Fang Peng

Classifying respiratory sounds is essential for the early diagnosis of respiratory diseases. However, the computational cost of existing computer-aided respiratory sound analysis approaches limits their practical deployment in real-time clinical environments. This study presents embedded PANNs-based intelligent auscultation with XGBoost (EPIX), an intelligent embedded auscultation instrumentation system that integrates an electronic stethoscope interface, a PANNs-based acoustic feature extraction module, an optimized eXtreme gradient boosting (XGBoost) classifier, and a two-stage decision mechanism for reliable real-time respiratory sound classification and measurement. For hardware implementation, EPIX integrates a Raspberry Pi 5 and a dedicated expansion module, enabling real-time inference for electronic stethoscopes and batch inference through an external USB drive, providing flexible processing capabilities. EPIX achieved an SC score of 0.7618 on the SPRSound dataset, while significantly reducing computational resource requirements. These capabilities make EPIX well-suited for embedded measurement scenarios with limited computational resources. The processing workflow on the device provides consistent measurement performance and enables practical use in portable and bedside intelligent auscultation instruments.

With the advancement of deep learning algorithms, semantic segmentation algorithms for remote sensing images (RSIs) have been widely applied. Segment Anything Model (SAM), as a universal image segmentation model, has been utilized in segmentation tasks. However, SAM’s training on natural images results in a lack of understanding of the general information specific to RSIs. Mamba, as a sequence modeling method grounded in the state space model, can directly capture the global dependencies inherent in RSIs. Its integration with SAM significantly enhances segmentation accuracy in complex scenes. Therefore, SFE-SAM is proposed to enhance Mobile SAM’s capability to comprehend complex scenes and capture global contextual information by introducing Mamba. In SFE-SAM, the SSM Tokening Adapter (STA) is employed to fine-tune the extractor of Mobile SAM, establishing long-range feature correlations. Meanwhile, the Dual Feature Enhancement Module (DFEM) is designed to fuse cross-dimensional features. Finally, the Semantic Information Enhance Module (SIEM) is utilized to improve multi-scale information extraction and dynamic feature enhancement. Experimental results demonstrate that SFE-SAM achieves superior segmentation performance compared to SOTA methods on the LandCover.ai and Vaihingen datasets.

To enhance robotic adaptability in dynamic environments, this study proposes a multimodal framework for skill transfer. The framework integrates vision-based kinesthetic teaching with surface electromyography (sEMG) signals to estimate human impedance. We establish a Cartesian-space model of upper-limb stiffness, linearly mapping sEMG signals to end-point stiffness. For flexible task execution, dynamic movement primitives (DMPs) generalize learned skills across varying scenarios. An adaptive admittance controller, incorporating sEMG-modulated stiffness, is developed and validated on a UR5 robot. Experiments involving elastic-band stretching demonstrate that the system successfully transfers human impedance characteristics to the robot, enhancing stability, environmental adaptability, and safety during physical interaction.

To enhance robotic adaptability in dynamic environments, this study proposes a multimodal framework for skill transfer. The framework integrates vision-based kinesthetic teaching with surface electromyography (sEMG) signals to estimate human impedance. We establish a Cartesian-space model of upper-limb stiffness, linearly mapping sEMG signals to endpoint stiffness. For flexible task execution, dynamic movement primitives (DMPs) generalize learned skills across varying scenarios. An adaptive admittance controller, incorporating sEMG-modulated stiffness, is developed and validated on a UR5 robot. Experiments involving elastic band stretching demonstrate that the system successfully transfers human impedance characteristics to the robot, enhancing stability, environmental adaptability, and safety during physical interaction.

Coordination engineering strategy for optimizing the catalytic performance of single‐atom catalysts (SACs) has been rapidly developed over the last decade. However, previous reports on copper SACs for nitrate reduction reactions (NO₃RR) have mostly focused on symmetric coordination configurations such as Cu‐N₄ and Cu‐N₃. In addition, the mechanism in terms of the regulation of coordination environment and catalytic properties of SACs has not been well demonstrated. Herein, we disrupted the local symmetric structure of copper atoms by introducing unsaturated heteroatomic coordination of Cu−O and Cu−N to achieve the coordination desymmetrization of Cu‐N₁O₂ SACs. The Cu‐N₁O₂ SACs exhibit an efficient nitrate‐to‐ammonia conversion with a high FE of ~96.5 % and a yield rate of 3120 μg NH₃ h⁻¹ cm⁻² at −0.60 V vs RHE. As indicated by in situ Raman spectra, the catalysts facilitate the accumulation of NO₃⁻ and the selective adsorption of *NO₂, which were further confirmed by the theoretical study of surface dipole moment and orbital hybridization. Our work illustrated the correlation between the coordination desymmetrization and the catalytic performance of copper SACs for NO₃RR.

Coordination engineering strategy for optimizing the catalytic performance of single‐atom catalysts (SACs) has been rapidly developed over the last decade. However, previous reports on copper SACs for nitrate reduction reactions (NO₃RR) have mostly focused on symmetric coordination configurations such as Cu‐N₄ and Cu‐N₃. In addition, the mechanism in terms of the regulation of coordination environment and catalytic properties of SACs has not been well demonstrated. Herein, we disrupted the local symmetric structure of copper atoms by introducing unsaturated heteroatomic coordination of Cu−O and Cu−N to achieve the coordination desymmetrization of Cu‐N₁O₂ SACs. The Cu‐N₁O₂ SACs exhibit an efficient nitrate‐to‐ammonia conversion with a high FE of ~96.5 % and a yield rate of 3120 μg NH₃ h⁻¹ cm⁻² at −0.60 V vs RHE. As indicated by in situ Raman spectra, the catalysts facilitate the accumulation of NO₃⁻ and the selective adsorption of *NO₂, which were further confirmed by the theoretical study of surface dipole moment and orbital hybridization. Our work illustrated the correlation between the coordination desymmetrization and the catalytic performance of copper SACs for NO₃RR.

Electron-rich Rh sites of CuRh NSs enable the selective adsorption of *NH₂OH, thus achieving high FE and yield rate of ammonia.

Tracking and grasping moving objects are hot topics in the field of robots, which provides great potential in the industrial scene and human–computer cooperation. Based on the kernel correlation filter and vision 3-D reconstruction, this article proposes a visual-based tracking and grasping method for moving targets. An improved algorithm based on the kernel correlation filter is proposed for object tracking. A scale pool is constructed, and a scale filter is trained to solve the problem of algorithm scale adaptation. At the same time, the judgment mechanism of tracking results, secondary detection, and modification update mechanisms are added to improve the robustness of the algorithm. Combined with an RGB-D camera, the target is reconstructed to obtain the 3-D pose of the target. The tracking and intercepting strategy is adopted to grasp the moving target. The proposed method is proven to have good performance through dataset comparison tests and experiments on real robot systems.

This topic studies how to use the sEMG signal portable external equipment to collect the surface EMG signal using the EMG instrument. In the signal data processing part, the causes of signal interference are analyzed from the hardware perspective, and the main frequency bands of its occurrence are recorded. In the code part, the data is processed through an analog filter. The communication mode uses TCP/IP protocol to achieve cross system interprocess communication. In the actual test process, there is unavoidable slight delay. At the same time, the voltage changes brought by the real-time adjustment mode of CPU and other computer hardware will also affect the accuracy of surface EMG signals. At the same time, the signal acquisition rate of the EMG is much higher than the transmission rate, so real-time transmission will lead to memory overflow. Therefore, this paper uses the fixed threshold algorithm as the control decision in the outgoing process, and tests the integral mean value of six different data sets and the grasping speed and reaction force of the gripper. It is finally found that when the data set has 20 elements, the delay is not large and the data is not easy to be blocked due to good real-time performance.