Search

In precision medicine and clinical pain management, the creation of quantitative, objective indicators to assess somatosensory sensitivity was essential. This study proposed a fusion approach for decoding human somatosensory sensitivity, which combined multimodal (quantitative sensory test and neurophysiology) features to classify the dataset on individual somatosensory sensitivity and reveal distinct types of brain activation patterns. Sixty healthy participants took part in the experiment on somatosensory sensitivity that implemented cold, heat, mechanical punctate, and pressure stimuli, and the resting-state electroencephalography (EEG) was collected using BrainVision. The quantitative sensory testing (QST) scores of the participants were clustered using the unsupervised k-means algorithm into four subgroups: generally hypersensitive (HS), generally non-sensitive (NS), predominantly thermally sensitive (TS), and predominantly mechanically sensitive (MS). Furthermore, two types of power spectral density (PSD), band-based PSD (BB-PSD) and frequency-based PSD (FB-PSD), and two types of inter-electrode connectivity (IEC), band-based connectivity (BBC) and frequency-based connectivity (FBC), derived from resting-state EEG were subjected to feature selection with a proposed prior-compared minimum-redundancy maximum-relevance (PCMRMR) protocol. Their effectiveness was then tested by the supervised classification tasks using support vector machine (SVM), k-nearest neighbor (kNN), random forest (RF), and Gaussian classifier (GC). Brain networks of four somatosensory types were revealed by decoding fused multimodal data, namely type-averaged connectivity. The data from sixty healthy individuals were divided into training (n =59) and validation (n =1) datasets according to leave-one-subject-out (LOSO) criteria. The FBC was identified, which can serve as better brain signatures than BB-PSD, FB-PSD, and BBC to classify subjects as HS, NS, TS, or MS groups. Using the SVM, kNN, RF, and GC models, the best accuracy of 87% was obtained when classifying participants into HS, NS, TS, or MS groups. Moreover, the brain networks were decoded from HS, NS, TS, and MS groups by decoding the type-averaged connectivity fused from somatosensory phenotypes and selected FBC. It indicated that quantified multi-parameter somatosensory sensitivity could be achieved with acceptable accuracy, leading to considerable possibilities for using objective pain perception evaluation in clinical practice.

Three-phase, three-level Vienna rectifiers are widely used in power supplies for the advantages of high power density, low switching voltage stress and low grid side current harmonics. However, due to the presence of positive and negative sequence components under the unbalanced weak power grid, the control structure using the traditional proportional integral (PI) control is complicated and requires decoupling and multiple coordinate transformations, and it has longer dynamic response time. This paper presents an unbalanced current control strategy that combines improved sliding mode control (SMC) with new reaching law and multi proportional resonance (MPR) control of Vienna rectifier under unbalanced weak power grid. This paper establishes the mathematical model of Vienna rectifier under unbalanced weak current network. In the $\alpha \beta $ coordinate system, dual second-order generalized integrator (DSOGI) is used to separate the positive and negative sequence components of the voltage. Suitable sliding mode surfaces and sliding mode convergence laws were designed. Finally, the results of simulations and experiments were used to verify the effectiveness of the proposed method.

In an automobile, the door panel type-B constitutes the interior compartment of the door, primarily composed of screws and white installations forming its structural framework. However, automated manufacturing and maintenance procedures often struggle to accurately detect these components due to their pronounced resemblance to other elements on the panel. Computer vision techniques present a viable solution to this challenge. In this paper, we propose the Hyper-Progressive Single Shot Detector (HPSSD), an object detection algorithm designed to address the aforementioned challenge. Our proposed HPSSD builds on the Single Shot Detector (SSD) algorithm and introduces several enhancements to improve its detection capabilities. The first modification involves replacing the VGG-16 backbone with a ResNet-50 module. Furthermore, we incorporated the Residual Convolutional Block Attention Mechanism (RCBAM) to boost the algorithm’s functionality. To enlarge the receptive fields of each pixel-an essential step for enhancing detection accuracy-we executed multi-dilated convolutions. In the final stage of our development process, we embedded a three-stage progressive attention mechanism (PAM). The PAM is instrumental in generating refined feature maps, which serve as the foundation for precise object detection on the door panel dataset comprising 1200 images. After running 50k iterations on the door panel dataset, the HPSSD displayed a promising mean average precision of 98.2% at a speed of 21 frames per second (FPS). Our results suggest that the HPSSD, with its ability to deliver real-time, accurate detection, is an ideal tool for improving the quality inspection of door panels in smart factories.

This paper reports the demonstration of a high performance 4H-SiC floating junction junction barrier Schottky (FJ_JBS) rectifier with a 30μm, 6×1015 cm-3-doped epitaxial layer. Extensive simulations have been performed to design, optimize and analyze the structure of the FJ_JBS rectifier. The fabricated FJ_JBS shows that breakdown voltage (BV) and differential Ron,sp are 3.4 kV, yielding the highest BV value reported for 4H-SiC FJ diodes, and 5.67 mΩ·cm2, respectively. Compared with the conventional JBS, the BV value of FJ_JBS increases by 33.3% and the Ron,sp only slightly rises by 6.2%. The corresponding Baliga figure-of-merit (BFOM) (4 BV2/Ron_sp) of this FJ_JBS diode is 8.16 GW/cm2.

In this paper, an intelligent adaptive jerk control (IAJC) with dynamic compensation gain for the permanent magnet linear synchronous motor (PMLSM) servo system was proposed to improve robustness and tracking performance against nonlinear and time-varying uncertainties. First, the dynamic model of the PMLSM servo system was investigated. Subsequently, the model-based feedforward control was designed for parametric uncertainties. Then, an adaptive jerk control (AJC) was adopted to restrain external load disturbance, nonlinear friction and unmodeled dynamics of the servo system. The adaptive feedback gain of jerk was updated by an exponential function. However, the uncertainties of the PMLSM servo system were unavailable in advance, it was difficult to design the adaptive feedback gain in practice. Thus, in the following part, the IAJC was further developed in which a dynamic compensation gain was designed using a double-loop recurrent feature selection fuzzy neural network (RFSFNN) to compensate for approximation deviation and suppress the chattering phenomenon. The learning algorithms of the double-loop RFSFNN were derived and the stability of the closed-loop system was proved by the Lyapunov approach. Finally, the experimental results demonstrate that the proposed IAC scheme can achieve robust precise tracking performance.

A stable and simple Yb:YAG rod amplifier based on chirped pulse amplification in a water cooling system has been demonstrated. The output power of 38 W at a 250 kHz repetition rate with a spectrum width of 4 nm centered at 1030 nm, which support Fourier-transform limited pulse duration of 389 fs has been generated. The compressed power of 23.9 W with pulse duration of 985 fs has been obtained with a chirped volume Bragg grating compressor. The power stability is measured to be 1.414% when the output power is 35 W. The system is worthwhile to popularize for its good performance and inexpensive cost.

As a nonlinear feature learning method, deep canonical correlation analysis (DCCA) has got a great success in computer vision. Compared with kernel methods, deep neural networks can more easily process large amounts of training data and do not require referring to the training set at test time. However, in the real world, due to the noise disturbance and the limited number of training samples, within-set and between-set sample covariance matrices cannot usually be estimated accurately, which causes that the gradient direction deviates from the true one when training DCCA. It incorporates fractional-order within-set and between-set scatter matrices to reduce the deviations of sample covariance matrices for gradient direction correction. In addition, to make full use of convolutional network's feature extraction ability and fractional model in modifying neural network gradient directions, we further propose two novel convolutional network-based FDCCA methods, named convolutional neural network-based FDCCA (CNN-FDCCA) and two-convolutional neural networks based FDCCA (2CNNs-FDCCA), respectively. The experimental results on MNIST and RAVDNESS datasets show that FDCCA has better recognition rates than existing methods. The experiments on AT&T dataset show that CNN-FDCCA and 2CNNs-FDCCA have great robustness in processing images.

As the representative one of representation-based classification (RBC) methods, collaborative RBC (CRC) has drawn much attention in pattern recognition and machine learning recently. Moreover, the collaborative representation-based face recognition has been extensively studied because of the effective classification performance of CRC. CRC collaboratively represents each query sample as the linear combination of all the training samples and then classifies the query sample according to the categorical representation-based distances. However, most variants of CRC cannot fully consider the locality and discrimination of data and cannot well handle the noise data, which has negative effect on real-world classification problems, such as face recognition. In this paper, a new discriminative collaborative neighbor representation (DCNR) method for face recognition is proposed by integrating class discrimination and data locality. In the proposed method, the locality of data constrains collaborative representation of each query sample to find representative nearest samples of the query sample. Moreover, the class discrimination regularization is taken into account by employing the representation of each class for each query sample. Due to the existing noises, such as corruptions and occlusions in face recognition, we further propose robust DCNR (R-DCNR) for robust classification by using the ℓ1-norm representation fidelity. Extensive experiments on face databases demonstrate that the proposed methods achieve competitive classification performance, compared to the state-of-the-art representation-based classification methods.