High-density surface electromyography (HD-sEMG) and deep learning technology are becoming increasingly used in gesture recognition. Based on electrode grid data, information can be extracted in the form of images that are generated with instant values of multi-channel sEMG signals. In previous studies, image-based, two-dimensional convolutional neural networks (2D CNNs) have been applied in order to recognize patterns in the electrical activity of muscles from an instantaneous image. However, 2D CNNs with 2D kernels are unable to handle a sequence of images that carry information concerning how the instantaneous image evolves with time. This paper presents a 3D CNN with 3D kernels to capture both spatial and temporal structures from sequential sEMG images and investigates its performance on HD-sEMG-based gesture recognition in comparison to the 2D CNN. Extensive experiments were carried out on two benchmark datasets (i.e., CapgMyo DB-a and CSL-HDEMG). The results show that, where the same network architecture is used, 3D CNN can achieve a better performance than 2D CNN, especially for CSL-HDEMG, which contains the dynamic part of finger movement. For CapgMyo DB-a, the accuracy of 3D CNN was 1% higher than 2D CNN when the recognition window length was equal to 40 ms, and was 1.5% higher when equal to 150 ms. For CSL-HDEMG, the accuracies of 3D CNN were 15.3% and 18.6% higher than 2D CNN when the window length was equal to 40 ms and 150 ms, respectively. Furthermore, 3D CNN achieves a competitive performance in comparison to the baseline methods.

At present, in the field of video-based human action recognition, deep neural networks are mainly divided into two branches: the 2D convolutional neural network (CNN) and 3D CNN. However, 2D CNN’s temporal and spatial feature extraction processes are independent of each other, which means that it is easy to ignore the internal connection, affecting the performance of recognition. Although 3D CNN can extract the temporal and spatial features of the video sequence at the same time, the parameters of the 3D model increase exponentially, resulting in the model being difficult to train and transfer. To solve this problem, this article is based on 3D CNN combined with a residual structure and attention mechanism to improve the existing 3D CNN model, and we propose two types of human action recognition models (the Residual 3D Network (R3D) and Attention Residual 3D Network (AR3D)). Firstly, in this article, we propose a shallow feature extraction module and improve the ordinary 3D residual structure, which reduces the parameters and strengthens the extraction of temporal features. Secondly, we explore the application of the attention mechanism in human action recognition and design a 3D spatio-temporal attention mechanism module to strengthen the extraction of global features of human action. Finally, in order to make full use of the residual structure and attention mechanism, an Attention Residual 3D Network (AR3D) is proposed, and its two fusion strategies and corresponding model structure (AR3D_V1, AR3D_V2) are introduced in detail. Experiments show that the fused structure shows different degrees of performance improvement compared to a single structure.

Step length estimation is essential for people to overcome different types of environmental obstacles in daily life. Aiming at the problem of low estimation accuracy based solely on surface electromyography (sEMG), in this paper, a multi-source (MS) data fusion method based on auto-encoder-convolutional neural network (AE-CNN) is proposed to estimate the motion parameters of human lower limbs, which improve the accuracy of step length estimation. Specifically, the time-domain feature data of sEMG and insole pressure data from the lower limbs are first collected simultaneously to form the MS. Then, by applying AE as the fusion units, fused data are obtained and dimensionality reduction is realized. Finally, by integrating AE and CNN, the step length can be well estimated by utilizing the MS. For experimental validation, the performances of different data fusion units and regression units in step length estimation are compared. The results show that the proposed method can achieve good estimation. The normalized root mean square error (NRMSE) and Pearson correlation coefficient (PCC) of proposed method reach 0.0479±0.0263 and 0.8273±0.1082, respectively. The NRMSE of the best trial is 0.0217, the PCC is 0.9774. Further, experiments at different walking speeds are carried out. The results show that the combination of AE and CNN has good performance with the increase of walking speed. In addition, a dimensionality reduction experiment based on AE is also carried out, and the optimal dimension is obtained.

The restoration of walking capability is very important after central nervous system (CNS) injury. Rehabilitation training robot can replace therapists to help patients do walking training by accurately simulating the human lower limb movement, and achieve the purpose of rehabilitation ultimately. In order to further improve the rehabilitation robots' intelligent and rehabilitation training effect and completely replace therapists to some extent, a new design method is proposed in this paper: firstly, the pleasant mechanical structure design method for the rehabilitation robot is put forward. Then, the rehabilitation robot's perceptual system is designed, which includes the detection methods of bioelectric and physical signals and the effective integration of the sensor network based on multi-agent system are presented. Finally, the superiority of the rehabilitation robot designed in this paper is described through results analysis.

It is imperative to have high adaptive techniques for sensing and manipulating biological targets at the nanoscale. This necessity becomes particularly crucial when dealing with fragile living bio-organisms like viruses, where the expression of capsids is closely linked to viral functions and genome constitution. Therefore, the development of a comprehensive system for dissecting and measuring viruses holds significant implications for the pharmaceutical industry and drug manufacturing. Leveraging the sub-nanometer spatial resolution and controllable tip-cantilever architecture of atomic force microscopy (AFM), a probe-laser system has been integrated as a self-sensing robotic end effector. To address intrinsic challenges in AFM-based robotic systems such as the lack of real-time monitoring, low scanning rates, and nonlinear motion caused by piezoelectric actuators, an augmented reality robotic system has been implemented. This system incorporates stereoscopic vision, a haptic feedback controller, a position recovery scheme, and a real-time force control algorithm. The integration of these components enhances the system’s capability to accurately dissect virus capsids. Operators can now perform highly efficient nanoscale tasks with multidimensional perception, utilizing the combination of stereoscopic vision and haptic force control. The position correction during manipulation can achieve a frame rate of over 30 frames per second, imperceptible to the operator, enabling closed-loop operation control. By adopting the proposed nanorobotic system in virology studies, it becomes possible to achieve accurate manipulation and dissection of SARS-CoV-2 pseudovirus capsids, and derive multi-parametric properties such as structural integrity, protein fragment thickness, and adhesive forces. The established nanobot system and experimental results serve as a guiding platform for high-accuracy evaluation in drug manufacturing development.

This paper presents an analytical model for calculating the detailed air-gap magnetic field distribution in interior permanent magnet synchronous motors (IPMSMs) with mixed eccentricity. In order to improve the efficiency of model solving, a modeling strategy combining the equivalent magnetic circuit network method and the subdomain method was adopted. Specifically, the magnetic field distribution of the rotor was modeled by using the magnetic circuit analysis method, and the magnetic field distributions of the air-gap, slot opening and stator slot along the radial direction were modeled in different regions according to their structure ruler. Then, the influence of bridge saturation was considered. Moreover, based on the analysis of the air-gap geometric structure with mixed eccentricity, a detailed spatiotemporal analytical model of the air-gap magnetic field was established, which provides a more accurate description of the mixed eccentricity composed of static and dynamic rotor eccentricities of different severity. The analytical models were compared with the corresponding models established by the finite element method, which proved the accuracy and validity of the models established in this paper. Finally, some key features related to radial and tangential air-gap flux density were extracted, which can significantly reflect the characteristics of eccentricities. The main findings reported in this paper will be of benefit for developing methods for early identification and diagnosis of eccentricity faults in IPMSMs.