FACE: Few-shot Adapter with Cross-view Fusion for Cross-subject EEG Emotion Recognition
Organizations: the Guangdong Provincial Key Laboratory of Computational AI Models and Cognitive Intelligence, the School of Computer Science and Engineering, South China University of Technology, Guangzhou 510006, China · the Pazhou Lab, Guangzhou 510335, China · Engineering Research Center of the Ministry of Education on Health Intelligent Perception and Paralleled Digital-Human, Guangzhou, China
Abstract
Cross-subject EEG emotion recognition is challenged by significant inter-subject variability and intricately entangled intra-subject variability. Existing works have primarily addressed these challenges through domain adaptation or generalization strategies. However, they typically require extensive target subject data or demonstrate limited generalization performance to unseen subjects. Recent few-shot learning paradigms attempt to address these limitations but often encounter catastrophic overfitting during subject-specific adaptation with limited samples. This article introduces the few-shot adapter with a cross-view fusion method called FACE for cross-subject EEG emotion recognition, which leverages dynamic cross-view fusion and effective subject-specific adaptation. Specifically, FACE incorporates a cross-view fusion module that dynamically integrates global brain connectivity with localized patterns via subject-specific fusion weights to provide complementary emotional information. Moreover, the few-shot adapter module is proposed to enable rapid adaptation for unseen subjects while reducing overfitting by enhancing adapter structures with meta-learning. Experimental results on three public EEG emotion recognition benchmarks demonstrate the effectiveness of FACE under few-shot cross-subject evaluation.
Figures & tables
| Dataset | Num of Subjects | Data per Subject per Session | Official Length | DE Samples per Subject per Session |
| SEED | 15 | 5 trials × 3 classes | 1s | 3394 |
| SEED-IV | 15 | 6 trials × 4 classes | 4s | 822 851 |
| SEED-V | 16 | 3 trials × 5 classes | 4s | 541 681 |
| Method | #Num of unlabeled target samples | #Num of labeled target samples | SEED Mean / STD (%) | SEED-IV Mean / STD (%) | |
| Supervised Learning | IAG [ 26 ] | ✗ | ✗ | 86.30 / 6.91 | 62.64 / 10.25 |
| GMSS [ 20 ] | 86.5 / 6.22 | 73.48 / 7.41 | |||
| PGCN [ 19 ] | 84.59 / 8.68 | 73.69 / 7.16 | |||
| Semi-Supervised learning | SDDA [ 8 ] | ALL | 225/ 35 | 91.08 / 7.70 | 81.58 / 8.72 |
| SSML [ 29 ] | 10-shot | 88.59 | - | ||
| Method | #NUS | #NLS | SEED-V Mean / STD (%) |
| IAG [ 26 ] | ✗ | ✗ | 59.68 / 9.44 |
| Progressive GCN [ 56 ] | 71.40 / 9.43 | ||
| PGCN ∗ [ 19 ] | 65.21 / 10.10 | ||
| JAGP [ 57 ] | ALL | ✗ | 75.43 / 7.57 |
| SSAA-MSDA ∗ [ 53 ] | 63.09 / 10.01 | ||
| PR-PL ∗ [ 54 ] | 72.62 / 12.67 |
| CVF | FSA | SEED | SEED-IV | SEED-V | ||||||
| 3-shot | 5-shot | 10-shot | 3-shot | 5-shot | 10-shot | 3-shot | 5-shot | 10-shot | ||
| 81.80 7.79 | 86.60 6.27 | 89.70 5.36 | 80.37 4.33 | 84.65 3.75 | 88.21 3.07 | 78.31 5.23 | 82.63 4.70 | 85.20 5.17 | ||
| ✓ | 90.84 5.26 | 93.70 3.85 | 95.36 3.03 | 82.57 4.44 | 87.03 3.86 | 90.73 3.26 | 87.60 3.80 | 91.08 3.38 | 94.63 2.23 | |
| ✓ | 84.51 7.04 | 88.86 5.48 | 94.02 3.77 | 83.48 4.16 | 88.75 2.74 | 95.10 1.30 | 87.81 2.81 | 94.09 1.58 | 98.23 0.87 | |
| ✓ | ✓ | 91.66 3.72 | 93.96 2.70 | 96.72 1.81 | 83.91 3.87 | 89.51 3.13 | 95.95 1.47 | 89.55 2.98 | 95.20 1.42 | 98.95 0.36 |
| SEED | SEED-IV | SEED-V | |
| DGCN | 86.60 6.27 | 84.65 3.75 | 82.63 4.70 |
| Conv3 | 88.85 5.39 | 80.74 6.32 | 90.53 4.73 |
| CVF (DGCN+Conv3) | 93.70 3.85 | 87.03 3.86 | 91.08 3.38 |
| DGCNN | 83.60 6.73 | 62.75 6.96 | 69.55 11.75 |
| Conv3 | 88.04 4.88 | 79.63 5.81 | 90.41 4.64 |
| CVF (DGCNN+Conv3) | 88.55 5.91 | 67.47 7.00 | 87.63 6.14 |
| SEED | SEED-IV | SEED-V | |
| Self Attention | 78.38 7.91 | 70.59 7.04 | 65.62 8.29 |
| Cross Attention | 71.91 9.83 | 59.69 4.98 | 73.17 9.70 |
| CVF w/o MAML | 93.28 4.10 | 86.58 3.82 | 90.52 3.42 |
| 92.93 3.75 | 86.15 3.40 | 90.79 3.84 | |
| CVF(ours) | 93.70 3.85 | 87.03 3.86 | 91.08 3.38 |
| SEED | SEED-IV | SEED-V | |
| Linear Probing | 84.40 6.57 | 84.24 4.03 | 83.64 4.45 |
| LoRA | 87.55 3.78 | 83.14 3.57 | 79.29 5.62 |
| Serial Adapter | 87.58 5.87 | 85.11 4.13 | 84.75 5.44 |
| Residual Adapter | 88.09 5.83 | 86.96 3.46 | 88.06 3.46 |
| FSA(ours) | 88.86 5.48 | 88.75 2.74 | 94.09 1.58 |
| Method | SEED-IV (1s) | SEED-IV (4s) | SEED-V (1s) | SEED-V (4s) |
| PR-PL [ 54 ] | 81.32 8.53 | – | – | – |
| 71.96 7.56 | 74.46 4.56 | 64.19 8.48 | 72.62 12.67 | |
| 81.89 9.57 | 84.64 4.49 | 76.30 11.17 | 83.52 11.17 | |
| SSAA-MSDA [ 53 ] | – | 76.23 9.02 | – | – |
| 67.91±6.47 | 73.65 6.73 | 53.69±6.09 | 63.09 10.01 | |
| 78.40±10.34 | 83.73 7.64 | 62.75±5.43 | 72.57 10.55 |
| SEED | SEED-IV | SEED-V | ||
| 90.63 4.43 | 83.75 4.30 | 89.55 2.98 | ||
| 91.66 3.72 | 83.91 3.87 | 89.37 3.16 | ||
| 90.51 4.52 | 83.89 4.26 | 89.46 2.72 | ||
| 93.34 2.89 | 89.51 3.13 | 95.20 1.42 | ||
| 93.96 2.70 | 89.30 2.94 | 95.09 1.39 | ||
| 93.50 3.17 | 89.37 3.04 | 95.16 1.26 |
EEG Emotion Recognition
Electroencephalography (EEG) emotion recognition uses brain signals to classify or track affective states across people and recording settings.
See all 34 papers