In order to improve the accuracy and stability of satellite clock bias prediction, a combined satellite clock bias prediction method based on the entropy weight method is proposed. Firstly, the method adopts a quadratic polynomial model and a gray model to make a single prediction of satellite clock bias and generate two sets of prediction results. Then, by calculating the entropy of error information of the two sets of prediction results, it determines the weights of each model and realizes the optimal fusion of the models. Finally, the entropy weight combination method is used to obtain a higher precision prediction result. Four different types of BeiDou-3 satellites were randomly selected for the prediction test by using the precision satellite clock bias products released by the GNSS Analysis Center of Wuhan University. The results show that the method can provide high-precision short- and medium-term predictions of BeiDou-3 satellite clock bias, and its 6-h average prediction accuracy and stability are 0.22ns and 0.46ns, respectively, which are 72.15% and 48.84% higher than the average prediction accuracy of quadratic polynomial and gray models, and the stability is 70.00% and 20 .69% higher, respectively.

With the increasing popularity of network technology and information concepts, Big Data (BD) and cloud computing technologies have emerged and are widely used in all industries. BD technology can fully utilize the application value of information. Cloud computing can store a large amount of data, and improve data usage efficiency, so as to make full use of the value of data. At the same time, data security issues have become increasingly important. However, due to various factors, many users are faced with data breaches and privacy violations in the BD cloud computing environment. In this case, there are some data security risks, and the key issue is to consider how to avoid these risks. By analyzing the relationship and differences between BD and cloud computing, this article studied the issues and influencing factors of Artificial Intelligence (AI) data security in the BD cloud computing environment, and proposed corresponding optimization strategies, so as to improve data security and provide users with a brand new experience. Through comparison, it could be seen that the data processing speed and legal integrity after using the optimization strategy significantly improved. Among them, data processing speed increased by 7.2% and legal integrity increased by 10.4%. BD cloud computing could effectively improve AI data security performance.

Real-time pedestrian detection, as a key technology in the field of computer vision, has broad application demands in intelligent surveillance, autonomous driving, robot navigation, and other areas. To address the problem that high-computational-power models are difficult to deploy on edge devices, this paper proposes a real-time pedestrian detection scheme based on the lightweight YOLOv5-tiny model. The study uses a pedestrian subset of the COCO dataset for model training, optimizes the anchor box dimensions through the K-means clustering algorithm to adapt to pedestrian target characteristics, and tests the model performance on ordinary CPU and GPU environments. Experimental results show that the optimized model can achieve a detection speed of 23.6 FPS with a recall rate of 82.3% on the Intel Core i7-10700 CPU; on the NVIDIA GTX 1650 GPU, the frame rate increases to 45.2 FPS and the recall rate rises to 84.7%, which can meet the real-time and detection accuracy requirements in low-computational-power scenarios.

This study addresses the practical demands for whole lifecycle management of infrastructure projects in large hospitals by systematically analyzing the inherent limitations in conventional management approaches, including non-standardized processes, information silos, and insufficient decision-making support. Guided by principles of modularity, standardization, and intelligentization, we propose a Hospital Infrastructure Management Information System (HIMIS) that comprehensively covers all project phases from initiation and design to construction and handover. The system features a three-tier architecture, i.e., user interaction layer, application logic layer, and database layer, integrating core modules for project approval, implementation, and handover. Advanced digital technologies, such as blockchain-based data authentication, BIM model collaboration, and AI-driven predictive alerts, are incorporated to enable full-chain data traceability and risk early-warning capabilities. Furthermore, this research delineates technical pathways for deep integration with hospital budgeting and auditing systems, providing a theoretical framework and practical reference for resolving information silos in medical infrastructure management and establishing a digital infrastructure foundation for smart hospitals.

With the continuous advancement of technology and the increasing demand for temperature measurement accuracy in various industries, the design and implementation of a high - performance digital thermometer have become an urgent problem to be solved. In this study, a digital thermometer was successfully designed, which could be widely applied to multiple fields, such as water temperature measurement, temperature monitoring in daily life, and temperature control in industrial production processes. In the hardware architecture of this thermometer, the NTC thermistor and NY8B062D single - chip microcomputer were respectively assigned key functional roles. The former was used to sense the subtle changes in ambient temperature and convert them into electrical signals, while the latter was responsible for the subsequent conversion, analysis, and processing of these electrical signals. The temperature data processed by the single - chip microcomputer would be transmitted to the display module through a specific communication protocol, thus realizing the real - time display of the measured temperature values. In order to effectively overcome the impact of the nonlinear characteristics of the thermistor itself on the measurement accuracy, this study adopted the multisegment linear regression technology. Through the collection and analysis of a large number of experimental data, multiple temperature segmentation intervals were determined, and the optimal linear regression equations were fitted for each interval. At the same time, combined with the 12 - bit ADC technology, the resolution and quantization accuracy of signal sampling were greatly improved. After a series of rigorous experimental tests and optimization adjustments, the measurement accuracy of this thermometer was significantly enhanced, and its measurement error was successfully limited within ±1%, providing reliable technical support for accurate temperature measurement in practical applications.

To address the issues of low diagnostic accuracy in traditional rolling bearing fault diagnosis models and the ineffective extraction of spatial and temporal features from vibration signals, this paper proposes a rolling bearing fault diagnosis method based on the improved sparrow search algorithm optimized VMD and multi-scale convolutional neural networks. First, the improved sparrow search algorithm is employed to adaptively optimize the penalty parameter and mode count in variational modal decomposition (VMD). This achieves finer frequency band segmentation and effectively suppresses energy leakage, thereby yielding high quality frequency domain representations. Second, a multi-scale convolutional neural networks (MSCNN) is constructed, with feature level fusion implemented. Subsequently, a bidirectional long short-term memory networks (BiLSTM) is introduced to model the temporal dependencies of the fused features, enabling fault mode learning. A softmax layer is employed to achieve multi-class classification. Finally experimental results and comparisons based on the CWRU bearing dataset demonstrate the effectiveness of the proposed method in the rolling bearing fault classification task, providing significant application value for achieving efficient and reliable bearing fault detection.

Existing research methods for recognizing EEG (Electroencephalogram) signals in motor imagery (MI) often overlook the dynamic changes of brain networks over time, resulting in insufficient classification accuracy for MI tasks. This article addresses the recognition problem of dynamic changes in brain networks during MI tasks and applies an EEG signal classification method based on multi-layer time-varying functional brain networks. This article uses the BCI (Brain-Computer Interface) Competition IV 2a dataset to preprocess the raw EEG signals through bandpass filtering and CSP (Common Spatial Pattern) algorithm. The EEG signals of the MI task are divided into 7 1-second time windows with a step size of 0.5 seconds. Within each time window, Pearson correlation coefficients between EEG channels can be calculated to generate corresponding brain networks, and multiple time-varying functional brain networks can be constructed by stacking the brain networks from multiple time windows. The network topology features, node degree, clustering coefficient, network efficiency, and multi-layer network features of each window can be extracted, including Multiplex Clustering Coefficient (MCC), Multiplex Participation Coefficient (MPC), and inter layer correlation coefficient. By dividing the dataset through 10 fold cross validation, the random forest algorithm can be used to classify and recognize four types of motion imagination tasks. The experimental results show that the average recognition rate of the article’s method in four types of MI tasks reached 89.19%. This method can improve the classification accuracy of MI tasks and enhance a comprehensive understanding of the dynamic changes in brain networks during the process of MI.

Aiming at the problem of insufficient feature extraction in some deep learning-based gearbox fault diagnosis models under small sample conditions leading to lower fault diagnosis accuracy and larger number of parameters, in this paper, a lightweight gearbox fault diagnosis method based on depth-separable cascade residual block and feature weighting module is proposed. Firstly, the one-dimensional original signal of the gearbox is used as the input of this model, which reduces the loss of information in data processing. Then the depth-separable cascade residual block is constructed, which utilizes the depth-separable convolution with a cascade residual structure to maximize the extraction of fault information while reducing the amount of feature parameters. Finally, the feature weighting module strengthened the model's identification and exploitation of key features by calculating the contribution of each channel and giving them weighting. The experimental validation is given by the gearbox dataset of Southeast University, and the experimental results show that the proposed method achieves 99.99% fault diagnosis accuracy under the original signal, and 99.60% under the SNR=6dB noise environment, which shows that the proposed method has high fault diagnosis accuracy and low complexity under the small sample condition.

National music is a cultural treasure with unique charm and charm in the traditional Chinese culture, which has high research value and broad social influence. This paper aimed to explore the construction of ethnic music knowledge map based on big data analysis. This paper proposed the knowledge map of ethnic music and big data clustering analysis, and studied the experimental results of constructing the knowledge map of ethnic music based on this research. The experimental results of this paper showed that the knowledge map provided a scientific framework and method for the exploration of ethnic music knowledge. It made a reasonable explanation and evaluation for researchers in the field of ethnic music and pointed out the direction for the development of ethnic music. This paper identified 24 high-frequency keywords, which are the basis of co-word analysis. Among them, "national music" appeared 1940 times, and "ethnomusicology" appeared 465 times. "Folk music" appeared 415 times, and "music tradition" appeared 276 times. "Chinese music" appeared 270 times. Multivariate statistical methods are often used in co word analysis. These are the central links in co word analysis. Clustering analysis was used to classify keywords in ethnic music, thus revealing the current hot topics in ethnic music. In a word, the construction framework of ethnic music knowledge map based on big data analysis is conducive to the development of ethnic music.