In the realm of unmanned aerial vehicle (UAV) swarm intent recognition, conventional approaches predominantly focus on the attributes derived from singular targets at discrete instances. This trend leads to a significant limitation: the inability to effectively harness and capture the collective feature information of the entire swarm over temporal sequences. To address this gap, this study introduces a comprehensive end-to-end UAV swarm intent recognition approach. Initially, this method utilizes the distance threat coefficient and angular threat coefficient between UAVs to construct the graphical structural representation of the UAV swarm. Subsequently, an innovative deep learning framework, designated as attention-pool based on graph attention network and long short-term memory, which integrates a graph attention network, a novel graph pooling strategy, and a long short-term memory, is developed. This architecture can process the graphically structured data derived from the swarm modeling and accurately deduce the collective intent. Through experimental validation and analyses against existing methodologies, as well as ablation studies, it is evidenced that the model outperforms state-of-the-art methods in terms of accuracy of intent recognition.

Recognizing human behavior is essential for early interventions in cognitive rehabilitation, particularly for older adults. Traditional methods often focus on improving third-person vision but overlook the importance of human visual attention during object interactions. This study introduces an egocentric behavior analysis (EBA) framework that uses transfer learning to analyze object relationships. Egocentric vision is used to extract features from hand movements, object detection, and visual attention. These features are then used to validate hand-object interactions (HOI) and describe human activities involving multiple objects. The proposed method employs graph attention networks (GATs) with transfer learning, achieving 97% accuracy in categorizing various activities while reducing computation time. These findings suggest that integrating the EBA with advanced machine learning methods could revolutionize cognitive rehabilitation by offering more personalized and efficient interventions. Future research can explore real-world applications of this approach, potentially improving the quality of life for older adults through better cognitive health monitoring.

Key point detection in football matches can provide effective support for understanding football videos and recognizing player movements. This study analyzes the position information of players in football matches using key point detection methods. A cascaded convolutional neural network for detection and regression is combined to design a key point detection model for the impact on football game commentary and player control rate. A first level key point detection network based on ResNet and a second level regression network based on heatmap are proposed by combining the fused stacked feature extraction and optimizing the original structure. The experiment confirmed that the model has an accuracy of 95.62% in detecting key points. At the same time, the coordinate error near the goal and other areas is relatively small. However, there is a significant coordinate error in the size exclusion zone. This research provides a certain reference for the analysis of future football matches.

Regional folk songs have a rich history and are filled with cultural values. In this paper, first, the style characteristics of regional folk songs are briefly introduced. Using four regional folk songs from the northwest, northeast, southwest, and Hakka as examples, time domain, frequency domain, and mel-frequency cepstral coefficient (MFCC) features were extracted. Finally, the bidirectional long short-term memory (BiLSTM)-based music classification algorithm is used to realize the classification of folk songs from different regions. It was found that using the time-frequency domain + MFCC as features produced better results in music classification than using only the time-frequency domain or only MFCC features. The BiLSTM algorithm achieved an accuracy of 0.8339 and an F1 value of 0.8201 for the 10 s fragment set, both of which were better than those of the K-nearest neighbor, support vector machine, and other classification algorithms. The results show that the approach used in this study to categorize regional folk songs is reliable and that it can be applied to real folk songs.

This paper introduces an artificial intelligence (AI) interactive system featuring a self-growing memory network designed to enhance self-efficacy, reduce loneliness, and maintain social interaction among the elderly. The system dynamically analyzes and processes user-written diaries, generating empathic and personalized responses tailored to each individual. The system architecture includes an experience extraction model, a self-growing memory network that provides a contextual understanding of the user’s daily life, a chat agent, and a feedback loop that adaptively learns the user’s behavioral patterns and emotional states. By drawing on both successful and challenging experiences, the system crafts responses that reinforce the self-efficacy of the user, fostering a sense of accomplishment and engagement. This approach improves the psychological well-being of elderly users and promotes their mental health and overall quality of life through consistent interaction. To validate our proposed method, we developed a diary application to facilitate user interaction and collect diary entries. Over time, the system’s capacity to learn and adapt further refines the user experience, suggesting that AI-driven solutions hold significant potential for mitigating the effects of declining self-efficacy on mental health and social interactions. With the proposed system, we achieve an average system usability scale score of 77.3 (SD = 5.4) and a general self-efficacy scale score of 34.2 (SD = 3.5).

To address the difficulty in identifying human lower-limb movement intentions, low accuracy of classification models, and weak generalization ability, this study proposes a motion intention recognition method that combines an improved pelican optimization algorithm (IPOA) and a hybrid kernel extreme learning machine (HKELM). First, we collect the surface electromyography (sEMG) signals of subjects in six motion modes and perform feature parameter extraction under non-ideal conditions. On this basis, we establish a dataset of the relationships between the feature parameters and gait movements. Second, we build a motion intention classification model based on relational data using the HKELM to solve the problems of low modeling accuracy and weak generalizability. Third, the IPOA is used to optimize the parameters related to the HKELM, and a differential evolution algorithm is introduced to improve the population quality and prevent the algorithm from falling into a local optimal solution. The experimental results show that the IPOA exhibits better optimization accuracy and convergence speed for four classical benchmark functions. Its average classification accuracy, average classification recall, and average F-value are 94.45%, 94.47%, and 94.46%, respectively, which are significantly higher than those of other intention recognition algorithms. Therefore, the proposed method has high classification accuracy and generalization performance.

Mobile ad hoc networks (MANETs) are decentralized networks of mobile devices connected by wireless links, enabling various applications across domains. Research on MANETs often relies on simulations for efficient and replicable evaluations due to the absence of central administration and complex real-world scenarios. Effective simulation necessitates the integration of mobility models to orchestrate node movement. In this study, we explore existing mobility models, propose evaluation metrics, and conduct simulations to characterize, compare, and rank these models. Additionally, we analyze ad hoc routing protocols, focusing on packet delivery ratio, average packet arrival time, and total control packets. Notable protocols such as AODV, DSR, DSDV, and TORA are scrutinized, considering factors like node count and mobility, to provide comprehensive insights into their performance and efficacy.

Teleoperated robots are attracting attention as a solution to the pressing labor shortage. To reduce the burden on the operators of teleoperated robots and improve manpower efficiency, research is underway to make these robots more autonomous. However, end-to-end imitation learning models that directly map camera images to actions are vulnerable to changes in image background and lighting conditions. To improve robustness against these changes, we modified the learning model to handle segmented images where only the arm and the object are preserved. The task success rate for the demonstration data and the environment with different backgrounds was 0.0% for the model with the raw image input and 66.0% for the proposed model with segmented image input, with the latter having achieved a significant improvement. However, the grasping force of this model was stronger than that during the demonstration. Accordingly, we added haptics information to the observation input of the model. Experimental results show that this can reduce the grasping force.

Grasping is a fundamental action in daily life and particularly evident during mealtime situations where various grasping actions occur with tableware such as chopsticks, spoons, forks, bowls, and cups, each serving specific purposes. While tableware usage varies across regions and cultures, recognizing grasping actions is crucial for assessing performance in daily activities. In this study, we focus on assessing grasping functionality in terms of tableware usage during meals and propose a method for identifying hand movements. In recent years, there has been a surge in developing approaches for hand pose estimation and gesture recognition using deep learning. However, these approaches encounter common challenges, including the need for large-scale datasets, hyperparameter tuning, significant time and computational costs, and limited applicability to incremental learning. To address these challenges, we propose an ensemble approach employing extreme learning machines to recognize grasp postures. In addition, we apply spatiotemporal modeling to extract the relationship between grasp postures and the surrounding tools during mealtimes.

Advancements in robotics technology and information and communication technology have significantly affected agriculture, especially systems that aid workers. Soil cultivation often requires prolonged bending, which results in physical strain. A commercially available work wheelchair reduces back strain, but lacks power assistance and relies on leg movements. In a previous study, we developed an electric wheelchair with a fabric pressure-distribution sensor on the seat. This sensor, combined with a fuzzy controller, aids movement by measuring the center of pressure fluctuations and reducing leg strain. However, it does not recognize user’s actions, such as standing or sitting, and requires the wheelchair to stop. This study introduced a user-action recognition system using a wheelchair seat pressure sensor. This system accurately recognizes user actions through deep learning time-series data. We evaluated data augmentation and multi-user data to enhance the performance. The results show improved prediction accuracy for action recognition in the test scenarios.

In the knowledge economy era, innovation is the main engine that drives high-quality development of the manufacturing industry. Exploring the mechanism behind intellectual capital, its constituent elements that stimulate innovation within manufacturing enterprises, and the influence of executive motivation emerges as a crucial imperative, considering intellectual capital as the sum of intangible assets of enterprises. Unraveling this intricate interplay becomes essential in comprehending the driving forces behind innovation in this sector. Based on this, the present study analyzed the relationship between intellectual capital, executive incentives, and corporate innovation level using a sample of manufacturing companies listed in Shanghai and Shenzhen A-shares from 2012–2021. The empirical findings demonstrate the positive impact of human capital, structural capital, and relational capital on the innovation levels of manufacturing firms. Moreover, executive compensation incentives and equity incentives also contribute positively to enhancing firm innovation. Notably, these executive incentives play a vital moderating role in strengthening the relationship between intellectual capital dimensions and firm innovation. Intellectual capital dimensions have a stronger impact on innovation levels in non-state-owned manufacturing enterprises than in their state-owned counterparts. Moreover, the positive influence of intellectual capital on innovation is more significant in highly marketized regions compared to regions with lower levels of marketization.

Evaluating employee performance in enterprises is beneficial for improving talent competitiveness, operation level production, and overall enterprise performance. At present, many performance appraisal methods lack objectivity and fairness, which is not conducive to the long-term development of enterprises. Therefore, more research into scientific and effective performance appraisal methods is required. This study takes the grassroots employees of Enterprise A as an example, improving the existing performance appraisal evaluation indices to evaluate employee performance from three dimensions, including achievement, ability, and attitude, as determined by index weights using the analytic hierarchy process approach. An improved back-propagation neural network (BPNN) method is then designed to obtain performance appraisal and evaluation results. The error between the output of the improved BPNN method and expected output was small. Of the 20 extracted samples, the maximum and minimum error values were 0.05 and 0.01, respectively, and the average error was 0.03. The improved BPNN method evaluated only one out of 20 samples incorrectly, and the accuracy of the improved BPNN method was 96.21%, which is 19.88% and 10.75% higher than those of support vector machine and standard BPNN, respectively. The findings demonstrate that the improved BPNN method can be used in the appraisal and assessment of enterprise employee performance and has practical application value.

Air quality issues have become a major environmental concern, with severe air pollution significantly reducing air quality and posing threats to human health. Accurate air quality prediction is crucial for preventing individuals from suffering the detrimental effects of severe air pollution. Recently, deep learning methods based on spatiotemporal graph neural networks (GNNs) have made considerable progress in modeling the temporal and spatial dependencies within air quality data by integrating GNNs with sequential models. Unfortunately, previous work often treats temporal and spatial dependencies as independent components, neglecting the intricate interactions between them. This oversight prevents the models from fully exploiting the complex spatiotemporal dependencies in the data, adversely affecting their predictive performance. To address these issues, we propose a general spatiotemporal interaction framework for air quality prediction. This framework models the bidirectional interactions between temporal and spatial dependencies in a data-driven manner. Furthermore, we designed a spatiotemporal feature extraction module and a dynamic adversarial adaptive graph learning module based on this framework. We introduce the Spatial-Temporal Interaction based Dynamic Adversarial Adaptive Graph Neural Network, capable of capturing the complex interactions between spatiotemporal dependencies and learning the dynamic spatial topology among sites by incorporating the competitive optimization concept of generative adversarial networks. Extensive experiments on two real-world datasets demonstrate the effectiveness of the proposed method, outperforming existing baseline models.

With the deepening research on music works, music transcription algorithms have been increasingly studied. This study examined the recognition of piano-playing notes using a music-transcription algorithm. First, the characteristics of MelSpec, LogSpec, and the constant Q-transform (CQT) are briefly introduced. Then, a convolutional recurrent neural network (CRNN) transcription algorithm, which includes four convolutional blocks and one bidirectional long short-term memory (BiLSTM) structure, was designed. The recognition performance of this method was analyzed using the MAPS dataset. LogSpec was found to have the best recognition performance for piano-playing notes when used as an input feature. In the CRNN structure, the recognition performance for piano-playing notes was the best when four convolutional blocks were used. Compared with the convolutional neural network (CNN), BiLSTM, and CNN-hidden Markov model algorithms, the F1-values of the CRNN algorithm were 84.9%, 92.24%, and 79.27% for frames, notes, and offsets, respectively, achieving the best recognition results. The results verify that the CRNN transcription algorithm is effective for the recognition of piano-playing notes and can be applied in practice.

Currently, power communication wireless sensor networks (WSNs) exhibit the characteristics of uncertainty and complexity. Furthermore, the application environment is more complex, resulting in nodes damaging easily, power communication breakdown, and economic losses. Dynamic monitoring of power communication WSN nodes is necessary. To this end, a constraint data maximum entropy Bayesian network (BN) parameter learning algorithm is used to solve the problem of small data sample size in monitoring and improve the quality of power communication WSN node fault diagnosis. After the latest verification, it is found that the correct rate of WSN node fault diagnosis under small data set is as follows: in normal conditions, the fault rate is 96%, the sensor fault is 94%, the power supply fault is 100%, the wireless communication fault is 90%, and the processor fault is 91%. It can be seen that even under small sample data sets, better diagnostic accuracy can be brought.

Cooperative control of multiple heavy-haul trains can improve the safety and efficiency of heavy-haul railway transportation. However, the influence of internal and external unknown disturbances for multiple heavy-haul trains is a serious obstacle, which will lead to imprecise train operation control. To address this issue, a cooperative active disturbance rejection control for heavy-haul trains is proposed. First, a multi-mass point longitudinal dynamic model of heavy-haul trains is established to meet the actual operation. Second, a cooperative active disturbance rejection controller is designed to estimate and compensate for the disturbance caused by the interaction between the trains and environment. Moreover, the extended state observer is leverage to estimate the nonlinear disturbance online, which enhances the resistance of multiple heavy-haul trains to nonlinear time-varying disturbance and suppresses the overshoots of train velocities and inter-train distance. Finally, the performance of the proposed method is verified in two different simulation scenarios: acceleration and deceleration conditions. The simulation results show that the proposed method reduces the maximum relative displacement by 38.9% and the velocity error by 54.5%.

Cooperative control for virtual coupling systems of multiple heavy-haul trains can improve the safety and efficiency of heavy-haul railway transportation. However, the false data injection attack for the virtual coupling system is a serious obstacle, which will lead to imprecise train operation control. To address this issue, a deep learning-based false data injection attack (FDIA) detection for virtual coupling systems of heavy-haul trains is proposed. First, the cyber-physical model of the virtual coupling system is established. Second, a cooperative control law is designed for the virtual coupling system, and the effects of the FDIA on the virtual coupling system is analyzed. Then, the unsupervised autoencoder method is introduced to achieve the false data injection attack detection. The autoencoder network model is trained with normal operation data and tested with abnormal operation data. The performance of the proposed method is verified in four different simulation scenarios: normal case, velocity attack case, position attack case, and joint attack case. Simulation results show that the proposed method can effectively increase the detection accuracy and reduce the error rate with other supervised methods.

In this paper, the industry classification of digital economy is determined, the comprehensive index framework of digital economy development level in China is constructed, and the input-output analysis method and index method are used to measure the digital investment status of manufacturing industry in China. At the same time, based on the realistic demand that China’s manufacturing industry needs to climb the regional value chain, this paper constructs the forward participation index, backward participation index, regional value chain position index, and regional value chain Balassa, and analyzes the current situation of China’s manufacturing industry in RCEP region. Construct a semi-parametric additive model to explore the impact of digital investment on China’s manufacturing industry’s participation in the RCEP regional value chain. In addition, this paper uses the community analysis method to build the RCEP regional manufacturing digital input trade network, and comprehensively evaluates the trade network pattern of digital input from multiple dimensions. The results show that: (1) the digital investment in China’s manufacturing industry is increasing year by year. (2) The indicators of manufacturing industry’s participation in RCEP regional value chain show a fluctuating situation, which is generally consistent with China’s economic development. (3) Within a certain range, digital investment will promote the participation of China’s manufacturing industry in the RCEP regional value chain. China is gradually becoming an important participant in RCEP regional trade. According to the corresponding measurement results, it provides policy suggestions for the deep integration of China’s manufacturing industry and digital economy.

The integrated knowledge graph summarization model improves summary performance by combining text features and entity features. However, the model still has the following shortcomings: the knowledge graph data used introduce data noise that deviates from the original text semantics; and the text and knowledge graph entity features cannot be fully integrated. To address these issues, a knowledge graph summarization model integrating attention alignment and momentum distillation (KGS-AAMD) is proposed. The pseudo-targets generated by the momentum distillation model serve as additional supervision signals during training to overcome data noise. The attention-based alignment method lays the foundation for the subsequent full integration of text and entity features by aligning them. Experimental results on two public datasets, namely CNN / Daily Mail and XSum, show that KGS-AAMD surpasses multiple baseline models and ChatGPT in terms of the quality of summary generation, exhibiting significant performance advantages.

Enhancing the precision of supply chain management and reducing operational costs are crucial for the development of the cross-border e-commerce market. However, existing research often overlooks the demand uncertainty caused by seasonal variations and the challenges of handling returns in logistics. Therefore, this paper proposes a SARIMA-CNN-BiLSTM prediction model that effectively captures both the seasonal and nonlinear characteristics of cross-border e-commerce supply chains. Additionally, by incorporating the returns process, a supply chain distribution optimization model is developed with the objective of minimizing total operational costs. The model is solved using an improved whale optimization algorithm. In validation with real-world data, the SARIMA-CNN-BiLSTM model achieved a mean absolute percentage error reduction of 6.479 and 7.703 compared to convolutional neural network (CNN) and BiLSTM models, respectively. Moreover, the chosen optimization algorithm reduced the cost by 231,310 CNY, 62,564 CNY, and 131,632 CNY compared to the whale optimization algorithm, genetic algorithm, and particle swarm optimization, respectively. The proposed approach provides robust support for cross-border e-commerce enterprises in reducing costs and enhancing efficiency in their operations.