arXiv subjects

Guobing Sun

Publications and source records attributed to Guobing Sun.

5 recordsLinked to original sources

arXiv2026

Dense and Query-Set Prediction for J-Peak Detection in Pillow-Based Ballistocardiography

Pillow-based ballistocardiography (BCG) enables unobtrusive cardiac monitoring, but J-peak detection is commonly learned as dense sequence labeling although the desired output is a sparse event set. This letter asks a narrower question: how do dense and query-set outputs differ when the data, encoder, validation protocol, and event evaluator are controlled? We formulate one-dimensional point-set prediction with 64 learned queries and Hungarian assignment, and compare it with a shared-backbone dense Transformer and a U-Net--BiLSTM. Evaluation uses five-subject leave-one-subject-out testing, three independently seeded runs, validation-only postprocessing selection, and strict one-to-one peak association. The dense Transformer attains the highest subject-wise pooled F1 ($0.786\pm0.105$) and precision ($0.817\pm0.089$), whereas U-Net--BiLSTM obtains $0.780\pm0.094$ F1. Set+DN reaches $0.764\pm0.112$ F1 but the lowest beat-count error ($0.862\pm0.584$ beats/epoch), compared with $2.790\pm1.356$ for the dense Transformer. DN changes set-model F1 by only $+0.004$. The results identify distinct event-accuracy and count-fidelity operating points; they do not establish universal superiority of either output formulation.

eess.SP↗

arXiv2026

SomnoNet: A Lightweight and Interpretable Framework for Sleep Staging Using Single-Channel EEG

Automated sleep staging from single-channel electroencephalography (EEG) is attractive for scalable sleep assessment, but practical systems must jointly address accuracy, efficiency, and clinical interpretability. We propose SomnoNet, a hierarchical raw-EEG framework motivated by expert scoring practice. The model first extracts multi-scale local rhythm representations from short temporal chunks and then integrates intra-epoch organization and inter-epoch context using hierarchical temporal modeling. On two large public benchmarks, SomnoNet achieves 80.9\% accuracy, 79.0\% macro-F1, and 0.739 kappa on Physio2018, and 88.0\% accuracy, 80.7\% macro-F1, and 0.831 kappa on SHHS. To support resource-constrained deployment, we further develop SomnoNet-Nano, a frozen-encoder compact variant that reuses the learned morphology encoder and replaces the original temporal stack with a lightweight sequence unit. SomnoNet-Nano contains 0.049M parameters, runs in 29.49 ms per 30-s epoch on an i7-12700F CPU under FP32 inference, and retains 99.5\% and 99.3\% of the full-model accuracy on Physio2018 and SHHS, respectively. Finally, rhythm-aware decision analysis visualizes segment-level model evidence and relates predictions to clinically meaningful EEG patterns. These results suggest that SomnoNet balances predictive performance, compactness, and transparent decision support for single-channel EEG sleep staging.

eess.SP↗

arXiv2026

SI-FID: Noise-Aware Fine-Tuning for Perceptual Quality Assessment of Stitched Images

Accurate evaluation of stitched image quality is essential for advancing stitching algorithms, yet existing objective metrics often diverge from human perception because they insufficiently capture stitching-specific artifacts such as ghosting and misalignment. To address this limitation, we propose SI-FID, a noise-aware extension of the Fréchet Inception Distance tailored for stitched-image assessment. Instead of modifying the FID formulation itself, SI-FID adapts the underlying feature representation through contrastive fine-tuning with controlled perturbations introduced via data augmentation, thereby enhancing sensitivity to subtle stitching-induced distortions. A pre-trained InceptionV3 encoder is calibrated using both original and perturbed samples, yielding a perceptually aligned feature space for distribution-based quality evaluation. Experiments on two complementary benchmark datasets demonstrate that SI-FID improves rank correlation with human subjective scores by over 25\% relative to conventional metrics, providing a more reliable and perceptually consistent indicator for stitched image quality.

eess.IV↗

arXiv2026

MAD: A Multimodal and Multi-perspective Affective Dataset with Hierarchical Annotations

This work presents MAD (Multimodal Affection Dataset), a multimodal emotion dataset designed for affective computing and neurophysiological modeling. MAD is built upon synchronous collection of diverse physiological signals (EEG, ECG, EOG, EMG, PPG, and BCG) together with tri-view RGB-D facial videos, enabling the observation of emotional dynamics from neural, physiological, and behavioral perspectives. The dataset consists of synchronized recordings from 18 participants and introduces two key contributions. First, it provides temporally aligned multimodal data that jointly capture central neural activity, peripheral physiological responses, and overt facial expressions. Second, it incorporates a three-level emotion annotation framework spanning stimulus elicitation, subjective cognition, and behavioral expression, supporting joint modeling of the full emotion process. To validate the dataset, we conduct systematic benchmark experiments covering intra-subject EEG emotion recognition, cross-subject EEG transfer learning, consistency analysis and emotion classification with cardiac-related signals, multimodal physiological fusion, and multi-view facial emotion recognition. The experimental results demonstrate that MAD supports consistent and comparable performance across both unimodal and multimodal settings, establishing it as a reliable benchmark for emotion recognition and cross-modal affective analysis, and as a valuable resource for studying emotion mechanisms across multiple levels.

eess.SP↗

arXiv2024

MixCut:A Data Augmentation Method for Facial Expression Recognition

In the facial expression recognition task, researchers always get low accuracy of expression classification due to a small amount of training samples. In order to solve this kind of problem, we proposes a new data augmentation method named MixCut. In this method, we firstly interpolate the two original training samples at the pixel level in a random ratio to generate new samples. Then, pixel removal is performed in random square regions on the new samples to generate the final training samples. We evaluated the MixCut method on Fer2013Plus and RAF-DB. With MixCut, we achieved 85.63% accuracy in eight-label classification on Fer2013Plus and 87.88% accuracy in seven-label classification on RAF-DB, effectively improving the classification accuracy of facial expression image recognition. Meanwhile, on Fer2013Plus, MixCut achieved performance improvements of +0.59%, +0.36%, and +0.39% compared to the other three data augmentation methods: CutOut, Mixup, and CutMix, respectively. MixCut improves classification accuracy on RAF-DB by +0.22%, +0.65%, and +0.5% over these three data augmentation methods.

cs.CV↗