Systems and Soft Computing最新文献

As the information age develops, the scale and form of data become more diverse and diverse. Therefore, people need to use effective means to process information. For large-scale data mining problems, a clustering-based kernel matrix inner product filtering method is introduced to decompose the original quadratic programming problem into multiple sub-problems to support parallel training. And a Spark-based multiple submodels parallel support vector machine is proposed. By introducing open-source tools such as OpenCV, image feature extraction can be performed on large-scale video data. Finally, combined with the designed parallel support vector machine algorithm, video facial and expression recognition is carried out. These experiments confirmed that the research method achieved a maximum acceleration ratio of 2090 times when processing Covtype datasets. The research model could achieve an accuracy of over 99 %. Under the maximum data scale experiment, the research model improved prediction accuracy by 21 percentage points with an acceptable additional time cost of about 4 min only. Task parallel processing could more fully utilize cluster performance, increasing by approximately 3.5 m/s² from 30 to 150 cores. The research model had the highest recognition accuracy for facial expressions, further demonstrating the effectiveness and superiority of this method. The research method has improved the efficiency of big data analysis and mining, and is of great significance in parallel analysis of video data.

A multi-modal feature based motion emotion analysis model based on a fusion deep learning model is proposed for the problem of analyzing the motion emotions of participants in the joint exercise quality expansion task. This model involves three major modalities: EEG signals, peripheral physiological signals, and facial expression signals, and processes and fuses the information of these three main modalities to achieve the effect of processing multi-dimensional motor emotional information. At the same time, this study introduces the design concept of residual networks, using self attention modules and multi head mutual attention modules to process different modal features. The results showed that the combination of peripheral physiological modality and facial expression modality had the highest accuracy among the three modality combinations, with an accuracy rate of 88.8 %. The feature fusion method based on the cascaded residual attention mechanism module has better accuracy and F1 Score performance than other methods. In addition, different emotional states can be effectively identified and distinguished in these three modalities, indicating that the model has a wide range of possibilities in practical applications.

Automatically recognizing sheep breeds is highly valuable for the sheep farming industry, allowing farmers to pinpoint their specific business needs. Accurately distinguishing between sheep breeds poses a challenge for numerous farmers with limited expertise. Although biometric-based identification offers a feasible solution, its application becomes impractical when assessing large numbers of sheep in real-time. Therefore, the implementation of an automatic sheep classification model that can replicate the breed identification skills of a sheep breed expert can come in handy. This would be particularly beneficial for novice farmers who could utilize handheld devices for breed classification. To address this objective, we propose employing a convolutional neural network (CNN) model capable of rapidly and accurately identifying sheep breeds from low-resolution images. Our experiment utilized a dataset of 1680 facial images representing four distinct sheep breeds. We conducted experiments on the dataset using various EfficientNet models and found that EfficientNetB5 achieved the highest performance with 97.62 % accuracy on a 10 % test split. The classification model we developed has the potential to assist sheep farmers in efficiently distinguishing between different breeds, facilitating more precise assessments and sector-specific classification for various businesses within the industry.

It is necessary to ensure security and community safety around High Voltage Transmission Poles (HVTP). Legislation requires a safety perimeter around HVTP and the High Voltage Lines (HVL) where no building and tree can be located. However, surveying thousands of kilometers of circuit is an expensive and challenging task that is currently performed by human inspection. Therefore, the use of automatic detection methods enables to facilitate the inspection is necessary to reduce time and cost. Convolutional Neural Network (CNN) is proposed in this work to detect, from Google Earth images, buildings and trees within the safety perimeter of HVTP. A dedicated 3 class (House, forest and HVTP) dataset of approximately 1 million tiles with a resolution of 0.09 m/pixel is created. Tiles size for trees and building classes is 64 × 64 pixels while for the HVTP 128 × 128 pixels is used. Three CNN models are built and optimized to classify each of these classes. Models validation shows that, except for houses where the accuracy is only 84 %, the other two classes have an accuracy of over 89 %. Moreover, by analyzing the classified HVTP, type can be identified. Finally, buildings and trees within the safety perimeter around the HVTP can be identified and displayed on the image demonstrating the usefulness of the tool.

Online network education offers convenience, however, the inefficiency and time-consuming nature of question-answering models negatively impact the demand for online learning. To address this issue, the study puts forward the development of an automatic English question-answering model. The improved model leverages a term frequence-inverse document frequency approach and an unsupervised participle algorithm based on deep learning. The precision and promptness of the question-answering model were enhanced by refining the weighted allocation of the term frequence-inverse document frequency algorithm and the unsupervised word-splitting algorithm. The validation shows that the improved precision rate is 68.14%, which is 34.37% and 50.45% more than the other two methods, respectively. The precision rate, recall rate, and F1 value for semantic similarity calculation improved by 9.23%, 9.22%, and 9.71%, respectively, compared to the traditional method. The validation experiments of the automatic English question-answering model indicate that its average accuracy was 94.68%, surpassing other models by 4.77%. The average answer time for the four types of questions was 30.52 ms, and the average answer time for the cause questions was 11.45 ms. The results show that the proposed English automatic question-answering model has better accuracy and timeliness of answering questions, and the improved accuracy for weight calculation is better. The English automatic question-answering model integrating word frequency-inverse document frequency and participle algorithm can satisfy the basic needs of teachers and students in online teaching, course question-answering, etc., which is of positive significance for the development of online education in the context of the Internet.

Accurate short-term photovoltaic (PV) power forecasting can reduce the un- certainty of PV power generation, which is crucial for grid operation as well as energy dispatch. Considering the influence of seasonal and meteorological factors on short-term PV power prediction, a short-term PV power predic- tion method based on meteorological similarity day and sparrow search algo- rithm and bi-directional long and short-term memory network combination (SSA-BiLSTM) is proposed. Firstly, the correlation between meteorological factors and PV power generation is calculated by using Pearson coefficients, getting the strongly correlated meteorological factors affecting PV power generation; afterwards,the historical data of the strongly correlated meteorological factors are clustered by fuzzy C-means clustering to achieve meteorological similar day clustering; then, the best similar day is selected from the meteorological similar day according to the test day seasonal features and meteorological data, and Forming a training set with historical data, and training the original BiLSTM network. the SSA algorithm was used to find the optimal BiLSTM network parameters. Finally, Using the optimal parameters construct BiLSTM network to achieve short-term PV power prediction. The experiments were conducted with historical data from a PV power plant in Xinjiang, and also compared with existing prediction algorithms.The results show that the accuracy of PV power prediction in different weather conditions is 33.1 %, 31.9 % and 24.1 % higher than that under the same intelligent optimization algorithm and different neural networks, the accuracy of PV power prediction in different weather conditions is 27.9 %, 24.7 % and 18.0 % higher than that under the different intelligent algorithms and same neural network. Therefore, the algorithm in this paper has better accuracy in short-term PV power prediction under different seasons and different weather conditions.

Objective

The purpose of this paper is to utilize a convolutional neural network (CNN) to identify the types of basketball footwork of athletes as a way to assist in the training of basketball players' footwork and to improve their performance in the game.

Methods

A traditional CNN algorithm was improved to a dual-model CNN (DMCNN) algorithm, where convolutional feature extraction was performed separately on both the acceleration and angular velocity data of footwork. The two features were then merged and subjected to principle component analysis (PCA) dimensionality reduction for identifying different types of footwork. In subsequent simulation experiments, ten basketball players' footwork data were collected using sensors. The improved CNN algorithm was used for footwork recognition and compared with the support vector machine (SVM) and traditional CNN algorithms.

Results

The experimental results showed that the acceleration and angular velocity signals of different basketball footwork had distinct differences. The comprehensive recognition precision of DMCNN for footwork types was 98.8 %, and the comprehensive recall rate and overall F value were 97.8 % and 98.2 %, respectively. Its recognition time was 1.23 s. For the traditional CNN algorithm, the comprehensive precision was 87.5 %, the comprehensive recall rate was 85.7 %, and the overall F value was 86.6 %. Its recognition time was 1.99 s. As for the SVM algorithm, the comprehensive precision was 74.2 %, the comprehensive recall rate was 73.2 %, and the overall F value was 73.7 %. The recognition time was 3.68 s.

Novelty

The novelty of this article lies in using two separate CNNs to extract convolutional features from acceleration and angular velocity, respectively. These features are then combined and reduced dimensionality using PCA, thereby improving both recognition accuracy and efficiency.

As basic education is increasingly digitized, the need for better teaching and learning quality also rises. Teaching reform is crucial to achieve this, and incorporating the Levenberg-Marquardt (L-M) into the Radial Basis Function (RBF) can help establish a fair online teaching evaluation system. The experimental results showed that the convergence ability of the model was significantly improved compared with the traditional RBF neural network. The overall mean square error of the improved model was 10°. The actual value prediction accuracy of the improved model is higher than that of the Backpropagation (BP). When the actual value was at its peak, the accuracy reached 98 %, the overall fluctuation range of absolute error was low, the highest absolute error value reached 0.78, and the average absolute error was below 0.5. With targeted improvements, teachers and students could better understand and change their own learning situations, as reflected in empirical evaluations.

This paper presents a Takagi-Sugeno (T-S) fuzzy type 1–2 controller for load frequency control (LFC). Most of the fuzzy controllers implemented for LFC in the past have Mamdani inference. Mamdani inferences give a fuzzy set that is defuzzified to form an output, whereas the T-S inference doesn't require defuzzification as its output is either constant or 1st order linear polynomial expression of inputs. T-S linear output dependency on its inputs helps it operate at different operating conditions of a dynamic nonlinear system. In this work, a combination of both constant and linear outputs for T-S fuzzy are used to implement a controller for LFC. A two-area tie-line power system is used for demonstration purposes. LFC has gained more importance with the introduction of deregulated renewable energy sources (RES) access to the grid. The proposed controller demonstrates higher stability for almost 30% load variation than its predecessors’ fuzzy controllers.

Regarding the problem of automatic detection in art teaching classroom behavior, the research combines the YOLOv5 algorithm in the deep learning algorithm and adds a two-way feature information pyramid function with weighting capability to the neck part of the algorithm to achieve performance-based algorithm improvement. This research prunes and optimizes the model for the campus technology implementation problem to improve the robustness and ease of implementation of the model. The model is designed in line with the model of the art teaching classroom behavior training set, and the applied experimental method is adopted for analysis. The results show that the average accuracy of all classes of state classification is 0.973 level after model improvement, the average accuracy of all classes of state classification is 0.970 level after model pruning, and the realizability of the model is significantly enhanced while the performance and efficiency are improved. Therefore, the research-designed classroom behavior detection and analysis model for art teaching can effectively detect the types of classroom behaviors of students in the process of art teaching with excellent performance, providing an effective way to ensure the quality of student learning in classroom teaching.