Qisong Wang

Electroencephalography (EEG)-based emotion recognition provides an important physiological basis for affective brain–computer interfaces. However, high-dimensional EEG features vary across trials, sessions, and subjects. Euclidean distance and conventional source-only LMNN cannot jointly learn a discriminative geometry and align cross-domain emotion distributions. To address this problem, we propose Adaptive Large Margin Nearest Neighbor (AdaLMNN), an unsupervised domain-adaptation framework for EEG emotion recognition. AdaLMNN learns a shared Mahalanobis projection through normalized large-margin discrimination, adaptive margins, and refreshed source neighborhoods. It further uses confidence-guided target pseudo-labels, class-balanced conditional alignment, and scale regularization with scheduled optimization to reduce source–target mismatch without using target ground-truth labels. We evaluated AdaLMNN on synthetic data and the SEED and SEED-IV datasets using SVM, KNN, and MLP under three protocols. AdaLMNN substantially improved EEG emotion-classification performance, particularly under within-session and session-level shifts. Although cross-subject gains remained dataset- and classifier-dependent, the results demonstrate the potential of AdaLMNN as an interpretable and extensible representation-learning module for EEG emotion recognition under heterogeneous recording conditions.

Accurate inter-subject pain intensity assessment using EEG remains a major challenge due to substantial inter-subject variability. This study introduces a novel framework that leverages pain-related brain dynamics and transfer learning to enable reliable inter subject pain intensity classification. The pro-posed method first quantifies pain sensitivity from resting state EEG to identify source subjects with comparable neural pain signatures. High-activation pain brain sources are subsequently localized and remapped between source and target subjects. A classifier is trained to evaluate transfer suitability across subjects, and balanced distribution adaptation is applied to align brain source features, mitigating inter-subject variability. The adapted model infers pseudo labels for the target EEG, which guide the pain response extraction. Final classification is determined by selecting the model exhibiting the minimal cross-domain discrepancy between brain source and pain-evoked EEG features. Experimental evaluations on real EEG datasets demonstrate that the proposed method significantly out performs three existing approaches in inter-subject pain intensity classification. The proposed method effectively overcomes the problem of poor reliability in inter-subject pain intensity classification, providing a robust and clinically viable solution.

The human modeling technology has a broad application background in popular fields, such as security inspection, medical modeling, virtual fitting, and virtual human in the metauniverse. Addressing the limitations of current optical modeling methods in accurately representing the body surface covered by clothes accurately and meeting the precise location requirements of the acquisition equipment for heterodyne signal fusion, this article utilizes optical modeling based on a time-of-flight (ToF) camera and combines it with millimeter-wave technology. It also applies heterodyne signal fusion, leveraging the characteristics of the human body’s skeleton and edge points, to achieve the accurate modeling of both clothes and body surface. A human modeling system and method are proposed, based on the fusion of alien signal technology with millimeter-wave technology and ToF cameras, to solve the shortcomings of deep learning methods in estimating body surface estimation. The feasibility of the method in potential application areas such as anthropometry and visual security screening is finally verified through experiments. The measurement errors and visual effects in security screening are analyzed to demonstrate the practical value of the method.

Soil organic matter (SOM) is a fundamental indicator of soil health and a major component of the global carbon cycle; its accurate quantification is essential for sustainable agriculture. Conventional chemical assays yield only point-based soil measurements and miss the spatial distribution of soil elements; airborne hyperspectral remote sensing has emerged as a promising approach for the quantitative measurement and characterization of SOM. Inversion models translate hyperspectral data into quantitative SOM estimates. However, existing models rely solely on a single preprocessing pathway, limiting their ability to fully exploit available spectral information. We address these limitations by developing a marginal contribution-driven spectral fusion network (MC-SFNet) that conducts feature-level fusion of heterogeneous preprocessing outputs within a physics-guided deep architecture. Moreover, the combination of data-driven fusion and the Kubelka–Munk (KM) model yields more physically interpretable spectral features, advancing beyond prior purely data-driven methods. We validated MC-SFNet on a self-constructed remote sensing, high-throughput hyperspectral dataset comprising 200 black soil samples from Northeastern China (400–1000 nm, 256 bands). Experimental results show that our network reduces the RMSE by 10.7% relative to the prevailing generalized hyperspectral soil-inversion model. The proposed method provides a novel preprocessing pathway for forthcoming airborne high-throughput hyperspectral missions to extract soil-specific spectral information more effectively and further enhance large-scale SOM retrieval accuracy.

The pain-evoked potential electroencephalogram (EEG) is an effective electrophysiological indicator for pain assessment, yet its extraction is challenging due to interference from background activity and involuntary blinks. Although existing blink artifact-removal methods show efficacy, they face limitations such as the need for reference signals, neglect of individual differences, and reliance on user input, hindering their practical application in clinical pain assessments. In this paper, we propose a novel framework applying adaptive quadrature mirror filter banks (AQMFB) with discrete wavelet transform (DWT) to remove blink artifacts in pain EEG. Unlike traditional DWT methods that apply fixed wavelets across subjects, our method adapts wavelet construction based on the characteristics of EEG. Experimental results demonstrate that AQMFB-DWT outperforms four leading methods in removing blink artifacts with minimal distortion of pain information, all within an acceptable processing time. This technique is a valuable preprocessing step for enhancing the extraction of pain-evoked potentials.

Emotion recognition from electroencephalography (EEG) signals is increasingly emerging as a critical research focus in brain-computer interfaces (BCIs). However, challenges such as the scarcity of emotion labels and distribution discrepancies in EEG signals significantly hinder the practical application of EEG-based emotion recognition. To overcome these challenges, this paper fully exploits the continuity of emotion-related EEG data and proposes an unsupervised domain adaptation with pseudo-label propagation, termed DAPLP, for cross-domain EEG emotion recognition. Specifically, we first perform global distribution alignment between the source and target domains and utilize the source classifier to generate pseudo-labels for the target domain. From these predictions, reliable pseudo-labels are then selected to guide label propagation, and the propagation process is further optimized with correct and smooth techniques. Systematic experiments conducted on the SEED, SEED-IV, and SEED-V datasets reveal that the proposed DAPLP accomplishes competitive performance compared to advanced existing methods, reaching average accuracies of 89.44%/74.57%/69.15% in cross-subject evaluation and 96.41%/82.20%/84.70% in cross-session evaluation, respectively. Moreover, our proposed DAPLP exhibits strong practical potential and robust performance in unsupervised cross-domain emotion recognition.

Surface electromyography (sEMG) based hand gesture recognition has garnered widespread attention in rehabilitation robotics due to its non-invasive measurement and intuitive motion decoding. However, affected by various factors such as individual differences, achieving cross-user adaptability and longterm reliability for sEMG classification poses a significant challenge. Existing domain adaptation methods primarily focus on global distribution alignment to mitigate statistical distribution discrepancies across domains, yielding significant achievements. Nevertheless, these methods often overlook fine-grained category-level subdomain distribution alignment, leading to discriminative structure confusion and subdomain misalignment, which hinder cross-subject and cross-session gesture recognition. To tackle these issues, this paper proposes a plug-and-play subdomain adaptation method called PPSDA to enhance the classification performance and generalization ability for gesture recognition across domains. Specifically, handcrafted features are extracted and utilized for source domain supervised training to preserve discriminative structures. Subsequently, source and target domains co-training is performed, wherein the local Maximum Mean Discrepancy (LMMD) is minimized to capture fine-grained information on relevant subdomains for subdomain distribution alignment. To validate the performance of the proposed PPSDA, we recruited 12 healthy subjects and developed an sEMG-driven exoskeleton rehabilitation glove for cross-subject and cross-session evaluations. Extensive experimental results demonstrate the effectiveness and superiority of the proposed PPSDA over existing domain adaptation approaches.