To mitigate error accumulation and physical implausibility in long-term human motion prediction for HRC, this paper proposes the Dynamics-aware Spatio-Temporal Graph Transformer (DST-GT). By integrating spatial graph attention with multi-scale temporal attention, the model captures longrange dependencies and enforces biomechanical realism through velocity, acceleration, and joint feasibility constraints. Results on the PEC dataset show that DST-GT outperforms ST-GCN and AGCRN in 1.65s-3.3s forecasting, effectively suppressing trajectory drift and enhancing collaborative stability.

This study proposes an innovative WalkBCI real-time motor imagery brain-computer interface system to solve the problems of time-consuming model calibration and low EEG decoding accuracy in stroke patients because of individual differences. WalkBCI integrates generative adversarial networks and transfer-learning techniques to generate motor imagery feature data via RM-GAN using the resting data of target subjects, which shortens the time of calibration data acquisition and reduces patient fatigue. Meanwhile, the system enhances classification performance by combining source domain data during transfer learning optimization. The system is evaluated in both offline and online experiments to assess its practicality and stability. The study results show that WalkBCI outperformed traditional methods on the stroke patient dataset. In the offline experiments, the model's accuracy, precision, recall, and F 1-score of the model increased by 3.3%, 2.6%, 3.8%, and 3.4%, respectively. In online experiments, the system maintained stable classification performance with a maximum accuracy of 72.5%, while reducing calibration time by 58%. WalkBCI effectively lowered subjects' fatigue, and the comparison experiments revealed average reductions of 7.5 points in FS-14 scores and 74.05 in frontal theta/beta values. This study offers an efficient and stable solution for real-time EEG decoding, particularly well-suited for motor-imagery tasks in stroke patients.

OBJECTIVES:To investigate the electroencephalography (EEG) features in patients with spinal cord injury (SCI) experiencing neuropathic pain (NP) by analyzing the γ (30-80 Hz) frequency band. METHODS:Participants were divided into three groups: the PWP group (SCI patients with pain), the PNP group (SCI patients without pain) and AB group (able-bodied). Then we recorded and preprocessed the EEG data and calculated the power spectral density (PSD). Then, the amplitude of γ band oscillation (GBO) and the strength of brain network connections among the different groups were compared. RESULTS:In the PWP group, early GBO was observed in the FPZ channel in the resting state, and the main effect was concentrated in the 55-60 Hz range. In the high-frequency γ band (55-80 Hz) of the FPZ channel, the PWP group showed a higher GBO than the other groups. Moreover, during motor imagination (MI), the PWP group showed significantly stronger functional connectivity than the AB group, but significantly weaker than the PNP group. CONCLUSION:The GBO related to NP is highly consistent with the frequency domain manifestations of chronic pain. NP causes persistent abnormal increases in the γ-band PSD in multiple brain regions and the GBO changes in the prefrontal cortex have the strongest correlation with NP symptoms.

Background Electroencephalography (EEG) signals can be used to measure neuronal activity in different regions of the brain through electrodes. Objective To enhance the decoding of motor imagery (MI) EEG signals in spinal cord injury (SCI) patients, this study proposes a feature fusion graph convolutional neural network (F-GCN) model that integrates wavelet-based time-frequency features and functional topological relationships among EEG electrodes, aiming to improve classification accuracy and provide guidance for rehabilitation. Study design This study included 10 patients with spinal cord injuries as the experimental group, and 10 healthy individuals as the control group. After the experiment began, the subjects underwent 2-min recordings of their EEG signals in resting states with eyes open or closed, with records for each state repeated twice. The participants were then asked to imagine the movements of their left hand, and right hand. The entire process of MI consists of four task stages, with each stage containing three tasks. Each task randomly appears 10 times. Methods Time–frequency features of MI-EEG signals were extracted using a continuous wavelet transform to enhance the effectiveness of decoding raw EEG signals. Functional and statistical analyses of brain regions during MI were conducted based on the extracted time–frequency features. Based on this, the motor intentions of patients with SCI were decoded using a GCN that integrates the functional topological relationships of the electrodes. Results The proposed network achieved a classification accuracy of 92.44 % for MI task recognition. Furthermore, the fusion of wavelet features demonstrated superior performance in classification and recognition. Conclusions The results of this study confirm the efficacy of wavelet fusion in advancing MI feature decoding, enhancing the understanding of neurological conditions, such as SCI, and offering promising prospects for improving rehabilitation methods.

Traditional machine learning methods struggle with efficiency when processing large-scale data, while deep learning approaches, such as convolutional neural networks (CNN) and long short-term memory networks (LSTM), exhibit certain limitations when handling long-duration sequences. The choice of convolutional kernel size needs to be determined after several experiments, and LSTM has difficulty capturing effective information from long-time sequences. In this paper, we propose a transfer learning (TL) method based on Transformer, which constructs a new network architecture for feature extraction and classification of electroencephalogram (EEG) signals in the time-space domain, named TS-former. The frequency and spatial domain information of EEG signals is extracted using the Filter Bank Common Spatial Pattern (FBCSP), and the resulting features are subsequently processed by the Transformer to capture temporal patterns. The input features are processed by the Transformer using a multi-head attention mechanism, and the final classification outputs are generated through a fully connected layer. A classification model is pre-trained using fine-tuning techniques. When performing a new classification task, only some layers of the model are modified to adapt it to the new data and achieve good classification results. The experiments are conducted on a motor imagery (MI) EEG dataset from 16 spinal cord injury (SCI) patients. After training the model using a ten-time ten-fold cross-validation method, the average classification accuracy reached 95.09 %. Our experimental results confirm a new approach to build a brain-computer interface (BCI) system for rehabilitation training of SCI patients.

Objective.Obstructive sleep apnea (OSA) is a prevalent sleep disorder. Accurate sleep staging is one of the prerequisites in the study of sleep-related disorders and the evaluation of sleep quality. We introduce a novel GraphSleepFormer (GSF) network designed to effectively capture global dependencies and node characteristics in graph-structured data.Approach.The network incorporates centrality coding and spatial coding into its architecture. It employs adaptive learning of adjacency matrices for spatial encoding between channels located on the head, thereby encoding graph structure information to enhance the model's representation and understanding of spatial relationships. Centrality encoding integrates the degree matrix into node features, assigning varying degrees of attention to different channels. Ablation experiments demonstrate the effectiveness of these encoding methods. The Shapley Additive Explanations (SHAP) method was employed to evaluate the contribution of each channel in sleep staging, highlighting the necessity of using multimodal data.Main results.We trained our model on overnight polysomnography data collected from 28 OSA patients in a clinical setting and achieved an overall accuracy of 80.10%. GSF achieved performance comparable to state-of-the-art methods on two subsets of the ISRUC database.Significance.The GSF Accurately identifies sleep periods, providing a critical basis for diagnosing and treating OSA, thereby contributing to advancements in sleep medicine.

Motor imagery (MI) engages a broad network of brain regions to imagine a specific action. Investigating the mechanism of brain network reorganization during MI after spinal cord injury (SCI) is crucial because it reflects overall brain activity. Using electroencephalogram (EEG) data from SCI patients, we conducted EEG-based coherence analysis to examine different brain network reorganizations across different frequency bands, from resting to MI. Furthermore, we introduced a consistency calculation-based residual graph convolution (C-ResGCN) classification algorithm. The results show that the [Formula: see text]- and [Formula: see text]-band connectivity weakens, and brain activity decreases during the MI task compared to the resting state. In contrast, the [Formula: see text]-band connectivity increases in motor regions while the default mode network activity declines during MI. Our C-ResGCN algorithm showed excellent performance, achieving a maximum classification accuracy of 96.25%, highlighting its reliability and stability. These findings suggest that brain reorganization in SCI patients reallocates relevant brain resources from the resting state to MI, and effective network reorganization correlates with improved MI performance. This study offers new insights into the mechanisms of MI and potential biomarkers for evaluating rehabilitation outcomes in patients with SCI.

Objective: This study focuses on enhancing the applicability of brain-computer interface (BCI) systems for spinal cord injury (SCI) patients through improvements in electroencephalography (EEG) signal acquisition and decoding performance. Methods: This study developed an EEG acquisition optimization scheme leveraging virtual reality (VR) technology and integrated it with a deep learning (DL) framework for multifeature fusion to improve decoding performance. To enhance immersion, VR motor imagery (MI) task scenarios (e.g., fruit picking) were designed, reducing interference from traditional screen cues and optimizing the quality of EEG signal acquisition at its source. The framework uses a continuous wavelet transform algorithm to extract time-frequency features and incorporates a hybrid architecture combining convolutional neural networks and transformer models (C-CTNet) to enable efficient multidimensional feature fusion and decoding. Results: EEG signals in VR scenes exhibit a significant event-related desynchronization (ERD)/event-related synchronization (ERS) phenomenon. In addition, phase-locked value (PLV) analysis reveals enhanced functional connectivity in the brain network. Motion intention decoding using C-CTNet achieved classification accuracies of 86.11% on the BCI Competition IV 2a dataset, 89.53% on the PhysioNet EEG Motor Movement/Imagery dataset, and 92.41% on a dataset of ten SCI patients, significantly surpassing traditional methods. Conclusion: The findings confirm that integrating VR enhances EEG signal acquisition quality, while the proposed multifeature fusion framework improves BCI decoding performance. Significance: This method improves the accuracy of motor intention recognition while enhancing the stability and reliability of EEG signals, thereby providing key support for intelligent, personalized, and immersive neurorehabilitation. It shows promise for speeding functional recovery in SCI patients and laying the groundwork for clinical application and widespread adoption of BCI systems.

Objective. With the rapid development of brain-computer interface (BCI) technology, steady-state visual evoked potential (SSVEP) has emerged as an effective method for high-efficiency information transmission. However, traditional single-frequency stimulation methods face limitations in command set scalability and visual comfort.Approach. To address these issues, we propose a novel binocular SSVEP stimulation paradigm for brain-controlled unmanned vehicles. (UV) This system uses a checkerboard and phase encoding for stimulus presentation, encoding a single target with two frequencies to expand the command set. The frequencies are set between 30-35 Hz to enhance visual comfort. By leveraging polarized light technology, each eye receives distinct frequencies, suppressing intermodulation components and reducing the stimulated area for each eye. We also introduce an improved filter bank dual-frequency task-discriminant component analysis (FBD-TDCA) algorithm.Main results. Experimental results show that, in a 15-command simulation, only six frequencies successfully encoded all commands, achieving comparable performance to traditional single-frequency paradigms. Furthermore, the FBD-TDCA algorithm outperformed existing methods such as filter bank task-related component analysis and filter bank canonical correlation analysis, achieving a classification accuracy of 89.27% ± 3.67 and an information translate rate of 163.87 ± 14.32 bits min-1, with statistical significance confirmed through pairedt-tests. The system's practical application was further demonstrated in an online 12-command UV control task. Participants achieved an average classification accuracy of 90.34% ± 8.75%, with most maintaining low path deviation rates during navigation tasks.Significance. The proposed binocular SSVEP stimulation paradigm and FBD-TDCA algorithm address the limitations of traditional methods, offering enhanced command set scalability, improved visual comfort, and superior performance, paving the way for more efficient and user-friendly BCI applications in real-world scenarios.

Stroke, an abrupt cerebrovascular ailment resulting in brain tissue damage, has prompted the adoption of motor imagery (MI)-based brain-computer interface (BCI) systems in stroke rehabilitation. However, analyzing electroencephalogram (EEG) signals from stroke patients poses challenges. To address the issues of low accuracy and efficiency in EEG classification, particularly involving MI, the study proposes a residual graph convolutional network (M-ResGCN) framework based on the modified S-transform (MST), and introduces the self-attention mechanism into residual graph convolutional network (ResGCN). This study uses MST to extract EEG time-frequency domain features, derives spatial EEG features by calculating the absolute Pearson correlation coefficient (aPcc) between channels, and devises a method to construct the adjacency matrix of the brain network using aPcc to measure the strength of the connection between channels. Experimental results involving 16 stroke patients and 16 healthy subjects demonstrate significant improvements in classification quality and robustness across tests and subjects. The highest classification accuracy reached 94.91% and a Kappa coefficient of 0.8918. The average accuracy and F1 scores from 10 times 10-fold cross-validation are 94.38% and 94.36%, respectively. By validating the feasibility and applicability of brain networks constructed using the aPcc in EEG signal analysis and feature encoding, it was established that the aPcc effectively reflects overall brain activity. The proposed method presents a novel approach to exploring channel relationships in MI-EEG and improving classification performance. It holds promise for real-time applications in MI-based BCI systems.

Border security stands as a fundamental element in ensuring the property safety of citizens. Despite significant strides in artificial intelligence technology for border security and intrusion practices, the majority of research on intrusion signal classification primarily focuses on enhancing accuracy, with limited consideration of resource constraints. To construct a compact and efficient model that meets the requirements of small computing devices, this paper proposes a boundary security intrusion event recognition algorithm based on a distributed fiber optic sensing system. This algorithm combines the Gram Angle Field (GAF) and knowledge distillation network. The proposed approach establishes an efficient model with fewer parameters and computational resources. Compared to traditional vibration sensing systems, the Distributed Acoustic Sensing System (DAS) better leverages the advantages of distributed fiber extension. It utilizes the GAF to transform one-dimensional temporal signals into two-dimensional images, effectively filtering the impact of power fluctuations in the optical path on intrusion signals, and extracting deeper temporal features into the image. Knowledge distillation transfers the feature information trained by the teacher model to the smaller student model for intrusion signal recognition. The recognition accuracy reaches up to 98%, with a detection response time of approximately 0.78 s. Research results indicate that this approach can be utilized in developing high-precision lightweight intrusion signal detection models.

IntroductionTransformer network is widely emphasized and studied relying on its excellent performance. The self-attention mechanism finds a good solution for feature coding among multiple channels of electroencephalography (EEG) signals. However, using the self-attention mechanism to construct models on EEG data suffers from the problem of the large amount of data required and the complexity of the algorithm.MethodsWe propose a Transformer neural network combined with the addition of Mixture of Experts (MoE) layer and ProbSparse Self-attention mechanism for decoding the time-frequency-spatial domain features from motor imagery (MI) EEG of spinal cord injury patients. The model is named as EEG MoE-Prob-Transformer (EMPT). The common spatial pattern and the modified s-transform method are employed for achieving the time-frequency-spatial features, which are used as feature embeddings to input the improved transformer neural network for feature reconstruction, and then rely on the expert model in the MoE layer for sparsity mapping, and finally output the results through the fully connected layer.ResultsEMPT achieves an accuracy of 95.24% on the MI EEG dataset for patients with spinal cord injury. EMPT has also achieved excellent results in comparative experiments with other state-of-the-art methods.DiscussionThe MoE layer and ProbSparse Self-attention inside the EMPT are subjected to visualisation experiments. The experiments prove that sparsity can be introduced to the Transformer neural network by introducing MoE and kullback-leibler divergence attention pooling mechanism, thereby enhancing its applicability on EEG datasets. A novel deep learning approach is presented for decoding EEG data based on MI.

Transformer neural networks based on multi-head self-attention are effective in several fields. To capture brain activity on electroencephalographic (EEG) signals and construct an effective pattern recognition model, this paper explores the multi-channel deep feature decoding method utilizing the self-attention mechanism. By integrating inter-channel features with intra-channel features, the self-attention mechanism generates a deep feature vector that encompasses information from all brain activities. In this paper, a time-frequency-spatial domain analysis of motor imagery (MI) based EEG signals from spinal cord injury patients is performed to construct a transformer neural network-based MI classification model. The proposed algorithm is named time-frequency-spatial transformer. The time-frequency and spatial domain feature vectors extracted from the EEG signals are input into the transformer neural network for multiple self-attention depth feature encoding, a peak classification accuracy of 93.56% is attained through the fully connected layer. By constructing the attention matrix brain network, it can be inferred that the channel connections constructed by the attention heads have similarities to the brain networks constructed by the EEG raw signals. The experimental results reveal that the self-attention coefficient brain network holds significant potential for brain activity analysis. The self-attention coefficient brain network can better illustrate correlated connections and show sample differences. Attention coefficient brain networks can provide a more discriminative approach for analyzing brain activity in clinical settings.

Pain is an experience of unpleasant sensations and emotions associated with actual or potential tissue damage. In the global context, billions of people are affected by pain disorders. There are particular challenges in the measurement and assessment of pain, and the commonly used pain measuring tools include traditional subjective scoring methods and biomarker-based measures. The main tools for biomarker-based analysis are electroencephalography (EEG), electrocardiography and functional magnetic resonance. The EEG-based quantitative pain measurements are of immense value in clinical pain management and can provide objective assessments of pain intensity. The assessment of pain is now primarily limited to the identification of the presence or absence of pain, with less research on multilevel pain. High power laser stimulation pain experimental paradigm and five pain level classification methods based on EEG data augmentation are presented. First, the EEG features are extracted using modified S-transform, and the time-frequency information of the features is retained. Based on the pain recognition effect, the 20-40[Formula: see text]Hz frequency band features are optimized. Afterwards the Wasserstein generative adversarial network with gradient penalty is used for feature data augmentation. It can be inferred from the good classification performance of features in the parietal region of the brain that the sensory function of the parietal lobe region is effectively activated during the occurrence of pain. By comparing the latest data augmentation methods and classification algorithms, the proposed method has significant advantages for the five-level pain dataset. This research provides new ways of thinking and research methods related to pain recognition, which is essential for the study of neural mechanisms and regulatory mechanisms of pain.

ObjectivesTo identify potential treatment targets for spinal cord injury (SCI)-related neuropathic pain (NP) by analysing the differences in electroencephalogram (EEG) and brain network connections among SCI patients with NP or numbness.Participants and methodsThe EEG signals during rest, as well as left- and right-hand and feet motor imagination (MI), were recorded. The power spectral density (PSD) of the θ (4–8 Hz), α (8–12 Hz), and β (13–30 Hz) bands was calculated by applying Continuous Wavelet Transform (CWT) and Modified S-transform (MST) to the data. We used 21 electrodes as network nodes and performed statistical measurements of the phase synchronisation between two brain regions using a phase-locking value, which captures nonlinear phase synchronisation.ResultsThe specificity of the MST algorithm was higher than that of the CWT. Widespread non-lateralised event-related synchronization was observed in both groups during the left- and right-hand MI. The PWP (patients with pain) group had lower θ and α bands PSD values in multiple channels of regions including the frontal, premotor, motor, and temporal regions compared with the PWN (patients with numbness) group (all p < 0.05), but higher β band PSD values in multiple channels of regions including the frontal, premotor, motor, and parietal region compared with the PWN group (all p < 0.05). During left-hand and feet MI, in the lower frequency bands (θ and α bands), the brain network connections of the PWP group were significantly weaker than the PWN group except for the frontal region. Conversely, in the higher frequency bands (β band), the brain network connections of the PWP group were significantly stronger in all regions than the PWN group.ConclusionThe differences in the power of EEG and network connectivity in the frontal, premotor, motor, and temporal regions are potential biological and functional characteristics that can be used to distinguish NP from numbness. The differences in brain network connections between the two groups suggest that the distinct mechanisms for pain and numbness.

The inertial measurement unit is widely used in the design and implementation of the motion tracking system. This paper designs a ICM20948-based motion tracking system that supports wireless transmission. Our system consists of 15 motion collection modules, data receiver and PC. The motion collection module is responsible for collecting, calibrating and fusing motion data. The data receiver packs the motion data and transmits it to the PC through the USB transmission module. To reduce the effect of system error on the motion data, we use the 6-sided calibration algorithm to calibrate the accelerometer, use ellipsoidal algorithm to calibrate the magnetometer, and use the deviation value to calibrate the gyroscope. An adaptive extended Kalman filter is used for fusion of motion data to obtain accurate attitude corner information. We use NRF24L01 to build a transmission hardware and innovatively propose a wireless data transmission mechanism to ensure the transmission integrity of data. Our motion tracking system supports wireless transmission and can ensure the real-time, synchronization and stability when collecting and transmitting motion data.

Objective. Electroencephalogram (EEG) signals exhibit temporal-frequency-spatial multi-domain feature, and due to the nonplanar nature of the brain surface, the electrode distributions follow non-Euclidean topology. To fully resolve the EEG signals, this study proposes a temporal-frequency-spatial multi-domain feature fusion graph attention network (GAT) for motor imagery (MI) intention recognition in spinal cord injury (SCI) patients. Approach. The proposed model uses phase-locked value (PLV) to extract spatial phase connectivity information between EEG channels and continuous wavelet transform to extract valid EEG information in the time-frequency domain. It then models as a graph data structure containing multi-domain information. The gated recurrent unit and GAT learn EEG's dynamic temporal-spatial information. Finally, the fully connected layer outputs the MI intention recognition results. Main results. After 10 times 10-fold cross-validation, the proposed model can achieve an average accuracy of 95.82%. Furthermore, this study analyses the event-related desynchronization/event-related synchronization and PLV brain network to explore the brain activity of SCI patients during MI. Significance. This study confirms the potential of the proposed model in terms of EEG decoding performance and provides a reference for the mechanism of neural activity in SCI patients.

ObjectiveThe objective of this study was to analyze the changes in connectivity between motor imagery (MI) and motor execution (ME) in the premotor area (PMA) and primary motor cortex (MA) of the brain, aiming to explore suitable forms of treatment and potential therapeutic targets.MethodsTwenty-three inpatients with stroke were selected, and 21 right-handed healthy individuals were recruited. EEG signal during hand MI and ME (synergy and isolated movements) was recorded. Correlations between functional brain areas during MI and ME were compared.ResultsPMA and MA were significantly and positively correlated during hand MI in all participants. The power spectral density (PSD) values of PMA EEG signals were greater than those of MA during MI and ME in both groups. The functional connectivity correlation was higher in the stroke group than in healthy people during MI, especially during left-handed MI. During ME, functional connectivity correlation in the brain was more enhanced during synergy movements than during isolated movements. The regions with abnormal functional connectivity were in the 18th lead of the left PMA area.ConclusionLeft-handed MI may be crucial in MI therapy, and the 18th lead may serve as a target for non-invasive neuromodulation to promote further recovery of limb function in patients with stroke. This may provide support for the EEG theory of neuromodulation therapy for hemiplegic patients.

The ECG signal is often accompanied by noise, which can affect its shape characteristics, so it is important to perform signal de-noising. However, the commonly used signal noise reduction methods, such as wavelet or filter transformation, often prioritize high-frequency signals over low-frequency ones, leading to the loss of low-frequency band features or difficulties in capturing them. We propose a fusion reconstruction framework that combines hash autoencoder and margin semantic reinforcement to enhance low-frequency band features. Specifically, for labeled samples, margin semantic reinforcement identifies and corrects weight discrepancies among bands with similar waveforms but different labels to amplify the low-frequency signals associated with the label and reduce irrelevant ones. Meanwhile, hash autoencoder utilizes a semantic hash dictionary to reconstruct the original signal and mitigate noise pollution. For unlabeled samples, the hash autoencoder is utilized to generate pseudo-labels, followed by the reproduction of the aforementioned enhanced reconstruction process. The final step involves weighting the two types of signals, enhanced with margin semantics and hash autoencoder reconstruction, to achieve the reconstruction objective of the original signal, facilitating recognition and detection tasks. Experiments conducted on different classical classifiers demonstrate that the reconstructed ECG signals can significantly improve their performance.