Search

Depression is a common mental illness with widespread effects worldwide. At present, most relevant studies are performed by using free-view experiments to identify depression based on eye movements, which identify depression by analyzing the fixation patterns of subjects. However, technical and cost barriers make the rapid and objective detection of depression challenging. This study aimed to develop an objective and convenient method that utilizes eye tracking technology to detect depression. In this study, we designed an eye-tracking algorithm to localize pupil position and presented a variance entropy method to measure the fluctuating characteristics of eye movements. Video data were collected from 115 participants and 35 controls (mean age 24.99 years, standard deviation 6.14 years). The pupil position time series data were formed by the eye-tracking algorithm, and the variance entropy of the pupil position time series was calculated. Based on eye movement variance entropy, we used three machine learning models for classification detection: Logistic Regression, K-Nearest Neighbors, and Random Forest. The accuracy of machine learning in identifying whether a person had depression or not was 97.5%. The accuracy of the four class depression classification (severe, moderate, mild, and no depression) was 82.5%.The accuracy of identifying whether a person had anxiety or not was 95%. The accuracy of identifying whether a person had suicidal tendencies or not was 85%. The results showed that variance entropy based on the eye movement time series can be used as a valid bioindicator for detecting depression.

Power loads usually have multiple cycles, and the traditional forecasting methods only use the historical loads under various types of cycles together as input features to construct the forecasting model, ignoring the deep features under multiple cycles. The simultaneous inputs will lead to the cycles overlapping and influencing each other, which will be difficult to deal with when building the model. Therefore, this paper proposes a GRO-Bagging day-ahead power curve forecasting model based on multi-cycle feature extraction. First, the multi-periodicity of the power load is analyzed. The one-dimensional time series is converted to two-dimensional according to the multiple cycles of power load. Then, the multi-periodic feature extraction is performed by a multi-size convolutional feature extraction layer with parallel selected based on the data characteristics and modeling mechanism, and Bootstrap Aggregating (Bagging) method is used to construct different prediction models for the datasets containing different periodic features; finally, Gold Rush Optimization (GRO) algorithm is introduced and improved by using the Tent chaotic mapping and elite strategy, and the improved optimization algorithm is used for the weight optimization of the model, the error feedback mechanism is introduced to achieve the weight dynamic Updating. To prove the superiority of the proposed model, a series of comparison experiments and ablation experiments are carried out on real datasets, and the results show that the proposed method has higher prediction accuracy, and prove that the multi-period feature extraction and dynamic weighting methods have a positive and active effect on load prediction.

The number of clever facial recognition systems has been growing as artificial intelligence and robotics have advanced. However, due to the limited collection of biometric features by intelligent facial recognition systems compared to authentication methods such as iris and fingerprint, there are errors in the recognition process, low recognition accuracy, and low operational efficiency. To improve laboratory security and the efficiency of intelligent facial recognition systems, a study was conducted to extract facial feature information through facial expression key points, and a spatiotemporal graph convolutional network fused with attention mechanism was used to organize and match feature data. Finally, an improved facial recognition system was designed. A simulation experiment was conducted, and the results showed that the intelligent laboratory facial recognition system constructed by the research had a facial recognition accuracy of 89% in different datasets. Compared with the Line Hausdorff distance facial recognition method, the facial feature data was more concentrated, proving that the facial recognition system designed by the research has high recognition accuracy, wide application range, and strong application value, It helps to improve the optimization of facial recognition systems in the laboratory and improve the operational efficiency of the laboratory.

A new design for a quasi-endfire spoof surface plasmon polariton (SSPP) leaky-wave antenna (LWA) is presented for wearable application. The antenna consists of an ultra-thin corrugated metallic structure screen-printed on a flexible textile substrate, which supports extremely confined spoof surface plasmon polaritons. To enable a highly directional leaky mode, two unit-cell designs with different surface impedances are incorporated to realize binary perturbations on the in-plane wavenumber. An auto-adaptive multi-objective optimizer (MOO) is utilized to intelligently design the surface impedance configuration, which achieves significant dimensional reduction compared to the periodically modified SSPP LWAs. A final miniaturized version with 28-unit-cells achieved about 70% size reduction in comparison to the longer design of 75 unit-cells. For proof of concept, the antenna is designed and optimized for operation at 6 GHz. A bandwidth of >200 MHz (5.90 GHz - 6.13 GHz) is achieved, centered around 6 GHz, for which the highly directional endfire pattern can be tilted to 22° and 14° for the 28 and 75 unit-call designs, respectively. The measured results agree well with the simulations. Meanwhile, experimental results show that the Specific Absorption Rate (SAR) is lower than 1.6 W/kg standard when the antenna is 2 mm away from the human phantom. This textile-based antenna realized with advanced screen-printing technology is extremely suitable for garment integration due to its high flexibility, low-profile, good fabrication accuracy, and robustness in its performance.

This paper focuses on developing the acoustic-based detection of arc magnet internal defects under different noises and object types. The major challenge in such a detection case is that the traditional approaches mainly rely on expert knowledge and experiential features, which results in a poor generalization performance of the algorithm for new types of detection objects and is susceptible to noise interference. This work presents a novel detection framework based on an end-to-end one-dimensional (1D) convolutional long short-term memory (LSTM) model, where both the spatial and temporal features of the measured acoustic signals are extracted and then jointed for determining the internal defects of arc magnets. In addition, the LSTM layers are employed behind the 1D convolutional neural network, which makes the number of time steps in the LSTM layers for the temporal feature extraction is much smaller than the length of the input segments, thus the computational complexity of the LSTM layers can be highly reduced. Experimental results show that our method is superior to existed methods in the detection accuracy for the internal defects of arc magnets, and the diagnosis time per a single arc magnet is controlled at the millisecond, making it appropriate for real-time applications. Furthermore, the robustness of the proposed framework is validated through experiments on different signal-to-ratios and multiple object types of arc magnets.

Great progress has been made in bifurcation control of systems described by ordinary differential equations. However, the control of Hopf bifurcation and Turing patterns is seldom reported in reaction-diffusion systems, which is formed by partial differential equations. In this paper, a hybrid control synthesis combining state feedback is firstly devised in the reaction-diffusion marine planktonic ecosystem. The Turing instability condition and Hopf bifurcation criterion are derived through carrying out the eigenvalue analysis of the controlled system. The numerical simulations show that the hybrid control strategy can not only suppress the formation of Turing patterns, but also delay or advance the Hopf bifurcation point. Therefore, the desired spatial dynamics behaviors can be generated by manipulate the control gain parameters, so as to achieve the purpose of maintaining the marine ecological balance.

Achieving accurate navigation and localization is crucial for Autonomous Underwater Vehicle (AUV). Traditional navigation algorithms, such as Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF), require the system model and measurement model for state estimation to obtain the AUV position. However, this may introduce modeling errors and state estimation errors which will affect the final precision of AUV navigation system to a certain extent. To avoid these problems, in this paper, we proposed a deep framework - NavNet - by taking AUV navigation as a deep sequential learning problem. Firstly, the proposed NavNet can take raw sensor data at different frequencies as input, which benefits from the sequential learning capability of Recurrent Neural Network (RNN). Secondly, NavNet takes advantage of a simplified attention mechanism and Fully Connected (FC) layers to output AUV displacements per unit time, which accomplishes low-frequency AUV navigation by accumulation of it. More importantly, there is no need for the model building and state estimation with NavNet, which avoids the import of relevant errors. We compare the performance of NavNet to EKF and UKF using collected data by running Sailfish in the sea. Experimental results show that NavNet has an excellent performance in terms of both the navigation accuracy and fault tolerance. In addition, a reliable fusion strategy of NavNet and conventional method is applied to achieve high-frequency AUV navigation. The experimental results show that the proposed architecture can be a reliable supplement to limit the error growth of conventional algorithms.

Objective: Laryngopharyngeal Reflux Disease (LPRD) is prevalent and has a range of symptoms. However, diagnosing LPRD is difficult because of the lack of specific symptoms or clinical gold standard. An objective and reliable test, which does not rely on a perfect clinical gold standard, is required in clinical practice. Methods: 60 normal volunteers and 74 confirmed Laryngopharyngeal Reflux (LPR) patients were labelled based on the combined consideration of the reflux symptom index, reflux finding score, 24h oropharyngeal pH monitoring and results of anti-reflux treatment. 72 candidate features were extracted from pH recordings and the most efficient feature combination was detected using a stepwise wrapper method. The labelled data were combined with 1552 unlabelled data for feature selection and model training using semi-supervised learning. A latent class model method was used to assess the proposed model based on 64 additional validation data and an imperfect clinical reference test. Results: A new score (named W score), which significantly improved the sensitivity of the LPRD test (82.67% vs. 24.09%) and has a relatively high specificity (80.19%), was proposed. W score concurs with the complicated clinical test. Conclusion: W score which significantly improves LPRD diagnostic efficiency, can aid the clinical diagnosis of LPRD. W score provides an objective, efficient, and reliable indicator for the application of anti-reflux treatment in clinical practice. Significance: More LPRD patients can benefit from anti-reflux treatment in clinical practice and fewer patients may suffer from, for example, the side effects of unnecessary long-term Proton Pump Inhibitors (PPI) treatments.

Assessment of lower limb motion function is crucial for clinicians to treat patients effectively. However, diagnosis accuracy is still affected by clinical experience. To reduce the influence of subjective factors, a new classification model using conditional deep belief network (CDBN) and long short-term memory (LSTM) is proposed. Firstly, the original data of gait are collected by motion capture system, and low-rank matrix completion is utilized to reconstruct missing data caused by the interruption of hands and clothes. Then CDBN is trained for extracting features and LSTM classifies the features from the completed data. Compared with deep belief network (DBN), CDBN is based on conditional restricted Boltzmann machine (CRBM) which improves the importance of past time-step to achieve higher accuracy for temporal sequences. And LSTM classifies features with time information. The matrix completion result verifies excellent recovery accuracy of the proposed method. Experimental results demonstrate that CDBN-LSTM is 4% higher than that of DBN. In addition, the proposed model outperforms traditional neural network on the train process and precision rate.

With Nd:CTGS crystal as self-frequency-doubling (SFD) medium, and Cr4+:YAG crystal as passively Q-switcher, a pulsed microchip green laser was demonstrated. The maximum average output power was 69 mW, and the corresponding pulse duration, repetition rate, single pulse energy, and peak power were 7.12 ns, 8.83 kHz, 7.81 μJ, and 1.01 kW, respectively. As we have known, they are the best results of diode pumped, pulsed SFD laser, and the simple structure, miniaturization, as well as low cost are very favorable for large-scale production and application.

Data imbalance issue generally exists in most medical image analysis problems and maybe getting important with the popularization of data-hungry deep learning paradigms. We explore the cutting-edge Wasserstein generative adversarial networks (WGANs) to address the data imbalance problem with oversampling on the minority classes. The WGAN can estimate the underlying distribution of a minority class to synthesize more plausible and helpful samples for the classification model. In this paper, the WGAN-based over-sampling technique is applied to augment the data to balance for the fine-grained classification of seven semantic attributes of lung nodules in computed tomography images. The fine-grained classification is carried out with a normal convolutional neural network (CNN). To further illustrate the efficacy of the WGAN-based over-sampling technique, the conventional data augmentation method commonly used in many deep learning works, the generative adversarial networks (GANs), and the deep convolutional generative adversarial networks (DCGANs) are implemented for comparison. The whole schemes of the minority oversampling and fine-grained classification are tested with the public lung imaging database consortium dataset. The experimental results suggest that the WGAN-based oversampling technique can synthesize helpful samples for the minority classes to assist the training of the CNN model and to boost the fine-grained classification performance better than the conventional data augmentation method and the two schemes of the GAN and DCGAN techniques do. It may thus suggest that the WGAN technique offers an alternative methodological option for the further deep learning on imbalanced classification studies.

This paper proposes to resolve the limitation of differential evolution (DE) that the difference between the individuals in search behavior has not yet been utilized effectively for guiding the evolution of the population. An adaptive guiding mechanism (AGM) based on the heuristic rules is thus suggested to make possible, individual-dependent guidance. The AGM mainly comprises three stages: construction, separation, and guidance. In the construction stage, the elite leadership team (ELT) is established with an adaptive control scheme by using good information of the population. In the separation stage, the ELT is divided into distinct elite groups that are allocated to different individuals based on their search behaviors. In the guidance stage, the leader that is chosen from the respective elite group, as well as the promising directions extracted from the population, are used together to guide the search of each individual. By incorporating AGM into DE, a novel algorithm framework, named DE with AGM (DE-AGM), is proposed to enhance the performance of DE. As a general framework, DE-AGM can be easily and seamlessly applied to most DE variants. The experimental results on 58 benchmark functions have demonstrated the competitive performance of DE-AGM.

In recent years, people with depression tend to be younger and normalized. Although current findings can identify the depressive tendency, the effects of recognition are unsatisfactory. Therefore, this paper aims to integrate multi-modal features and build a more objective and accurate model to identify the depressive tendency effectively. First, we combine mental health self-examination data and eye tracking data to extract multi-modal features. Explicitly, we integrate a variety of functions, including eye movements, memory features, cognitive style features, and network behavior. Second, we innovatively propose a scanning stacking model to capture complex nonlinear relationships between features. Finally, the effectiveness of our proposed method is verified by experiments. The experimental results show that our multi-modal based scanning stacking method is superior to the state-of-the-art detection methods.

The application of inductive power transmission system in trams makes trams get rid of the traditional overhead catenary, but the electromagnetic environment and safety problems brought by the inductive power transmission system also attract the public's attention. In this paper, the model of the wireless charging tram platform is built to analyze the electromagnetic environment when the tram stops at the platform. Then, the shielding shape of the platform is proposed. A mathematical model with the minimum shielding cost as the goal and the magnetic field and installation space as constraints, is proposed for the platform shielding model of wireless charging tram. The three-dimensional finite element simulation software is used to verify the designed platform shield. The results show that the designed platform shield can ensure the electromagnetic radiation at the platform reaches the standard, and the cost is low. The platform shielding scheme designed in this paper provides a theoretical basis for the establishment of electromagnetic protection measures in the practical application of the wireless charging tram.

Learning from class-imbalanced data is a challenging problem as standard classification algorithms are designed to handle balanced class distributions. Scholars solve this problem by modifying classifiers or and generating artificial data by oversampling. The former usually design corresponding classifier to adapt them to the imbalanced data, while the latter exploits the sampling algorithm, which are the data preprocessing steps independent of the classifier. In this paper, we propose a novel synergistic oversampling algorithm to combine the oversampling and classification into one without training the classifier repeatedly, which can generate new pertinent samples according to the classification performance of the classifier without repeat training or deep understanding of the classifier, so the generated samples can guarantee the performance improvement of the classifier. Moreover, The proposed framework enclosures the oversampling method without traditional parameters in oversampling methods. Experimental results on several real-life imbalanced datasets demonstrate the effectiveness and efficiency of the proposed algorithm in binary classification problems.

Cyber-physical system (CPS) is a kind of complex real-time hybrid system which involves deep interactions between computation processors, communication network, and physical environments are deemed as the key enablers of next generation computer applications. However, how to verify CPS effectively is always a great challenge. Based on current scientific works about CPS verification, this paper aims at identifying the gap of current studies and suggesting promising areas for the future works. For this purpose, we conduct a systematic mapping study over the topic on verification of cyber-physical system. We carry out a widely search of publications from 2006 to 2018 in 11 electronic databases. After the step of study selection, 80 papers are selected as primary studies for answering proposed research questions, focused questions, and statistical questions. According to these questions and their answers, this paper not only presents a quantitative and comprehensive analysis of verification challenges, abstraction methods, verification techniques, assistance tools, and verification scenarios that represent each step of verification works, but also summarizes CPS systematic natures, main routine of verification and future research directions. We believe that this survey can identify gaps in current research works and reveal new insights for the future works.

In this paper, a new method is used to direct generate Ince-Gaussian (IG) modes from a pulsed solid-state microchip laser. By moving a circular aperture horizontally, the gain size in the laser medium is adjusted, and different IG laser modes are obtained successively. The output laser possesses a stable pulse duration of 500-600 ps; at the same time, the repetition frequency varies between 7 and 16 kHz. To our knowledge, this is the first time that IG modes are obtained in the subnanosecond laser domain, which will be helpful for making large-energy high-peak-power devices.