mlsanityData to 14 Oct 2026

A Lightweight Transformer for Pain Recognition from Brain Activity

arXiv 2604.16491 ↗First posted 13 Apr 2026cs.CV cs.AIcited by 7

Not on the map. Why?

Abstract (arXiv)

Pain is a multifaceted and widespread phenomenon with substantial clinical and societal burden, making reliable automated assessment a critical objective. This paper presents a lightweight transformer architecture that fuses multiple fNIRS representations through a unified tokenization mechanism, enabling joint modeling of complementary signal views without requiring modality-specific adaptations or increasing architectural complexity. The proposed token-mixing strategy preserves spatial, temporal, and time-frequency characteristics by projecting heterogeneous inputs onto a shared latent representation, using a structured segmentation scheme to control the granularity of local aggregation and global interaction. The model is evaluated on the AI4Pain dataset using stacked raw waveform and power spectral density representations of fNIRS inputs. Experimental results demonstrate competitive pain recognition performance while remaining computationally compact, making the approach suitable for real-time inference on both GPU and CPU hardware.

Key facts. Cited by 7 papers in the map (data through 14 Oct 2026). Builds on 16 of its 68 references that are in the map. First posted on arXiv 13 Apr 2026.

Builds on

16 of 68 references.

References · 16

  1. 2023Transformer Encoder with Multiscale Deep Learning for Pain Classification Using Physiological Signalscited by 14
  2. 2024Twins-PainViT: Towards a Modality-Agnostic Vision Transformer Framework for Multimodal Automatic Pain Assessment using Facial Videos and fNIRScited by 14
  3. 2024Synthetic Thermal and RGB Videos for Automatic Pain Assessment utilizing a Vision-MLP Architecturecited by 12
  4. 2024A Full Transformer-based Framework for Automatic Pain Estimation using Videoscited by 11
  5. 2025Efficient Pain Recognition via Respiration Signals: A Single Cross-Attention Transformer Multi-Window Fusion Pipelinecited by 11
  6. 2025PainFormer: a Vision Foundation Model for Automatic Pain Assessmentcited by 10
  7. 2024Faces of Experimental Pain: Transferability of Deep Learned Heat Pain Features to Electrical Paincited by 9
  8. 2025Multi-Representation Diagrams for Pain Recognition: Integrating Various Electrodermal Activity Signals into a Single Imagecited by 9
  9. 2025Tiny-BioMoE: a Lightweight Embedding Model for Biosignal Analysiscited by 8
  10. 2024Multi-task Neural Networks for Pain Intensity Estimation using Electrocardiogram and Demographic Factorscited by 7
  11. 2024Transformer with Leveraged Masked Autoencoder for video-based Pain Assessmentcited by 7
  12. 2019A View on Edge caching Applicationscited by 4
4 more
  1. 2024Data Augmentation for 3DMM-based Arousal-Valence Prediction for HRIcited by 4
  2. 2025A Pain Assessment Framework based on multimodal data and Deep Machine Learning methodscited by 4
  3. 2025When and How to Express Empathy in Human-Robot Interaction Scenarioscited by 4
  4. 2024Virtual Reflections on a Dynamic 2D Eye Model Improve Spatial Reference Identificationcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 7.

7 papers.

  1. 2026A Unified Tokenization Framework for Pain Recognition using Heterogeneous 3D Modalitiescited by 7
  2. 2026ReFace: Reorganizing Facial Spatiotemporal Representations for Improved Pain Assessmentcited by 7
  3. 2026Towards a Unified Modality-Agnostic Multimodal Framework for Cognitive Workload Assessmentcited by 5
  4. 2026Beyond the Raw Waveform: Fusing Visual Representations of EDA for Stress Detectioncited by 3
  5. 2026A Multi-Scale Temporal Framework with Dynamic Fusion for EEG-Based Emotion Recognitioncited by 2
  6. 2026One-Block Transformer (1BT) for EEG-Based Cognitive Workload Assessmentcited by 1
  7. 2026MUPA2E: Multimodal Unified Perception with Asymmetric Attention for Emotion Assessmentcited by 0