This paper carries out a research on patients’ lower limb posture capture strategy based on the lower limb rehabilitation of patients with sports function injury. The study is based on the posture filtering algorithm and designed a lower limb joint localization model based on the quaternion Kalman filter. The model utilizes five IMUs to capture the patient’s lower limb movements to determine the posture of the patient’s critical limbs in three-dimensional space and establish the joint coordinate system. Based on the filtered pose quaternions, the joint coordinate system of the lower limb is solved to obtain the optimal estimation of the lower limb pose. The results of simulation experiments show that the algorithm of this paper can make the motion data smoother and satisfy the motion requirements. The valuation of this paper’s algorithm on the Z-axis in the single-axis rotation experiment is stable from – 90° to 90°, while the valuation on the X-axis and Y-axis is near 0°. And the error in the ankle motion trajectory is small, with a mean value of 1.36°. The example results illustrate that the rehabilitation system equipped with the algorithm of this paper is basically consistent with the thigh elevation curve of the optical method in the patient’s lower limb motion monitoring during walking, and the error is within 6°. The research in this paper provides a new technical means for lower limb rehabilitation training, which helps to improve the personalization and precision of rehabilitation training.

This paper is based on the definition of novel distribution system panoramic perception technology under the perspective of generative artificial intelligence. The preprocessed data are put into forward GRU neurons and reverse GRU neurons as model input variables for multi-task assisted training, and the model outputs distribution system perception results to complete the task of constructing a new distribution system panoramic perception model based on BiGRU. When the distribution system current and voltage data is zero, it will lead to a reduction in the current and voltage prediction accuracy of the distribution system of the ELM model, for this reason, it is proposed to use the genetic algorithm to optimize the ELM model, to achieve the modeling of the new distribution system prediction model based on the ELM-GA algorithm. Using the model constructed in this paper, panoramic perception and prediction analysis of the new distribution system is carried out. When the BiGRU model is deployed in the new distribution system, the BiGRU network’s system perception accuracy and error rate are 95.00% and 5.00%, respectively, which fully meets the user experience requirements of the new distribution system, and the relative errors of fault voltage and fault current prediction based on the ELM-GA algorithm for the new distribution system are less than 5%, which indicates that the ELM-GA distribution system prediction model has the characteristics of high robustness and high accuracy.

This paper analyzes public interest litigation and its salient features, and organizes the audit rules for the electronic transformation of litigation evidence. Aiming at the phenomenon of varying text length in litigation evidence, a joint CTC-Attention decoding model (HCADecoder) based on bigram hybrid labeling is proposed. Based on the existing research on computer vision technology for target number prediction, the stacked object occlusion problem existing in special scenes is proposed, and an algorithm for predicting the number of stacked objects combining planar density map and depth map is proposed. Combined with the public interest litigation evidence document corpus dataset, we analyze the recognition of basic elements of litigation evidence by text label recognition algorithm, and select the commonly used precision rate P, recall rate R and F1 value to evaluate the recognition results of basic elements. Subdivide the text length of litigation evidence and analyze the recognition accuracy of each algorithm on different text lengths. Bring the text label recognition algorithms into real cases to analyze the element extraction. For this paper, we propose monocular image target counting algorithm, which is brought into different scenarios for performance testing. This paper proposes text label recognition algorithm with evidence image target counting algorithm for litigation evidence text image recognition with mean value at 80%.

Innovation and entrepreneurship, as an important part of social and economic activities, has received more and more widespread attention. Based on the characteristics of the digital era, the study uses artificial intelligence to empower innovation and entrepreneurship education in colleges and universities. Optimize the allocation of innovation and entrepreneurship education resources in colleges and universities through multi-objective optimization algorithm. Construct an optimization model of resource allocation for innovation and entrepreneurship education in colleges and universities, and verify its resource optimization and allocation performance. Taking 13 colleges and universities in C city as the research object, the optimization of their innovation and entrepreneurship education resource allocation is processed. The MSS cumulative values of this paper’s multi-objective optimization model on the CPLX problem and the MATP problem are -1.400 and -1.033, respectively, which are the smallest among all models, with the best performance and ranked the first in resource allocation efficiency. After optimization, the resource allocation level of innovation and entrepreneurship education in all 13 colleges and universities has been improved, and the resource allocation among the colleges and universities is more balanced.The resource utilization efficiency of innovation and entrepreneurship education in the 13 colleges and universities has been improved by 17.02% on average.

The article aims to accelerate the growth and progress of young teachers in private applied colleges and universities and improve their teaching ability, combining with the knowledge graph, and designing a recommended algorithm based on deep reinforcement learning to improve teachers’ ability. Firstly, the growth and progress process of young teachers in private applied colleges and universities is defined as a dynamic development process, i.e., for different latitude abilities such as teacher ethics, professional knowledge, preteaching preparation, communication and cooperation, teaching ability training needs to be carried out gradually and in a certain order. Then the Knowledge Graph Teacher Competency Enhancement Recommendation Algorithm (KGDR) based on deep reinforcement learning and knowledge graph algorithm is constructed by combining deep reinforcement learning and knowledge graph algorithm. When performing top-𝑘 recommendation, the diversity value of the model at 𝑘 = 20 is 0.7876, and the model can provide more diverse paths for teacher ability improvement. After the application of the dynamic development mechanism of young teachers’ competence based on KGDR, the competence improvement of young teachers is significant and can reach the grade of “excellent”. The mechanism designed in this paper can be used as a reference for other colleges and universities.

The purpose of this study is to evaluate the comprehensive ability of students objectively by constructing the evaluation system of compound music talents based on multi-objective planning, so as to promote the quality improvement and excellent cultivation of compound music talents in higher vocational colleges. The selected evaluation indexes of composite music talent cultivation are empowered by using the combination assignment method, and the construction of multi-objective planning model for cross-border composite music talent cultivation is realized based on the setting of objective function, constraints and model solving method. The article forms an index system covering 6 dimensions and 24 indicators, successfully divides the interval length of five evaluation levels, and obtains the distribution of students in each level, with the largest proportion of students in level 3, which is 40.77%. In addition, the ratings of the level 1 indicators are 2.47 to 3.31, which are in the middle to lower level. According to the student groups of different grades and the evaluation results of the indicators, we can clarify the level of student cultivation, improve the music talent cultivation system, coordinate and improve the elements and resources of each dimension, and promote the cultivation of cross-boundary composite music talents in higher vocational colleges and universities.

The article uses the appropriate equipment for research data, designing the face and physiological signal emotion recognition network respectively, and putting its recognition features into the random forest classifier for training in order to realize the construction work of emotion recognition model. In-depth interpretation of the random forest algorithm based emotion recognition model in the application of information systems, combined with the research data, respectively, the emotion recognition model and system safety performance testing assessment. The emotion recognition model of this paper based on the 25% retention method has a recognition rate of 96.16% for the 14- dimensional B emotion features, which has the highest recognition efficacy and can well meet the system emotion recognition needs. The experimental group is found to be significantly different from the control group, and it is concluded that by introducing the emotion recognition model into the traditional information system, all three security performance indicators of the system are significantly improved.

In enterprise operations, multi-objective optimization involves multiple conflicting objectives such as cost escalation control, customer satisfaction, and production efficiency. Based on reinforcement learning algorithm, the article deals with multi-objective optimization problem in enterprise operation through the interactive learning between intelligent body and environment, for which a multi-objective operation efficiency improvement path for enterprise based on Q-learning scheduling is designed. The simulation data is utilized to generate the PDR tree structure, and subsequently, the intelligent body is prompted to complete the multi-objective operation learning of the enterprise through several iterations. On this basis, the intelligent body completes all the actions and generates scheduling strategies to improve operational efficiency. The model proposed in this paper can predict the demand changes of enterprises in the future time window and make the best decision to improve the operational efficiency. Under the model of this paper, the mean values of pure technical efficiency as well as scale efficiency of 10 firms in 2024 are 0.9 and 0.933, respectively, and they are predicted to continue to grow in 2025. The model reduces the firms’ average operating costs and administrative expenses, while employee compensation and fixed assets increase by 49.58% and 19.48%. Since the survey period, the TFP index of all 10 companies is greater than 1, which indicates that, the application of the model in this paper improves the operational efficiency of the companies.

This paper establishes a specific path for the realization of AI-enhanced learning on the content of Civic and Political Education, starting from the relevance, quality, novelty and intuitiveness of the teaching content. Through HTML parsing and other crawler technology to obtain the Civics education data on the news network, and extract the data characteristics of the Civics material, using the clustering rule algorithm, to classify the material. Decision tree calculation based on random forest is performed to dynamically expand and integrate the material, on this basis, using reinforcement learning recommendation algorithm, the Civic and political education content recommendation model is constructed, and the recommendation results of the algorithm are verified using simulation experiments. The experimental results show that the average success rate of the research-designed recommendation algorithm in the last 10 groups of experimental data is 25.218%, which is higher than that of the MK recommendation algorithm (18.03%), and the average time of the research-designed recommendation algorithm in the last 10 groups of data is 5.095s, which is more efficient than that of the MK recommendation algorithm (11.903s). After integrating the enhanced learning content recommendation in the Civics education, the students’ humanism scale score was 100.56±12.364, with a p-value of less than 0.05, which was significantly higher than that before teaching.

In response to the greening and decarbonization of economic development and in search of a path to improve the corporate efficiency of resource-consuming enterprises, the study explores the impact of the financial sharing model on the efficiency of resource-consuming enterprises. The research hypothesis is formulated after the preliminary analysis of related theories such as financial sharing and accounting information. After completing the selection of research samples and data collection, the research variables are defined, the regression analysis model of the impact of financial sharing model on enterprise efficiency is constructed, and empirical analysis is conducted. The research hypotheses proposed in the previous section are verified through regression analysis. Monte Carlo method is used to simulate the financial sharing model and resource-consuming enterprise efficiency, and the net present value of resource-consuming enterprises is simulated during the construction period and the operation period of the financial sharing model, respectively, so as to understand their enterprise efficiency. The results of the empirical study show that financial sharing can realize the improvement of enterprise efficiency. Enterprise efficiency can increase with the improvement of accounting information transparency and accounting information consistency. During the construction and operation periods of the financial sharing model, the mean enterprise NPV after five years of operation is $608.4 and $2,327.4 million, respectively, and the probability of positive NPV is 68% and 94%, respectively.