Search

IntroductionImproving the rationality and accuracy of classroom quality analysis is crucial in modern education. Traditional methods, such as questionnaires and manual recordings, are resource-intensive and subjective, leading to inconsistent results. As a solution, computer vision (CV) technologies have emerged as powerful tools for real-time classroom monitoring. This study proposes a novel Dense 3D Convolutional Residual Network (D3DCNN_ResNet) to recognize students’ expressions and behaviors in English classrooms.MethodsThe proposed method combines Single Shot Multibox Detector (SSD) for target detection with an improved D3DCNN_ResNet model. The network applies 3D convolution in both spatial and temporal domains, with shortcut connections from residual blocks to increase network depth. Dense connections are introduced to enhance the flow of high- and low-level features. The model was tested on two datasets: the CK+ dataset for expression recognition and the KTH dataset for behavior recognition.Results and DiscussionThe experiments show that the proposed method is highly efficient in optimizing model training and improving recognition accuracy. On the CK+ dataset, the model achieved an expression recognition accuracy of 97.94%, while on the KTH dataset, the behavior recognition accuracy reached 98.86%. The combination of residual blocks and dense connections reduced feature redundancy and improved gradient flow, leading to better model performance. The results demonstrate that the D3DCNN_ResNet is well-suited for classroom quality analysis and has the potential to enhance teaching strategies by providing real-time feedback on student engagement.

Most machine learning algorithms require a large set of training samples in order to achieve satisfactory performance. However, this requirement may be difficult to satisfy in practice. Take the one-shot learning (OSL) problem on texture recognition for example; the machine learning algorithm is difficult to achieve satisfactory results. In order to solve this problem, a novel multi-modal one-shot learning method for texture recognition is presented. First, in order to improve the robustness of identification and the anti-interference to noise, we addressed the nontravel texture recognition challenges of learn information about object categories from only one training sample by fusing varied modalities data, including image, sound and acceleration, which provides rich information regarding textures. Second, a novel dictionary learning model is designed, which contains the various modalities information, and can simultaneously learn the latent common sparse code for the different modalities. Third, an original regularization term is developed to enhance the degree of distinction of different classes. Furthermore, the common features of the three modalities are evaluated in the case of one-shot learning and used as the basis for feature selection. In the end, experiments were performed based on a data set which was published openly to validate the effectiveness of the presented method.

Sign language recognition aims to recognize meaningful movements of hand gestures and is a significant solution in intelligent communication between the deaf community and hearing societies. However, until now, the current dynamic sign language recognition methods still have some drawbacks with difficulties of recognizing complex hand gestures, low recognition accuracy for most dynamic sign language recognition, and potential problems in larger video sequence data training. In order to solve these issues, this paper presents a multimodal dynamic sign language recognition method based on a deep 3-dimensional residual ConvNet and bi-directional LSTM networks, which is named as BLSTM-3D residual network (B3D ResNet). This method consists of three main parts. First, the hand object is localized in the video frames in order to reduce the time and space complexity of network calculation. Then, the B3D ResNet automatically extracts the spatiotemporal features from the video sequences and establishes an intermediate score corresponding to each action in the video sequence after feature analysis. Finally, by classifying the video sequences, the dynamic sign language is accurately identified. The experiment is conducted on test datasets, including DEVISIGN_D dataset and SLR_Dataset. The results show that the proposed method can obtain state-of-the-art recognition accuracy (89.8% on the DEVISIGN_D dataset and 86.9% on SLR_Dataset). In addition, the B3D ResNet can effectively recognize complex hand gestures through larger video sequence data, and obtain high recognition accuracy for 500 vocabularies from Chinese hand sign language.

Tracking control, which is applied to the target mobile robots in the process of rushing toward the trainees, is one of the critical technologies in the advancement of anti-terrorist training. Considering the disadvantages of various types of traditional tracking methods, this paper proposes a novel laser ray tracking mechanism and a backstepping controller for the target mobile robot that is used in shooting ranges. The mechanism and principle of the laser ray tracking is illustrated in detail. Based on the unique structure, the light intensity distributions are measured to further locate the laser spots on the cut-ray boards. Then, the relationship between the positions of the laser spots on the cut-ray boards and the pose of the target mobile robot is demonstrated. According to the features of the tracking situation, a backstepping controller is designed to achieve the laser ray tracking. After that, the inverse kinematics of the wheeled skid-steering mobile robot is analyzed to map the linear and angular velocities of the robot to the velocities of its left wheels and right wheels. The conventional proportional–integral–derivative controller is applied in the experiments to compare with the proposed backstepping controller. The experimental results show that the proposed controller is more robust, and converges faster for the laser ray tracking.

Early warning on the ship deficiency is crucial for enhancing maritime safety, improving maritime traffic efficiency, reducing ship fuel consumption, etc. Previous studies focused on the ship deficiency exploration by mining the relationships between the ship physical deficiencies and the port state control (PSC) inspection results with statistical models. Less attention was paid to discovering the correlation rules among various parent ship deficiencies and subcategories. To address the issue, we proposed an improved Apriori model to explore the intrinsic mutual correlations among the ship deficiencies from the PSC inspection dataset. Four typical ship property indicators (i.e., ship type, age, deadweight and gross tonnage) were introduced to analyze the correlations for the ship parent deficiency categories and subcategories. The findings of our research can provide basic guidelines for PSC inspections to improve the ship inspection efficiency and maritime safety.

Maritime surveillance videos provide crucial on-spot kinematic traffic information (traffic volume, ship speeds, headings, etc.) for varied traffic participants (maritime regulation departments, ship crew, ship owners, etc.) which greatly benefits automated maritime situational awareness and maritime safety improvement. Conventional models heavily rely on visual ship features for the purpose of tracking ships from maritime image sequences which may contain arbitrary tracking oscillations. To address this issue, we propose an ensemble ship tracking framework with a multi-view learning algorithm and wavelet filter model. First, the proposed model samples ship candidates with a particle filter following the sequential importance sampling rule. Second, we propose a multi-view learning algorithm to obtain raw ship tracking results in two steps: extracting a group of distinct ship contour relevant features (i.e., Laplacian of Gaussian, local binary pattern, Gabor filter, histogram of oriented gradient, and canny descriptors) and learning high-level intrinsic ship features by jointly exploiting underlying relationships shared by each type of ship contour features. Third, with the help of the wavelet filter, we performed a data quality control procedure to identify abnormal oscillations in the ship positions which were further corrected to generate the final ship tracking results. We demonstrate the proposed ship tracker’s performance on typical maritime traffic scenarios through four maritime surveillance videos.

It is difficult to achieve ideal virtual surgery transparency and stability when virtual tissue stiffness and damping are high. Typically, the stability of the surgery system is improved, while its transparency is sacrificed. In order to achieve high transparency in virtual surgical interactions, a maximum output force controller based on passive theory is proposed in this work. This controller is then applied in a virtual surgery system. The maximum output force controller predicts the maximum allowable output force above which the system passivity is broken and limits the force presented to the operator to this amount. The main contributions of this work include the following two parts: firstly, the maximum output force controller is developed and applied to a virtual surgery system; secondly, a new criterion for transparency is presented and analyzed for the level of transparency that can be achieved for a virtual surgical system when the stability is guaranteed. Experimental results show that the maximum output force controller can guarantee stability of the virtual surgical interaction with maximum transparency even when the virtual tissue stiffness and damping are high. In addition, the maximum output force controller is a self-adaptive controller. It works well without modification, regardless of the virtual tissue stiffness and damping.

Herein, the problem of target tracking in wireless sensor networks (WSNs) is investigated in the presence of Byzantine attacks. More specifically, we analyze the impact of Byzantine attacks on the performance of a tracking system. First, under the condition of jointly estimating the target state and the attack parameters, the posterior Cramer−Rao lower bound (PCRLB) is calculated. Then, from the perspective of attackers, we define the optimal Byzantine attack and theoretically find a way to achieve such an attack with minimal cost. When the attacked nodes are correctly identified by the fusion center (FC), we further define the suboptimal Byzantine attack and also find a way to realize such an attack. Finally, in order to alleviate the negative impact of attackers on the system performance, a modified sampling importance resampling (SIR) filter is proposed. Simulation results show that the tracking results of the modified SIR filter can be close to the true trajectory of the moving target. In addition, when the quantization level increases, both the security performance and the estimation performance of the tracking system are improved.

In order to improve the precision and stability of puncture surgical operations to assist doctors in completing fine manipulation, a new of type puncturing needle sensor is proposed based on a fiber Bragg grating (FBG). Compared with the traditional puncture needle sensor, the new type of puncturing needle sensor is able to sense not only the axial force, but also the torque force during the puncture process. A spoke-type structure is designed near the needle tip. In order to eliminate the influence of temperature and realize temperature compensation, a reference fiber method using three FBGs is applied. FBG1 and the reference FBG2 are pasted on the upper and lower surfaces of the new-type elastic beam, and FBG3 is pasted into the groove on the surface of the new type of puncturing needle cylinder. The difference of Bragg wavelength between FBG1 and the reference FBG2 is calibrated with the torque force, while the difference between the Bragg wavelength of the FBG3 and the reference FBG2 is calibrated with the axial force. Through simulation and sensing tests, when the torque force calibration range is 10 mN·m, the torque average sensitivity is 22.8 pm/mN·m, and the determination coefficient R2 is 0.99992, with a hysteresis error YH and repetition error YR of 0.03%FS and 0.81%FS, respectively. When the axial force calibration rang is 5 N, the axial force average sensitivity is 0.089 nm/N, and the determination coefficient R2 is 0.9997, with hysteresis error YH and repetition error YR of 0.014%FS and 0.11%FS, respectively. The axial force resolution and torque resolution of the new type of puncturing needle sensor are 0.03 N and 0.8 mN·m, respectively. The experimental data and simulation analysis show that the proposed new type of puncturing needle sensor has good practicability and versatility.

Telerobot system plays an important role in executing task under hazard environment. As the computer networks such as the Internet are being used as the communication channel of telerobot systems, varying time delay causes the overall system unstable and reduces the performance of transparency. In this paper, we propose twelve operation modes with different control schemes for telerobot on the Internet with time delay. And an optimal operation mode with control scheme is specified for telerobot with time delay, based on the tradeoff between passivity and transparency properties. We experimentally confirm the validity of the proposed optimal mode and control scheme by using a simple one DOF master–slave manipulator system.