Pain is a highly subjective and multidimensional experience involving both physiological and psychological factors. Developing accurate and robust pain monitoring systems is essential for effective assessment and rehabilitation management. Recently, multimodal fusion combining physiological and behavioral signals has gained attention for capturing complex pain responses. Unlike video and audio modalities that may raise privacy concerns, physiological signals such as electrodermal activity (EDA) and electromyography (EMG) offer a privacy-preserving alternative. In this study, we propose a bi-modal fusion framework based on EDA and EMG signals for pain intensity classification. The framework integrates a Long Short-Term Memory (LSTM) network with a spatial attention mechanism to capture both local and global temporal dependencies, and incorporates an adaptive weighting module to enhance inter-modal feature interaction. To address the class imbalance issue, we introduce focal loss as the loss function and apply a data-level jump reduction strategy. Extensive experiments on 11 sub-datasets from the X-ITE pain database demonstrate that our method outperforms state-of-the-art (SOTA) approaches, achieving an average accuracy improvement of 3.88% and a 7.12% gain on the R-ETD sub-dataset. The model also exhibits more reliable performance on minority and challenging classes, highlighting its robustness across modalities and sub-datasets. Moreover, the proposed method offers improved computational efficiency and reduced resource consumption. These results suggest that the proposed framework is a promising solution for multimodal pain monitoring with strong potential for real-world clinical applications.

Previous approaches to pain intensity classification have typically relied on small sets of top-performing features to maximize accuracy. While effective in constrained scenarios, such strategies neglect the diverse range of modalities available in modern pain databases. In this work, we introduce a multimodal feature set (MMFS) that integrates heterogeneous features from each biosignal modality in the BioVid and X-ITE databases. Our approach captures a broad spectrum of complementary information, maintaining robustness even when individual modalities are unavailable. Experimental results show consistent performance improvements, with classification accuracy increasing by up to 8% overall and by 6% when evaluating all pain levels. Through detailed analyses of individual modalities, confusion matrices, and a modality ablation study, we demonstrate that the combined effect of multimodality and balanced information distribution drives these gains. Furthermore, feature importance analysis reveals which inputs contribute most to final predictions and which features are most beneficial for the more challenging pain classes.