0	80.1 $\pm$ 7.3	56.0 $\pm$ 5.7
prod. to play.	68.3 $\pm$ 7.4	79.0 $\pm$ 7.2	56.6 $\pm$ 3.8

Table 3: Cross-task decoding results for the pretrained model (mean

\pm

std across subjects, in %). Improvements over baseline are marked in bold.

Figure 1: Effect size of transfer learning for cross-task decoding. Bars show the mean improvement in F1 score (transfer learning minus training from scratch) for each task and train-test task pair. Error bars indicate the standard deviation across subjects.

Figure 2: Cross-task F1 asymmetry for each task pair. Each x-axis category shows two directions: circle markers correspond to the left-to-right direction in the label, and diamond markers to the reverse direction. For each model, the two direction-specific points are connected to highlight asymmetry. Error bars indicate the standard deviation across subjects.

Figure 3: Subject-level in-task F1 improvement with transfer learning. Each bar shows the improvement in F1 score (transfer learning minus training from scratch) for a single subject and task. Colors indicate the three tasks.

Explore similar work

MEBM-Speech: Multi-scale Enhanced BrainMagic for Robust MEG Speech Detection
Feb 27, 2026Li Songyi, Zheng Linze, Liang Jinghua +1Neural Decoding Speech Processing
Speech and Hearing Research Center, School of Intelligence Science and Technology, Peking University, Beijing, China 100871 · Center for BioMed-X Research, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China 100871
Do EEG Foundation Models Transfer to Speech? A Benchmark on Overt and Imagined Speech Decoding
Jul 29, 2026Owais Mujtaba Khanday, Mohamed Baha Ben Ticha, Sanae Belfrouh +2EEG Decoding EEG Foundation Models
Dept. Signal Theory, Telematics and Communications, University of Granada, Spain · Research Centre for Information and Communication Technologies (CITIC-UGR), Spain · Laboratory of Information Technologies, University of Chouaib Doukkali, Morocco +1
Bypassing Direct Reconstruction: Speech Detection from MEG via Large-Scale Audio Retrieval
May 13, 2026Boda Xiao, Bo Wang, Heping ChengNeural Decoding Neuroimaging
Center for BioMed-X Research, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China · Speech and Hearing Research Center, School of Intelligence Science and Technology, Peking University, Beijing, China · National Biomedical Imaging Center, State Key Laboratory of Membrane Biology, Institute of Molecular Medicine, Peking-Tsinghua Center for Life Sciences, College of Future Technology, Peking University, Beijing, China

Feb 27, 2026J/K move · Enter open · S save

MEBM-Speech: Multi-scale Enhanced BrainMagic for Robust MEG Speech Detection

Li Songyi, Zheng Linze, Liang Jinghua, Zhang Zifeng

Speech and Hearing Research Center, School of Intelligence Science and Technology, Peking University, Beijing, China 100871 · Center for BioMed-X Research, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China 100871

Neural Decoding Speech Processing Multiscale Time-Series Modeling

Open paper AskPDF arXiv

We propose MEBM-Speech, a multi-scale enhanced neural decoder for speech activity detection from non-invasive magnetoencephalography (MEG) signals. Built upon the BrainMagic backbone, MEBM-Speech integrates three complementary temporal modeling mechanisms: a multi-scale convolutional module for short-term pattern extraction, a bidirectional LSTM (BiLSTM) for long-range context modeling, and a depthwise separable convolutional layer for efficient cross-scale feature fusion. A lightweight temporal jittering strategy and average pooling further improve onset robustness and boundary stability. The model performs continuous probabilistic decoding of MEG signals, enabling fine-grained detection of speech versus silence states - an ability crucial for both cognitive neuroscience and clinical applications. Comprehensive evaluations on the LibriBrain Competition 2025 Track1 benchmark demonstrate strong performance, achieving an average F1 macro of 89.3% on the validation set and comparable results on the official test leaderboard. These findings highlight the effectiveness of multi-scale temporal representation learning for robust MEG-based speech decoding.

Abstract

Explore similar work

MEBM-Speech: Multi-scale Enhanced BrainMagic for Robust MEG Speech Detection

Do EEG Foundation Models Transfer to Speech? A Benchmark on Overt and Imagined Speech Decoding

Bypassing Direct Reconstruction: Speech Detection from MEG via Large-Scale Audio Retrieval