Sign language is an important way for deaf people to understand and communicate with others. Many researchers use Wi-Fi signals to recognize hand and finger gestures in a non-invasive manner. However, Wi-Fi signals usually contain signal interference, background noise, and mixed multipath noise. In this study, Wi-Fi Channel State Information (CSI) is preprocessed by singular value decomposition (SVD) to obtain the essential signals. Sign language includes the positional relationship of gestures in space and the changes of actions over time. We propose a novel dual-output two-stream convolutional neural network. It not only combines the spatial-stream network and the motion-stream network, but also effectively alleviates the backpropagation problem of the two-stream convolutional neural network (CNN) and improves its recognition accuracy. After the two stream networks are fused, an attention mechanism is applied to select the important features learned by the two-stream networks. Our method has been validated by the public dataset SignFi and adopted five-fold cross-validation. Experimental results show that SVD preprocessing can improve the performance of our dual-output two-stream network. For home, lab, and lab + home environment, the average recognition accuracy rates are 99.13%, 96.79%, and 97.08%, respectively. Compared with other methods, our method has good performance and better generalization capability.

HfO₂-based ferroelectrics have extended a familiar CMOS high-κ dielectric into a materials platform for nonvolatile memory, low-power logic, and in-memory hardware. This review examines thickness engineering in hafnia, especially Hf_0.5Zr_0.5O₂ (HZO), through the state produced during fabrication, the voltage and time available for switching, and the signal retained during operation. Representative quantitative comparisons retain stack, anneal, electrical state, and endpoint definitions; they show that conditioning and drive voltage can change the apparent thickness trend substantially. Capacitor memory, FeFETs, tunnel junctions, negative-capacitance transistors, and computing devices are compared through architecture-specific read margins and failure criteria. The main unresolved challenge is to isolate thickness effects from coupled process and interface changes and establish reproducible, device-specific operating windows.

With the rapid advancement of digital technologies and the growing emphasis on high-quality and sustainable higher education, application-oriented universities are undergoing profound digital transformation. This conceptual study develops a sustainability-oriented framework grounded in Socio-Technical Systems Theory and the Dynamic Capabilities perspective, with China’s Four Futures initiative serving as a contextual lens for identifying key transformation domains. The framework integrates four institutional dimensions: faculty development, pedagogical innovation, institutional governance, and digital resource integration. It distinguishes existing digital foundations from prospective AI-enabled extensions and illustrates selected elements through an exploratory smart laboratory governance pilot. The pilot demonstrates how shared data standards, university–college workflows, laboratory management, safety management, teaching operations, and mobile visualization can be aligned with the proposed framework. However, this study does not claim improvements in teaching quality, operational efficiency, or resource utilization. Instead, it identifies key implementation conditions related to data quality, privacy, security, model validation, human oversight, accessibility, life cycle costs, and environmental impact. The framework provides a basis for future longitudinal and multi-institutional evaluation of responsible AI-enabled digital transformation in application-oriented higher education.

Human Activity Recognition (HAR) based on Channel State Information (CSI) has attracted considerable attention as a privacy-preserving sensing paradigm. However, CSI-based HAR faces several challenges, including environmental noise, long-range temporal dependencies, and the anisotropic structure of CSI tensors. To address these challenges, this paper presents Wi-CSNet, a lightweight CSI-oriented framework that integrates Discrete Wavelet Transform (DWT) preprocessing, asymmetric-stride convolutions, and a Cross-Scanning State Space Duality (CS-SSD) block derived from Mamba2. DWT preprocessing compresses temporal signals while preserving motion-related trends and reducing input dimensionality. Asymmetric-stride convolutions balance feature scales across heterogeneous CSI dimensions, while the lightweight CS-SSD module captures global dependencies with only a 0.63% parameter overhead. Extensive experiments demonstrate that Wi-CSNet achieves accuracies of 97.81% on HHI, 99.92% on UT-HAR, and 100% on NTU-HAR. These results confirm the effectiveness and robustness of Wi-CSNet for fine-grained CSI-based activity recognition in complex environments.

In experimental pain studies involving animals, subjective pain reports are not feasible. Current methods for detecting pain-related behaviors rely on human observation, which is time-consuming and labor-intensive, particularly for lengthy video recordings. Automating the quantification of these behaviors poses substantial challenges. In this study, we developed and evaluated a deep learning, multistream algorithm to detect pain-related grooming behaviors in rats. Pain-related grooming behaviors were induced by injecting small amounts of pain-inducing chemicals into the rats’ hind limbs. Day-long video recordings were then analyzed with our algorithm, which initially filtered out non-grooming segments. The remaining segments, referred to as likely grooming clips, were used for model training and testing. Our model, a multistream recurrent convolutional network, learned to differentiate grooming from non-grooming behaviors within these clips through deep learning. The average validation accuracy across three evaluation methods was 88.5%. We further analyzed grooming statistics by comparing the duration of grooming episodes between experimental and control groups. Results demonstrated statistically significant changes in grooming behavior consistent with pain expression.