Search

Accurate face recognition technology is of great significance for face anti-counterfeiting. Due to illumination, posture, angle, and other reasons, the existing face liveness detection technology is difficult to adapt the environmental changes, resulting in low detection accuracy. To address this issue, this paper presents a novel high-performance face anti-spoofing detection method named RGCS_ConvNeXt. The data-enhanced face images are fed into the ConvNext network, which group convolution is added to extract the correlation between different features, and the coordinate attention mechanism is used to enhance the facial feature extraction capability both spatially and directionally. Then SPPF is used to extract the features at different scales to enhance the representation of the feature map. Finally, the facial key point detection technique is utilized to calculate the eye EAR value to achieve accurate face anti-counterfeiting recognition. The proposed algorithm shows an average classification error rate of 0.3%, 1.7%, 1.9%±1.5% and 2.8%±3.4%, respectively, on the four protocols of the OULU-NPU public dataset. On the Siw dataset, the average classification error rate is 0.69%, a reduction of 0.02% compared to the MA-Net network. The half-error rate on the MSU-MFSD dataset is 2.39%, a 0.21% reduction compared to the DPCNN network. The algorithm shows good accuracy on the OULU-NPU, MSU-MFSD and Siw datasets, reaching 99.64%, 98.40% and 99.25% respectively, 0.26% higher than the SE-FeatherNet network’s average accuracy.

The feature selection for interval-valued data (IVD) aims to identify representative features from a large set of features, which can reduce the model complexity, minimize the training time, and enhance the generalization ability of the model. Addressing the inter-feature correlations in IVD, we propose a feature selection method called the maximum information coefficient for interval-valued data (IVD_MIC). First, the method balances the relationship between the midpoint and radius of IVD with an adjustment factor, constructing the interval-valued data unified representation frame (URF). Based on the URF, the method measures the degree of correlation between two features by calculating the maximum information coefficient, and obtains the maximum information coefficient matrix for IVD. Then the features with strong correlation are progressively removed from three perspectives(row, column, and both row and column), generating a series of corresponding candidate feature subsets. Finally, IVD_MIC is validated on candidate feature subsets to obtain the final classification accuracy and optimal feature subset. The experiment results on synthetic and real-world datasets with different classifiers demonstrate that the overall performance of IVD_MIC surpasses other methods. The average accuracy of IVD_MIC is higher, improving by 0.23%, 0.53% and 0.45% compared to the second-best method on LIBSVM, CART Tree and KNN, respectively.

Expressions serve as intuitive reflections of a person’s psychological state, making the extraction of effective features for accurate facial expression recognition a crucial research problem. However, when facial information is incomplete, the existing convolutional neural networks face some challenges in extracting features. To address this issue, this paper introduces a pyramidal convolutional attention residual network(PCARNet) based on the ResNet18. PCARNet combines the pyramidal convolution module and an improved convolutional attention mechanism to effectively extract expression features and achieve high-precision facial expression recognition. The proposed model utilizes pyramidal convolution to extract facial expression features at multiple scales, capturing both global and local information of the face. Grouped convolution is employed to reduce the computational complexity and the number of parameters. Additionally, to avoid the adverse effects of channel dimensionality reduction on the attention mechanism and enhance the capacity for information exchange across channels, the Share MLP module within the convolutional attention mechanism was replaced by a one-dimensional convolution with adaptive kernel size. The improved convolutional attention mechanism assigns weights to the extracted multiscale features based on both channel and spatial dimensions, enhancing the representation of crucial facial features. Experimental results demonstrate the high recognition accuracy of the proposed method on public datasets such as Fer2013, RAF-DB, and CK+. The accuracies achieved are 73.725%, 87.516%, and 95.455%, respectively. Compared to other methods, the proposed approach shows improvements of at least 1.4%, 2.4%, and 0.25% on the respective datasets, confirming its high reliability and performance.

Face anti-spoofing refers to the computer determining whether the face detected is a real face or a forged face. In user authentication scenarios, photo fraud attacks are easy to occur, where an illegal user logs into the system using a legitimate user’s picture. Aiming at this problem and the influence of illumination in real-time video face recognition, this paper proposes a real-time face detection method based on blink detection. The method first extracts the image texture features through the LBP algorithm, which eliminates the problem of illumination changes to a certain extent. Then the extracted features are input into the ResNet network, and the facial feature extraction is enhanced by adding an attention mechanism is added to enhance the face feature extraction. Meanwhile, the BiLSTM method is used to extract the temporal characteristics of images from different angles or at different times to obtain more facial details. In addition, the fusion of local and global features is realized by SPP pooling, which enriches the expression ability of feature maps and improves detection accuracy. Finally, the eye EAR value is calculated by the face key point detection technology to achieve face anti-spoofing, and then the real-time face recognition against fraud is realized. The experimental results show that the algorithm proposed in this paper has good accuracy on NUAA, CASIA-SURF and CASIA-FASD datasets, which can reach 99.48%, 98.65% and 99.17%, respectively.

This article identifies energy-saving potential of the fifth generation (5G) Radio Access Network, and describes main energy-saving principles and technologies. It explores how to use network energy saving technologies, such as carrier shutdown, channel shutdown, and symbol shutdown in 5G network, that have been inherited from 4G. Some enhanced technologies for 5G like equipment deep sleep and symbol aggregation have also been introduced in this article. However, it is far from enough and an innovative energy-saving solution should be considered. To meet the requirements and development of intelligent and self-adaptive energy-saving solution, Artificial Intelligence (AI) and big data analysis are introduced to form a more precise energy-saving strategy based on site-specific traffic and site-related conditions, thus improving the efficiency and reducing the manpower. Finally, two commercial application practices of AI-based energy-saving solution are elaborated. One is the practice of AI-based service awareness energy saving for 4G/5G collaborative networks, the energy benefits can be improved up to 20%; The other practice is the adoption of a new architecture Active Antenna Unit (AAU) with beam pattern optimization, its energy benefits can be promoted by 30%. These two practices could help mobile network operators (MNOs) to achieve the most energy-efficient network with good network performance and lower Operating Expense (OPEX).

Many electrical equipment malfunction text messages are collected during power system operation and maintenance procedures. These texts usually contain crucial information for maintenance and condition monitoring. Because these power system malfunction texts are characterized by multidomain vocabularies, complex-syntactic structures, and long sentences, it is challenging to for automated systems to capture their semantic meaning and essential information. To address this issue, we propose a hybrid natural language processing (hybrid-NLP) algorithm to extract entities that represent electrical equipment. This algorithm is composed of a dictionary-based method, a language technology platform (LTP) tool, and the bidirectional encoder representations from a transformers-conditional random field (BERT-CRF) model. Significantly, the softmax output layer of the bidirectional encoder representations from the transformers (BERT) model is replaced by the conditional random field (CRF) algorithm to strengthen the contextual relationships between words and thus solve the local optimization of the preferred word label. The effectiveness of the proposed hybrid-NLP method is verified on a realistic dataset. Moreover, a statistical analysis is conducted to provide a reference for the operation and maintenance of power systems.

The mine safety performance evaluation and its law were taken as research objects; the human factor, mechanical equipment, management system, environment and technology were taken as mine safety state indexes; accident loss and disabling incident frequency rate as mine safety performance evaluation indexes. The classification standard of safety performance was determined according to the range of influence index and evaluation index, and then the influence index value of safety performance was discretized. According to the classification standard, the classical domain and node domain of safety performance evaluation index classification were determined. Combined with the weight of each safety performance evaluation index, the correlation degree of each mine safety performance evaluation was calculated, and then the mine safety performance was evaluated, and the mine safety performance evaluation decision system was constructed. Based on the dominance-based rough set, the evaluation law of mine safety performance evaluation decision system was analyzed. To generate more reliable laws, an efficient concept lattice reduction method was introduced to reduce the attributes of decision system. The example shows that when the concept lattice is used for reduction, there are 84 concepts in the Hasse diagram, and 24 of them can be identified. The decision system has 8 conditional attributes. And 255 power sets can be obtained under 8 conditional attributes. After eliminating the defect attribute, there are 148 reducible sets, among which there are 4 optimal reducible sets with two conditional attributes. The results show that the proposed method can be used to analyze the law of mine safety performance, so as to improve mine safety work.

In advanced driver-assistance systems (ADAS), the accuracy and real-time performance of pedestrian- and vehicle-detection algorithms based on vision sensors are crucial for safety. Here, a lightweight detection algorithm based on aggregated channel features (ACFs),consisting of a context pixel ACF (CP-ACF) pedestrian detector and a multiview ACF (Mv-ACF) vehicle detector, is proposed to rapidly and precisely understand road scenes. The former fuses local and context information to improve the robustness to pedestrian deformation, while the latter contains a number of subclass detectors to alleviate intraclass differences due to different viewing angles. Compared to the original ACF, the CP-ACF pedestrian detector reduces the average miss rate (AMR) by 6.34%. The Mv-ACF vehicle detector improves the average precision (AP) by 40.26% on average at easy, moderate and hard levels. This remarkable effectiveness is due to the spectrum clustering of multiview samples and the resulting integration of these subclass detectors via confidence score calibration, which reduces the intraclass differences of vehicles. Since feature extraction takes up 68.8% of the total detection time, a mechanism of feature sharing between pedestrian and vehicle detectors is advanced to reduce the time spent in feature extraction. A strategy based on ground-plane constraints (GPCs) is proposed to control false detection of pedestrians and vehicles by incorporating road prior information, which reduces the AMR by 1.07% for CP-ACF pedestrian detectors and improves the AP by 0.27% on average for Mv-ACF vehicle detectors. Thus, the proposed algorithm can effectively control false detection by road prior information.

In the community of Granular Computing, knowledge is interpreted as one classification ability of realistic or abstract objects. Generally, the concept of granularity is used for characterizing such an ability, which has been widely explored in literatures. To calculate the parameterized granularity, a naive approach is to find the granularity in terms of the parameter one by one. Nevertheless, such approach can only generate the single parameter based knowledge granularity, and the difference of knowledge granularities among different parameters may be slight. It follows that the knowledge granularity derived from single parameter may be lack of representativeness. In this paper, the continuous parameters based knowledge granularity is proposed, and the corresponding calculation approach is presented. Inspired by the thinking of definite integral in mathematical problems, the calculation approach is mainly implemented by following steps: firstly, the graph formed by granularity and parameter interval is divided into several small rectangles whose length of interval tends to be 0; secondly, the sum of area values of all the small rectangles is calculated; finally, the obtained area value divided by the whole length of parameter interval can be considered as the continuous parameters based knowledge granularity. This study suggests a new trend of handling problems related to knowledge from the viewpoint of continuity.

Zero-shot learning (ZSL) aims to recognize the novel object categories using the semantic representation of categories, and the key idea is to explore the knowledge of how the novel class is semantically related to the familiar classes. Some typical models are to learn the proper embedding between the image feature space and the semantic space, whilst it is important to learn discriminative features and comprise the coarse-to-fine image feature and semantic information. In this paper, we propose a discriminative embedding autoencoder with a regressor feedback (DEARF) model for ZSL. The encoder learns a mapping from the image feature space to the discriminative embedding space, which regulates both inter-class and intra-class distances between the learned features by a margin, making the learned features be discriminative for object recognition. The regressor feedback learns to map the reconstructed samples back to the the discriminative embedding and the semantic embedding, assisting the decoder to improve the quality of the samples and provide a generalization to the unseen classes. The DEARF model is validated extensively on the benchmark datasets, and the experiment results show that the DEARF model outperforms the state-of-the-art models, and especially in the generalized zero-shot learning (GZSL), significant improvements are achieved.

In the current electrical load profile analysis, considering the shortage of traditional methods on the typical load profile extraction of single consumers and the load profile feature extraction, this paper proposes an approach based on time series data mining. Firstly, this method reduces the dimension of the load profile of a single consumer based on the Piecewise Aggregate Approximation(PAA), and re-expresses the load profile of the consumer over a period based on the Symbolic Aggregate approXimation(SAX), representing the consumer's load profile with a symbolic string to extract the typical load profile. Then, combined with the load characteristic indices and time series-based features, the typical load profiles of different consumers are clustered based on the K-means algorithm to analyze the power consumption behaviors. Finally, this paper performs a case analysis with a UCI test data set, and the results show that the proposed approach can excavate typical power consumption behaviors of consumers and improve the electrical load profile analysis efficiency and the clustering quality.

In the presence of the increasing penetration of electric vehicles (EVs) and conflict of independent optimization objectives among each electric vehicle aggregator (EVA), real-time optimal scheduling (RTOS) of large-scale EVs based on dynamic non-cooperative game approach is proposed for optimal decision makings in a dynamic pricing market. First, real-time optimal scheduling framework is designed to describe the flow of energy and information. Then, equivalent model of large-scale EVs is formulated to address “curse of dimensionality” caused by a large number of decision variables. Then, the potential game theory is used to study the existence and uniqueness of the Nash equilibrium (NE) solution. Finally, a distributed approach based on alternating direction method of multipliers (ADMM) is designed to achieve the equilibrium. Case studies demonstrate that the proposed approach achieves peak load shifting and reduces cost of EVAs significantly. Furthermore, the proposed method obtains higher-quality solution compared with other methods and is more applicable for real-time optimal scheduling of large-scale EVs due to its high computation efficiency and privacy protection.

Considering the effects of disturbances in permanent magnet synchronous motor (PMSM), in this paper, a non-smooth composite control approach, which includes finite time disturbance observer (FTDO) for feedforward compensation and finite time control (FTC) for feedback control, is proposed to improve the anti-disturbance performance of PMSM. First, the FTDO is used to estimate the lumped disturbances, such as friction, parameter perturbation, and load variation. Then, the observed value is added to the speed controller as a feedforward compensation to eliminate the effect of disturbance. Second, FTC is introduced into the feedback control design part. In the end, by utilizing Lyapunov theory, the stability analysis of the overall closed-loop system is demonstrated. In contrast to the conventional asymptotically stable control strategy, the proposed composite scheme can provide not only a faster dynamic response but also a stronger capacity of disturbance rejection. The simulation and experimental tests are presented to demonstrate the superior properties of the proposed control scheme.

Edge computing handles delay-sensitive data and provides real-time feedback, while it brings data security issues to edge devices (such as IoT devices and edge servers). Side-channel attacks main threaten to these devices. Collision attack represents a powerful category of side-channel analysis in extracting security information from embedded cryptographic algorithms. Since its proposition in 2003, plenty of collision detection algorithms are presented, most of which enumerate all the values of target plaintext byte to find a collision. In this paper, we establish a relation between “Euclidean distance between traces” and “Hamming distance between values,” and take advantage of the distance information leaked from the power traces of encrypting an adaptively chosen plaintext to reduce the candidate plaintext space. Consequently, the collision is detected at a high pace. Moreover, this improvement is fault-tolerant, and its self-correction feature promotes the efficiency of attacks based on our method significantly. We take AES implemented with masks, which is usually employed in edge computing devices, for instance, to introduce our method and conduct experiments to verify its efficiency. According to the experimental results, for whole key recovery attacks, our method requires only 26.5% plaintexts, 32.2% traces, and much less than 10% computations of the collision-correlation attack launched by Clavier et al.

Thermoelectric generators are efficient devices to recover energy from the automotive exhaust gas. In this paper, conversion efficiency of automotive thermoelectric generator (ATEG) and the maximum electrical power generated by the ATEG, defining as the power output of the ATEG excluding the energy loss caused to the engine improved by optimizing the number of thermoelectric modules (TEMs) and its distribution pattern in an ATEG. An advanced numerical model of ATEG considering the effect of the heat transfer among the adjacent TEMs' rows is developed with Simulation-X software. In order to acquire the ATEG's optimal electrical performance, a 3-step optimization is applied. First, 17 independent factors (the number of TEMs in each row from 1 to 18) are assessed and the significant parameters are screened using Plackett-Burman design. Second, an experiment designed with a central composite design is performed to analyze the sensitivity of six selected factors and a surrogate model is built through response surface method. Then, conflicts in two objectives are settled with a multi-objective genetic algorithm. According to the optimization results of a given ATEG, the maximum electrical power generated by the ATEG is 139.47 W and the conversion efficiency is 2.51% under steady engine condition. Finally, the performances of the optimized design under different engine conditions are discussed. The results show that the maximum power generated by the ATEG and efficiency respectively increase by 49.8% and 106.5% after optimization when the exhaust inlet temperature is 805 K and the mass flow rate is 0.5 kg/s.