Search

The rapid ageing of society makes necessary the development of advanced technologies for the identification of frailty. In this paper, we present a system for mobility evaluation in frailty phenotype assessment. The system is equipped with wireless, small and non-invasive wearable sensors for an objective evaluation of mobility. The paper proposes an optimization of gait analysis algorithm using a dynamic threshold. The results obtained from a comparison with the gold standard show errors of 3.7% for double support, 5.1% for stride length, and 5.8% for stride speed. Moreover, a simple and automatic tool, which estimates postural and walking parameters to assist medical staff in assessing frailty, is developed.

The effective utilization of accumulated forestry science papers is of paramount significance in enhancing our understanding of the current state of forests and the formulation of strategies for forest environmental preservation. However, the present challenge lies in the deficient richness of metadata associated with these pivotal documents, rendering their comprehensive exploitation a formidable endeavor. Metadata from forestry science papers serves as a foundational cornerstone for the efficient management and utilization of these scholarly documents, playing an indispensable role in the advancement of research within the domain of forestry science. Constructing a training corpus and extracting distant semantic relationships is challenging inherent, the utilization of named entity recognition (NER) technology for metadata entity identification in forestry science papers remains an unexplored avenue. To overcome these limitations, this paper creates a specialized training corpus and introduces a novel few-shot NERframework tailored specifically for metadata extraction from forestry science papers. Within this innovative framework, a data augmentation layer, employing word replacement (WR) and enhanced mixup (EM), effectively addresses the issue of suboptimal performance resulting from a scarcity of training data. The semantic comprehension layer incorporates a multi-granularity dilated convolution neural network (MGDCNN) to capture and extract distant semantic associations. Moreover, a meta-learning-based reweighting layer is introduced to mitigate the adverse effects of low-quality augmented examples on the model. Experimental results conclusively demonstrate the efficacy of the proposed framework, yielding precision, recall, and F1 of 91.08%, 88.96%, and 90.00%, respectively. Compared to traditional models, precision, recall, and F1 can be improved by up to 10.69%, 7.48%, and 9.07%, respectively.

The research field of this paper is unsupervised learning in machine learning, aiming to address the problem of how to simultaneously resist feature attacks and improve model classification performance in unsupervised learning. For this purpose, this paper proposes a method to add an optimized loss function after the graph encoding and representation stage. When the samples are relatively balanced, we choose the cross-entropy loss function for classification. When difficult-to-classify samples appear, an optimized Focal Loss*() function is used to adjust the weights of these samples, to solve the problem of imbalanced positive and negative samples during training. The developed method achieved superior performance accuracy with the values of 0.721 on the Cora dataset, 0.598 on the Citeseer dataset,0.862 on the Polblogs dataset. Moreover, the testing accuracy value achieved by optimized model is 0.745, 0.627, 0.892 on the three benchmark datasets, respectively. Experimental results show that the proposed method effectively improves the robustness of adversarial training models in downstream tasks and reduces potential interference with original data. All the test results are validated with the k-fold cross validation method in order to make an assessment of the generalizability of these results.

Understanding the domain topic of software in terms of maintenance and reuse is important. However, the continual development of software and changes in its size make it difficult for engineers to understand. Recent research studies have sought to solve this problem by extracting the domain topic using various information searching techniques, such as latent semantic indexing (LSI) and latent dirichlet allocation (LDA), with the research on LDA-based techniques being particularly active. However, since the research has used only unstructured information, such as identifiers or notes, without including structured information, such as a method of calling information, it has caused problems in which extracted topics differ according to the program’s characteristics. This paper proposes a method of generating documents and extracting topics using both structured and unstructured information. In addition, indexes are generated based on the frequency of the identifier’s occurrence in the source code, and a system is proposed that extracts an association rule based on the method’s co-occurrence. Moreover, this paper suggests an information retrieval system that can provide highly reliable search results for user queries by combining domain topics, indexes with scores, and association rule information. We also develop the Topic EXtract And Search System (TEXAS2) system for this research and confirm high user satisfaction with the search results to their queries in a performance test.

In order to solve the problem of low detection efficiency and long working time in the traditional video surveillance system for abnormal behavior detection and identification methods. A multimodal abnormal behavior detection and identification method based on video surveillance is proposed and applied to an online video classroom concentration evaluation task for college students in English. The model works by capturing abnormal behaviors and facial expressions and building a joint network that fuses abnormal behaviors and facial expressions. By testing on two open-source datasets and self-built classroom real-time datasets, the results verify that the model in this paper has better recognition performance compared to current mainstream models while maintaining real-time performance. The model proposed in this paper provides a new way of thinking about building smart classrooms.

Predicting merger waves has been a classical yet challenging problem. In this paper, we propose approaches to predict industry merger waves relying on an integrated dataset including financial statements and supply data, as well as more than 60 thousand firm-level mergers and acquisitions records. We utilize 1000-dimension features—including common-used industry characteristics and novel supply network information—for predictions and train classifiers based on different machine learning methods. The experiments demonstrate the usefulness of our prediction approach, as the predicting precision reaches 91% on acquirers and 96% on targets. By further analysis, some patterns are well explained by financial theories, such as the well-known Tobin’s Q measurement. Especially, new influential factors on merger waves are revealed by the empirical analysis on micro-structure network features. To the best of our knowledge, this paper is one of the first attempts to explore merger waves prediction, and our approaches and findings introduce a new viewpoint for this field.

Bayesian inference is one of the important topics in modern statistics. The information of the parameter in Bayesian statistics which is regarded as some random variable will be updated by that of the posterior distribution. In other words, all the inferences in Bayesian statistics are based on the updated posterior information, which has been proven to be a very powerful technique. In this paper, we study the Bayesian inference in the framework of uncertainty theory based on the uncertain Bayesian rule developed by Lio and Kang in 2022. To be more precise, issues on the point estimation, credible intervals and hypothesis testing in Bayesian statistics under uncertain theory are explored, and one application of our method in an IQ test problem is also given in this paper.

Mobile charging provides a new way for energy replenishment in the Wireless Rechargeable Sensor Network (WRSN), where the Mobile Charger (MC) is employed for charging nodes sequentially via wireless energy transfer according to the mobile charging sequence scheduling result. Mobile Charging Sequence Scheduling for Optimal Sensing Coverage (MCSS-OSC) is a critical problem for providing network application performance; it aims to maximize the Quality of Sensing Coverage (QSC) of the network by optimizing the MC’s mobile charging sequence and remains a challenging problem due to its NP-completeness in nature. In this paper, we propose a novel Improved Q-learning Algorithm (IQA) for MCSS-OSC, where MC is taken as an agent to continuously learn the space of mobile charging strategies through approximate estimation and improve the charging strategy by interacting with the network environment. A novel reward function is designed according to the network sensing coverage contribution to evaluate the MC charging action at each charging time step. In addition, an efficient exploration strategy is also designed by introducing an optimal experience-strengthening mechanism to record the current optimal mobile charging sequence regularly. Extensive simulation results via Matlab2021 software show that IQA is superior to existing heuristic algorithms in network QSC, especially for large-scale networks. This paper provides an efficient solution for WRSN energy management and new ideas for performance optimization of reinforcement learning algorithms.

Object recognition is a challenging task in image processing and computer vision. In this paper, segmentation, feature extraction and classification methods are done for elephant recognition. Thresholding based segmentation technique is used for image segmentation and k-NN classifier is used for object recognition based on the shape features of the segmented image. Infrared elephant images are considered for experimentation. The database created by us for this type of object recognition includes elephant, bear, horse, pig, tiger, and cow and lion images. The recognition rate is calculated for performance evaluation. However, implementing such algorithms on software consumes more time as image sizes and bit depths grow larger. Hence this paper aims at hardware implementation of elephant recognition to reduce the computational time. The proposed hardware is prototyped inVirtex-4 xc4vlx25 FPGA using Xilinx System Generator (XSG) tool. The hardware/software co-simulation feature allows the input and output to be displayed on Matlab window while the processing is done through FPGA. The results indicate that when the category is elephant and if the recognition status is “yes”, recognition rate is 100%. If the category is not an elephant and if the recognition status is “no”, recognition rate is still 100% also indicates, the approach is successful in elephant recognition and the computation of segmentation algorithm and shape feature extraction (area, centroid, equivdiameter) in hardware reduces the computational time of elephant recognition by 89.65% as compared to software computation.

With the development of medical image technology, two-dimensional imaging technology bottlenecks highlighted, three-dimensional imaging technology needs more and more urgent. Cone beam projection computer reorganization tomography equipment principle, which is also called “CBCT”, is the X-ray generator with a lower ray around the projection body to do a circular DR digital projection. Then the data obtained in the “intersection” after projection of the projection body has reconstructed in the computer to obtain the three-dimensional image. Image processing has still faced with image processing problems, such as processing speed and stability. In this paper, the key technology of high-speed image processing system based on FPGA is studied, and a high-speed and high-definition cone-shaped beam projection computerized tomo-graphic image processing platform based on FPGA is realized. The design of this paper is based on the image processing algorithm module of the pipeline. According to the structural characteristics of FPGA, a time division-multiplexing algorithm based on buffer shift is proposed to solve the problem of time division multiplexing for shared resources in image processing process, which facilitates the allocation of buffer memory resources. In this paper, the parallel CORDIC algorithm has introduced to solve the problem of the transcendental function in the image algorithm. The iterative speed advantage is obvious, which provides a powerful guarantee for the later image registration.