Tao Wang

Wearable neurotechnology depends critically on continuous movement monitoring to characterize motor impairment and recovery in real-world settings. While joint torque serves as a clinically essential kinetic marker, estimating it directly on-device from inertial signals remains challenging due to stringent computational, memory, and energy constraints. Lightweight pipelines typically omit computationally expensive time–frequency processing; however, this omission degrades the observability of dynamics encoded in 1D IMU signals and diminishes the effectiveness of standard knowledge distillation strategies. To enable reliable on-device torque inference, we propose a Physically Guided Dual-Consistency Knowledge Distillation (PDC-KD) framework that explicitly integrates biomechanical priors into the learning process through two collaborative pathways: parameter-manifold alignment and physics-guided compensation. The student network receives guidance through Fisher-information-weighted parameter transfer, ensuring robust knowledge distillation despite significant model capacity mismatch. Furthermore, the framework incorporates a physics-guided regularization term that enforces dynamically consistent torque trajectories via a numerically stable Cholesky-parameterized constraint. Experiments demonstrate that the student model preserves teacher-level predictive accuracy while operating within the stringent resource constraints of edge devices (achieving a 98% parameter reduction, ∼2× faster inference, and ∼1 ms latency). Moreover, the proposed method yields torque estimates with enhanced dynamical consistency, providing an efficient biosignal-processing solution for wearable neurotechnology platforms demanding real-time movement analytics.

Unmanned aerial vehicles (UAVs) and satellites play a crucial role in maritime surveillance, yet ship detection in remote sensing imagery remains challenging due to small object sizes, arbitrary orientations, and cluttered backgrounds. Existing detectors struggle to simultaneously preserve fine-grained details for small ships and suppress background noise. To tackle this, we propose P2R-OBB, a YOLOv8-OBB–based framework that introduces an additional P2 feature pyramid level together with a dynamically recalibrated attention mechanism. The P2 feature pyramid retains high-resolution shallow features that are critical for small-object detection, yielding a substantial 17.5% mAP50 improvement on the complex DOTA v1-ship dataset. In parallel, the dynamic attention module adaptively recalibrates feature responses to emphasize ships while suppressing irrelevant background structures, delivering a 4.7% mAP50 gain on the SSDD+ dataset. When combined, these components exhibit a strong synergistic effect, achieving a substantial 11.1% absolute mAP50 improvement on the complex DOTA v1-ship dataset and setting a new state of the art for oriented ship detection. Our framework offers a robust and efficient solution, with its key contributions particularly demonstrated in detecting small and arbitrarily oriented ship targets in remote sensing imagery.

Biomechanical analysis of the human lower limbs plays a critical role in movement assessment, injury prevention, and rehabilitation guidance. Traditional gait analysis techniques, such as optical motion capture systems and biomechanical force platforms, are limited by high costs, operational complexity, and restricted applicability. In view of this, this study proposes a cost-effective and user-friendly approach that integrates inertial measurement units (IMUs) with a novel deep learning framework for real-time lower limb joint torque estimation. The proposed method combines time-frequency domain analysis through continuous wavelet transform (CWT) with a hybrid architecture comprising multi-head self-attention (MHSA), bidirectional long short-term memory (Bi-LSTM), and a one-dimensional convolutional residual network (1D Conv ResNet). This integration enhances feature extraction, noise suppression, and temporal dependency modeling, particularly for non-stationary and nonlinear signals in dynamic environments. Experimental validation on public datasets demonstrates high accuracy, with a root mean square error (RMSE) of 0.16 N·m/kg, Coefficient of Determination (R²) of 0.91, and Pearson correlation coefficient of 0.95. Furthermore, the framework outperforms existing models in computational efficiency and real-time applicability, achieving a single-cycle inference time of 152.6 ms, suitable for portable biomechanical monitoring systems.

Coreference resolution is a key task in Natural Language Processing. It is difficult to evaluate the similarity of long-span texts, which makes text-level encoding somewhat challenging. This paper first compares the impact of commonly used methods to improve the global information collection ability of the model on the BERT encoding performance. Based on this, a multi-scale context information module is designed to improve the applicability of the BERT encoding model under different text spans. In addition, improving linear separability through dimension expansion. Finally, cross-entropy loss is used as the loss function. After adding BERT and span BERT to the module designed in this article, F1 increased by 0.5% and 0.2%, respectively.

Sleep apnea is a sleep disorder caused by weakened or suspended breathing during sleep, which seriously affects the work and health of patients. The traditional polysomnography (PSG) detection process is complicated and expensive, which has attracted researchers to explore a rapid detection method based on single-lead ECG signals. However, existing ECG-based sleep apnea detection methods have certain limitations and complexities, mainly relying on human-crafted features. To solve the problem, the paper develops a sleep apnea detection method based on a residual attention mechanism network. The method uses the RR interval signal and the R-peak signal derived from the ECG signal as input, realizes feature extraction through the residual network (ResNet), and adds the SENet attention mechanism to deepen the mining of channel features. Experimental results show that the per-segment accuracy of the proposed method can reach 86.2%. Compared with existing works, its accuracy has increased by 1.1–8.1%. These results show that the proposed residual attention network can effectively use ECG signals to quickly detect sleep apnea. Meanwhile, compared with existing works, the proposed method overcomes the limitations and complexity of human-crafted features in sleep apnea detection research.

Aiming at the fact that traditional convolutional neural networks cannot effectively extract signal features in complex application scenarios, a sleep apnea (SA) detection method based on multi-scale residual networks is proposed. First, we analyze the physiological mechanism of SA, which uses the RR interval signals and R peak signals derived from the ECG signals as input. Then, a multi-scale residual network is used to extract the characteristics of the original signals in order to obtain sensitive characteristics from various angles. Because the residual structure is used in the model, the problem of model degradation can be avoided. Finally, a fully connected layer is introduced for SA detection. In order to overcome the impact of class imbalance, a focal loss function is introduced to replace the traditional cross-entropy loss function, which makes the model pay more attention to learning difficult samples in the training phase. Experimental results from the Apnea-ECG dataset show that the accuracy, sensitivity and specificity of the proposed multi-scale residual network are 86.0%, 84.1% and 87.1%, respectively. These results indicate that the proposed method not only achieves greater recognition accuracy than other methods, but it also effectively resolves the problem of low sensitivity caused by class imbalance.

Background: The planning and control of wind power production rely heavily on short-term wind speed forecasting. Due to the non-linearity and non-stationarity of wind, it is difficult to carry out accurate modeling and prediction through traditional wind speed forecasting models.Methods: In the paper, we combine empirical mode decomposition (EMD), feature selection (FS), support vector regression (SVR) and cross-validated lasso (LassoCV) to develop a new wind speed forecasting model, aiming to improve the prediction performance of wind speed. EMD is used to extract the intrinsic mode functions (IMFs) from the original wind speed time series to eliminate the non-stationarity in the time series. FS and SVR are combined to predict the high-frequency IMF obtained by EMD. LassoCV is used to complete the prediction of low-frequency IMF and trend.Results: Data collected from two wind stations in Michigan, USA are adopted to test the proposed combined model. Experimental results show that in multi-step wind speed forecasting, compared with the classic individual and traditional EMD-based combined models, the proposed model has better prediction performance.Conclusions: Through the proposed combined model, the wind speed forecast can be effectively improved.

Identification and detection of traffic identifiers are some of the most challenging tasks in the field of autonomous driving. For object detection, how to obtain deep semantic information and shallow spatial information has always been a permanent subject. This paper proposes a dense network structure for the detection of traffic identifiers; the deep structure of the residual network is used to obtain the high-level semantic information of the detection target; the regression operation is used to obtain the underlying spatial information; the detection structure is higher than the previous The accuracy of the network structure and the generalization ability of the network has been improved; ASPP using the attention mechanism; the main function of ASPP is to obtain a continuous feature map; using the ASPP of the attention mechanism, the prior balance between the receptive field and the feature resolution map is solved; In the paper, a collection of residual networks is used to obtain a variety of acceptance domains; through a regression operation, an effective interactive connection between the receptive field and the acceptance domain is obtained; a deep network using regression residuals allows geometric transformation of the model during data training The ability to improve has been improved; at the same time, the use of a dense network structure has realized the diversity of different receptive fields in the process of multi-scale feature extraction.