Person concentration recognition and evaluation system based on machine vision

Abstract

Disclosed in the present invention is a person concentration recognition and evaluation system based on machine vision. The technical method belongs to the field of computer vision and human-machine interaction in the field of artificial intelligence. During an assembly activity, the states of listeners and audiences are important links for feeding back the assembly effect, but there is no accurate scoring standard, and therefore the links are referred to as fuzzy links. Taking a party, a speech or a conference as an example, the evaluation system designed herein innovatively performs listener (audience) behavior discrimination across four dimensions, which are respectively expression screening, concentration recognition, facial distance monitoring and facial position monitoring; fuzzy evaluation can be better performed on the real-time states of the listeners (audiences) by means of a fuzzy algorithm; a particle swarm fuzzy algorithm scheme is provided to enable the state discrimination to break through the polarization limitation; and a membership function is set to define a fuzzy area, the degree of behaviors of the minority being different from behaviors of the majority is discriminated, and boxes of different colors are provided for marking.

Description

A person's concentration recognition and evaluation system based on machine vision Technical Field

[0001] The present invention relates to the fields of computer vision and human-computer interaction in the field of artificial intelligence, and in particular to a person concentration recognition and evaluation system based on machine vision. Background Art

[0002] The machine vision-based system for identifying and assessing human focus relies on multiple technologies, including face detection, eye tracking, expression analysis, posture recognition, and gaze direction estimation. Combined with deep learning and machine learning algorithms, it can analyze and assess an individual's focus level in real time. This system not only provides personalized learning feedback in education but also improves efficiency in workplace scenarios. By collecting and analyzing large amounts of data, the system provides users with valuable information, helping them better understand their focus status and make appropriate adjustments, ultimately achieving better results in their studies and work.

[0003] To address these issues, this patent proposes a machine vision-based person concentration recognition and assessment system. By collecting and processing real-time streaming data, it rapidly builds, optimizes, and deploys algorithm models, improving the efficiency and reliability of algorithm model reasoning:

[0004] 1. Lack of methods to assess concentration based on facial expressions: In the education sector: Low student engagement: Teachers who cannot determine students' concentration levels by observing their facial expressions may be unable to adjust their teaching strategies in a timely manner, resulting in reduced student engagement. Difficulties in personalized learning: The lack of accurate assessment of student concentration makes it more difficult to implement personalized learning plans, as teachers struggle to adjust teaching content and pace based on students' actual needs. Inaccurate assessment of learning outcomes: If educators cannot accurately assess students' interest and concentration through facial expressions, they may inaccurately assess their learning outcomes, thus affecting the quality of teaching. In the workplace: Decreased productivity: In the workplace, managers who cannot identify employees' concentration levels may be unable to effectively assign tasks or motivate employees, thus affecting overall productivity. Neglected employee well-being: The lack of monitoring employee concentration and stress levels may lead to overlooked mental health issues, which in the long term may increase stress and fatigue, affecting job satisfaction and well-being. Difficulties in personal development: For individuals, the lack of tools to monitor their concentration can make it difficult to identify shortcomings in their studies or work, thus hindering their ability to improve themselves. Inefficient time management: Individuals who are unable to accurately assess their focus may perform demanding tasks at inappropriate times, resulting in inefficient time management. Technology Development and Application: Limitations of Human-Computer Interaction: In the field of human-computer interaction, the lack of methods to assess user focus can lead to interaction designs failing to fully consider users' actual usage status, thereby reducing user experience.

[0005] 2. Lack of timely detection and prompt notification of lapses in focus: Reduced learning efficiency: Without immediate feedback and reminders, students or learners may not recognize their lapses in focus, resulting in inefficient learning. Prolonged periods of inefficient learning not only waste time but can also lead to poor learning outcomes. Decreased work productivity: In a workplace, employees who lack timely feedback on lapses in focus may continue to work inefficiently. This impacts not only individual performance but also the overall productivity of the team and organization. Ineffective time management: Without timely reminders of lapses in focus, individuals may miscalculate the time required to complete tasks, leading to poor time management. This can lead to excessive procrastination or misscheduling work and study plans, impacting the completion of other activities or tasks. Increased psychological stress: Prolonged periods of inattention, without timely adjustments and improvements, can lead to frustration and increased stress. This can lead to self-doubt, further impacting mental health. Stunted personal development: Focus is a key factor in personal growth and development. Without timely reminders and interventions for loss of concentration, individuals may miss the best opportunities to learn new skills and knowledge, thus affecting long-term career development and personal achievements. Decreased innovation and problem-solving abilities: Concentration is crucial for innovative thinking and solving complex problems.

[0006] 3. Lack of methods to quantify changes in focus: Obstacles to personal self-improvement: The lack of quantitative assessment tools for focus makes it difficult for individuals to accurately understand their focus levels in different environments or when performing different tasks. Consequently, individuals may be unable to effectively identify weaknesses in their focus, missing opportunities to improve and optimize their focus. Limitations to educational effectiveness: In the education sector, if teachers and educators cannot accurately assess students' focus, it will be difficult to provide personalized teaching plans. This can lead to inaccurate allocation of educational resources, affecting students' learning efficiency and effectiveness. Decreased work efficiency and productivity: In the workplace, if managers and teams cannot monitor employees' focus levels, it will be difficult to effectively arrange work tasks and rest time. Exacerbated mental health issues: Concentration issues are often associated with mental health conditions such as stress and anxiety. The lack of quantitative assessment methods for focus can prevent individuals from promptly identifying and addressing these mental health issues, exacerbating their negative impact. Limitations of the human-computer interaction experience: In human-computer interaction design, the lack of quantitative analysis tools for user focus can prevent designers from fully considering how users allocate their attention, impacting product usability and user experience.

[0007] 4. In the development of AI for gathering intelligence, there is a lack of AI learning methods that combine algorithms and principles to enable focus assessment: Insufficient Accuracy: Learning methods that lack a combination of algorithms and principles may result in inaccurate AI in assessing focus, failing to truly reflect the audience's level of focus, and thus impacting decision-making and adjustments. Poor Generalization: Without appropriate algorithms and principles for learning, AI may lack the ability to generalize to different scenarios and audience characteristics, making it incapable of adapting to the focus assessment needs of different situations. Difficulty in Optimization: Without guidance from algorithms and principles, AI learning methods may lack effective optimization paths, making it difficult to improve the accuracy and efficiency of focus assessment. Risk of Overfitting: Without the constraints of algorithms and principles, AI learning methods are prone to the risk of overfitting, with the learned model potentially overfitting the training data and failing to generalize effectively to new scenarios. Inability to Continuously Improve: Learning methods that lack a combination of algorithms and principles may hinder continuous improvement and iteration in AI's focus assessment capabilities, making it difficult to adapt to changing needs and environments. Therefore, an AI learning method that combines algorithms and principles to enable focus assessment is urgently needed. Technical issues

[0008] The main purpose of the present invention is to provide a person concentration recognition and evaluation system based on machine vision, which can effectively solve the problems in the background technology. Technical Solutions

[0009] To achieve the above object, the technical solution adopted by the present invention is:

[0010] The machine vision-based system for identifying and assessing human concentration consists of three main stages: the input layer, the computation module, and the output layer. In the input layer, the system uses convolutional neural network technology for facial recognition. This technology analyzes and processes images to identify faces and transmits them to the system. Next, in the computation module, the system employs four different techniques to distinguish facial expressions, concentration, and the distance and relative position between faces. These techniques include image classification, object detection and tracking, pose estimation, and deep learning. Through these techniques, the system can determine information such as the audience's expression, concentration level, and relative position to the speaker. Finally, in the output layer, the system marks qualified faces with a green box, unqualified faces with a red box, and questionable faces with a yellow box. The results are displayed on the screen or through a mobile app to alert participants in real time. The system also stores the captured facial recognition results as a time series dataset for later machine learning to output the overall recognition results.

[0011] Furthermore, the initial stage of the particle swarm algorithm is to capture feature data and generate a particle set, which is the input variable for the algorithm and the Hungarian genetic algorithm optimization; each example contains the following vector information [a1, a2, b1, b2, c1, c2, d1, d2]: The expression information collection: expression recognition status (a1): used to judge whether the participant's expression meets the requirements; whether it conforms to the public atmosphere (a2): used to judge whether the participant's expression adapts to the current environment; the concentration information collection: pupil recognition status (b1): used to evaluate the participant's concentration level, judged by the size and change of the pupil; whether it is facing the direction of the camera (b2 ): used to determine whether the participant is looking directly at the camera to indicate his or her concentration; face distance: face recognition status (c1): used to detect the participant's face and determine whether it is correctly identified; whether the distribution distance of adjacent faces is similar (c2): used to evaluate whether the distance between adjacent faces is within a reasonable range to determine whether there is a situation of being too dense or too dispersed; the face position information collection: face recognition status (d1): used to identify the participant's face and determine its position; whether the distance between the face and the shoulder or table is similar (d2): used to determine whether the distance between the participant's face and the shoulder or table is appropriate to ensure that the participant's posture meets the requirements.

[0012] Furthermore, according to the principle of triangular fuzzy numbers, this paper grades the semantic factors of the expert prior experience evaluation and uses fuzzy values ​​to represent different levels of evaluation; the evaluation levels from "very good" to "very poor" are represented by fuzzy values ​​1.0 to 0.2, where "very good" corresponds to a fuzzy value of 1.0, "good" corresponds to a fuzzy value of 0.8, "average" corresponds to a fuzzy value of 0.6, "poor" corresponds to a fuzzy value of 0.4, and "very poor" corresponds to a fuzzy value of 0.2; these fuzzy values ​​can help the system more accurately fuzzily evaluate the performance of the participants and provide corresponding feedback and grade judgment based on this. According to the changes in fuzzy values, the semantic factors in the expert prior experience evaluation can be graded; according to the principle of triangular fuzzy numbers, we can map fuzzy values ​​to corresponding evaluation levels; for example, a fuzzy value of 1.0 corresponds to "very good" and a fuzzy value of 0.8 corresponds to "good". According to these fuzzy values ​​and evaluation levels, corresponding grade divisions are provided. The algorithm preliminarily constructs the pheromone matrix of the particle swarm algorithm based on the fuzzified information of face recognition and uses the Hungarian genetic algorithm for multi-objective feedback optimization.

[0013] Furthermore, the eigenvalue vector is bundled with the data processed by the algorithm and the corresponding captured image and stored in the system background. This can effectively associate the data and the image so that the system can access and utilize the information at any time. At the same time, the system can also display the relevant information on the on-site large screen or the participant's mobile phone APP based on the evaluation results of the fuzzy evaluation link to remind participants to pay attention to the adjustment of assembly discipline and assembly effect, thereby promoting the quality and effect of the assembly. In addition, the system can also conduct individual and overall evaluations of the quality of the participants' assemblies throughout the semester based on each individual's time series status data set. The system can also give instructions on the potential improvement details of the participants and predict the speed of the participants' progress. These evaluation results will constitute evidence of the new digital evaluation of smart assemblies, providing strong support for the management and improvement of assemblies. In this way, the system can realize real-time monitoring and evaluation of assembly quality, providing a scientific basis for the improvement and optimization of assemblies.

[0014] Furthermore, in the initial stage of the particle swarm algorithm, feature data is captured and used to generate particle sets, which serve as input variables for the algorithm and Hungarian genetic algorithm optimization. Each example contains vector information such as expression information, concentration information, and face position information. Based on the status of these factors, the system divides the participant's situation into a green box (meets the requirements), a red box (does not meet the requirements), and a yellow box (doubtful or difficult to determine). The principle of triangular fuzzy numbers is used to grade the semantic factors evaluated by experts' prior experience, and fuzzy values ​​are used to represent different evaluation levels, such as "very good" corresponding to 1.0 and "very poor" corresponding to 0.2. These fuzzy values ​​help the system accurately evaluate the performance of participants and provide feedback and grade discrimination. The new algorithm constructs the pheromone matrix of the particle swarm algorithm based on the fuzzified information of face recognition, and uses the Hungarian genetic algorithm for multi-objective feedback optimization to obtain the optimal solution. It has the advantages of fast speed and low power consumption and can effectively evaluate assembly state behavior.

[0015] Furthermore, this paper adopts the principle of triangular fuzzy numbers to grade the semantic factors of expert prior experience evaluation, and maps different evaluation levels from "very good" to "very poor" to fuzzy values ​​1.0 to 0.2 respectively, so as to more accurately evaluate the performance of participants and provide feedback. Based on this, the system constructs the pheromone matrix of the particle swarm algorithm through face recognition fuzzification information, and applies the Hungarian genetic algorithm for multi-objective feedback optimization to quickly obtain the optimal solution and effectively evaluate the state behavior of the assembly. The eigenvalue vector is bound to the processed data and image for storage to achieve information association and access at any time. At the same time, the evaluation results are displayed on a large screen or mobile phone APP to promote the adjustment of assembly discipline and effect. The system can also evaluate the quality of participants' gatherings individually and as a whole, give improvement suggestions and predict the speed of progress, support the digital evaluation of smart gatherings, provide scientific support for management and improvement, realize real-time monitoring and improvement, and provide a scientific basis. Beneficial effects

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] 1. This invention aims to determine whether a person is focused based on facial changes: Improving learning and work efficiency: By promptly detecting and alerting individuals to changes in focus, individuals can quickly adjust their attention and reduce distractions, thereby improving learning and work efficiency. Promoting personalized learning and work environments: This technology can help educators and employers understand students' or employees' focus patterns, thereby designing more personalized learning plans and work schedules to meet the needs of different individuals. Enhancing time management skills: Monitoring focus and providing timely reminders can help individuals better manage their time and identify when specific tasks are most effective, thereby optimizing daily schedules and improving the quality of life and work. Reducing psychological stress and anxiety: By reducing inefficiencies and task delays caused by lapses in focus, individuals may experience less stress and anxiety, helping to maintain good mental health. Promoting self-awareness and self-improvement: This technology enables individuals to more clearly recognize changes in their focus, enhancing self-awareness and thus targetedly improving focus and self-management skills over the long term. Improving the human-computer interaction experience: When designing a human-computer interaction system, by monitoring the user's concentration and providing timely feedback, a more personalized and responsive user experience can be provided, thereby improving the product's attractiveness and user satisfaction. Promoting the development of intelligent technology: The implementation and application of this invention has promoted the development of related intelligent monitoring and analysis technologies, including machine learning, artificial intelligence, and biometrics, and has played a positive role in promoting scientific and technological innovation. Optimizing teamwork and organizational management: At the team and organizational level, by promptly identifying the concentration status of members, managers can more effectively assign tasks and arrange breaks, thereby improving teamwork efficiency and overall productivity.

[0018] 2. This invention aims to promptly detect lapses in concentration and provide prompt alerts: Rapid Response: Reducing system latency means the system can respond more quickly to commands or abnormal conditions. This is crucial for mining equipment requiring real-time control, ensuring timely adjustments to the equipment's operating status to avoid production interruptions or accidents. Efficient Operation: Reducing energy consumption not only helps reduce production costs but also minimizes environmental impact. By optimizing control strategies and equipment configuration, automated control methods can achieve efficient energy utilization and improve equipment operating efficiency. Improved Stability: Reducing system latency and energy consumption reduces the risk of equipment failure, thereby improving system stability. A stable system reduces maintenance costs and downtime, further improving production efficiency. Improved User Experience: For operators, a low-latency, efficient system means a smoother operating experience and higher work efficiency. Furthermore, reducing energy consumption helps reduce operating and maintenance costs, improving overall user satisfaction. Promoting Technological Innovation: With the continuous development of automated control methods, low latency and low energy consumption will become important areas of technological innovation.

[0019] 3. This invention aims to quantify changes in concentration: Objective Assessment: This quantification method provides an objective assessment method, reducing the bias of subjective judgment and making the monitoring and assessment of concentration more accurate and reliable. Easy Tracking and Comparison: Through quantification, individuals or managers can easily track changing trends in concentration and compare concentration levels over different time periods or under different conditions, helping to identify factors affecting concentration. Promoting Self-Awareness and Self-Management: This invention enables individuals to intuitively understand their concentration levels, enhancing self-awareness and adopting effective strategies for self-management and improvement, thereby improving concentration and productivity. Personalized Feedback and Intervention: The quantified data can serve as the basis for providing personalized feedback and developing intervention measures, helping individuals to conduct targeted training and adjustments based on their individual concentration characteristics. Optimizing Teaching and Training Methods: In the education and training fields, teachers and trainers can use quantified concentration data to evaluate the effectiveness of teaching methods and design learning plans for students or employees that are more suited to their concentration characteristics. Improving Team and Organizational Efficiency: For teams and organizations, understanding the concentration levels of their members can help to rationally allocate tasks and schedule work, thereby improving team collaboration efficiency and overall productivity. Supporting mental health and well-being: By monitoring and quantifying changes in concentration, possible mental health problems, such as attention deficit disorder (ADHD), can be identified in a timely manner, providing a basis for timely intervention and treatment.

[0020] 4. This invention aims to calculate the concentration level of a gathering using convolutional neural network technology based on the principle of triangular fuzzy numbers and Hungarian genetic algorithm optimization. The triangular fuzzy number principle can transform the complex problem of determining concentration into a fuzzy problem, allowing for consideration of varying degrees of concentration, making concentration calculation more flexible and accurate. Convolutional neural network technology is used to extract features. Convolutional neural networks (CNNs) have powerful feature extraction capabilities and can automatically learn and extract features related to concentration in gatherings, thereby better assessing the audience's concentration level. The Hungarian genetic algorithm is used to optimize parameters. The Hungarian genetic algorithm can efficiently search the parameter space and optimize the parameters of the convolutional neural network, achieving better performance in the concentration assessment task and improving the overall effectiveness of the system. Comprehensive consideration of multiple factors: By combining the triangular fuzzy number principle, convolutional neural network technology, and the Hungarian genetic algorithm, the influence of multiple factors on concentration, such as sound, image, and posture, can be comprehensively considered, improving the comprehensiveness and accuracy of concentration calculation. High scalability and adaptability: This method can adjust and optimize the model according to specific needs, flexibly adapting to different gathering scenarios and application scenarios, and has high scalability and adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] FIG1 is a flow chart of the system design of the present invention;

[0022] FIG2 is a distribution of trapezoidal function output values ​​of the present invention;

[0023] FIG3 is a flowchart of the algorithm optimization of the present invention;

[0024] FIG4 is a diagram showing the eigenvalue vectors of the fuzzy rating system of the present invention; Best Mode for Carrying Out the Invention

[0025] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0026] In the description of the present invention, it should be noted that the terms "upper," "lower," "inner," "outer," "front end," "rear end," "two ends," "one end," "the other end," and the like, indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0027] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "provided with," "connected," etc., should be understood in a broad sense. For example, "connected" may refer to a fixed connection, a detachable connection, or an integral connection; it may refer to a mechanical connection or an electrical connection; it may refer to a direct connection or an indirect connection through an intermediate medium; it may refer to internal communication between two components. Those skilled in the art will be able to understand the specific meanings of the above terms in the present invention in specific circumstances.

[0028] Example

[0029] Please refer to Figures 1-4, the present invention provides a technical solution:

[0030] The system for identifying and evaluating a person's concentration based on machine vision includes the following: the system is mainly divided into three stages: input layer, computing module and output layer. In the input layer, the system uses convolutional neural network technology for face recognition, which can analyze and process images to identify faces and transmit them to the system. In the computing module, the system uses four different technologies to distinguish expressions, concentration, distance between faces and relative positions, etc. These technologies include image classification, target detection and tracking, posture estimation, and deep learning. Through these technologies, the system can determine the expression, concentration level and relative position of the audience (audience) to the speaker. Finally, in the output layer, The system will mark the faces that meet the requirements as green boxes, those that do not meet the requirements as red boxes, and doubtful faces as yellow boxes, and display the results on the screen or remind participants in real time through the mobile phone APP. The system will also store the captured face recognition results as a time series data set for subsequent machine learning to output the overall recognition results. The initial stage of the particle swarm algorithm is to capture feature data and generate particle sets, which are the input variables for the algorithm and Hungarian genetic algorithm optimization; each example contains the following vector information [a1, a2, b1, b2, c1, c2, d1, d2]: Expression information collection: Expression recognition status (a1): used to determine whether the participant's expression meets the requirements Compliance with requirements; Whether it conforms to the public atmosphere (a2): used to judge whether the participant's expression adapts to the current environment; the concentration information collection: pupil recognition status (b1): used to evaluate the participant's concentration level, judged by the size and change of the pupil; Whether facing the camera direction (b2): used to determine whether the participant is looking directly at the camera to indicate his or her concentration; face distance: face recognition status (c1): used to detect the participant's face and judge whether it is correctly recognized; Whether the distribution distance of adjacent faces is similar (c2): used to evaluate whether the distance between adjacent faces is within a reasonable range, to judge whether there is too dense or too scattered; the face position information collection: Face recognition status (d1): used to identify the participant's face and determine its location; whether the distance between the face and the shoulder or table is similar (d2): used to determine whether the distance between the participant's face and the shoulder or table is appropriate to ensure that the participant's posture meets the requirements. Based on the principle of triangular fuzzy numbers, this paper classifies the semantic factors of the expert's prior experience evaluation and uses fuzzy values ​​to represent different levels of evaluation. The evaluation level ranges from "very good" to "very poor" with fuzzy values ​​of 1.0 to 0.2, respectively, where "very good" corresponds to a fuzzy value of 1.0, "good" corresponds to a fuzzy value of 0.8, "average" corresponds to a fuzzy value of 0.6, "poor" corresponds to a fuzzy value of 0.4, and "very poor" corresponds to a fuzzy value of 0.2; These fuzzy values ​​can help the system to more accurately evaluate the performance of participants, and provide corresponding feedback and grade judgment accordingly. According to the changes in fuzzy values, the semantic factors in the expert's prior experience evaluation can be graded; according to the principle of triangular fuzzy numbers, we can map fuzzy values ​​to corresponding evaluation grades; for example, a fuzzy value of 1.0 corresponds to "very good", and a fuzzy value of 0.8 corresponds to "good". According to these fuzzy values ​​and evaluation grades, corresponding grade divisions are provided. The algorithm preliminarily constructs the pheromone matrix of the particle swarm algorithm based on the fuzzy information of face recognition, and uses the Hungarian genetic algorithm for multi-objective feedback optimization. The eigenvalue vector is combined with the fuzzy matrix processed by the algorithm. The processed data and the corresponding captured images are bundled and stored in the system background, which can effectively associate the data and images so that the system can access and use this information at any time. The system can also display the relevant information on the large screen on site or the mobile phone APP of the participants based on the evaluation results of the fuzzy evaluation link to remind participants to pay attention to the adjustment of assembly discipline and assembly effect, thereby promoting the quality and effect of the assembly. The system can also conduct individual and overall evaluations of the quality of the participants' assemblies throughout the semester based on the time series status data set of each individual. The system can also give instructions on the potential improvement details of the participants and predict the speed of the participants' improvement. These evaluation results The results will constitute evidence for the new digital evaluation of smart assemblies, providing strong support for the management and improvement of assemblies. In this way, the system can realize real-time monitoring and evaluation of the quality of assemblies, providing a scientific basis for the improvement and optimization of assemblies. In the initial stage of the particle swarm algorithm, feature data is captured and used to generate particle sets. These sets serve as input variables for the algorithm and Hungarian genetic algorithm optimization. Each example contains vector information such as expression information, concentration information and face position information. According to the status of these factors, the system divides the participants' situations into green boxes (meeting the requirements), red boxes (not meeting the requirements) and yellow boxes (doubtful or difficult to determine), using the triangular fuzzy number principle. The semantic factors of the experts' prior experience evaluation are graded, using fuzzy values ​​to represent different evaluation levels, such as "very good" corresponding to 1.0 and "very poor" corresponding to 0.2. These fuzzy values ​​help the system accurately assess participant performance and provide feedback and grade discrimination. A new algorithm constructs a pheromone matrix for a particle swarm optimization algorithm based on the fuzzified information from face recognition and uses a Hungarian genetic algorithm for multi-objective feedback optimization to obtain the optimal solution. This algorithm is fast and low-power, and can effectively evaluate assembly state behavior. This paper uses the principle of triangular fuzzy numbers to grade the semantic factors of the experts' prior experience evaluation, mapping different evaluation levels from "very good" to "very poor" to fuzzy values ​​from 1.0 to 0, respectively.2. To more accurately assess participant performance and provide feedback, the system uses facial recognition to fuzzify information and construct a pheromone matrix for a particle swarm algorithm. It then applies a Hungarian genetic algorithm for multi-objective feedback optimization, quickly obtaining the optimal solution and effectively evaluating gathering status and behavior. Eigenvalue vectors are stored alongside processed data and images, enabling information association and ready access. Evaluation results are displayed on a large screen or mobile app, promoting adjustments to gathering discipline and effectiveness. The system also assesses the quality of gatherings individually and collectively, providing suggestions for improvement and predicting progress. This supports digital evaluation of smart gatherings, providing scientific support for management and improvement, enabling real-time monitoring and improvement, and providing a scientific basis.

[0031] It should be noted that the machine vision-based human concentration recognition and evaluation system, this patent relates to the technical field related to automated control on the industrial Internet. The method is mainly implemented through the following steps: The initial stage of the particle swarm algorithm is to capture feature data and generate a particle set, which is the input variable for the algorithm and the Hungarian genetic algorithm optimization. Each example contains the following vector information, [a1, a2, b1, b2, c1, c2, d1, d2]

[0032] The detailed explanation of the factors in the vector is shown in Table 1.

[0033]

[0034] Table 1 Explanation table of various factors of particle swarm optimization

[0035] (2) Calculation process

[0036] This paper classifies the semantic factors in the expert prior experience evaluation into very poor, poor, fair, good, and very good grades. Based on the principle of triangular fuzzy numbers, the fuzzy values ​​used to determine the grades in the four major indicators of the fuzzy evaluation system can be adjusted as shown in Table 2 below. The trigonometric function and fuzzy value output distribution are shown in Figure 2:

[0037]

[0038] Table 2 Prior experience fuzzy evaluation scale

[0039] (3) Optimize processes

[0040] The new algorithm takes advantage of its strengths and avoids its weaknesses. The basic idea is to preliminarily construct the pheromone matrix of the particle swarm algorithm based on the fuzzy information of face recognition, generate the initial solution set of the problem based on its randomness and strong global search ability, and then rely on the multi-objective feedback advantages of the Hungarian genetic algorithm [23-25] and strong iterative optimization ability to obtain the optimal solution to the problem, that is, to obtain the best evaluation results of the assembly state behavior at a relatively fast speed and relatively low power consumption.

[0041] (IV) Output results

[0042] At the output layer, the system labels the participants' faces based on the analysis results of the computational module: faces that meet the focus requirements are marked with a green frame, those that do not meet the requirements are marked with a red frame, and faces that are questionable are marked with a yellow frame. These results can be displayed on the screen or through a mobile app to provide real-time notifications to participants, allowing them to adjust their focus. Furthermore, the system stores the captured face recognition results as a time series dataset, facilitating subsequent analysis and summary of the overall recognition results using machine learning methods.

[0043] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A person concentration recognition and evaluation system based on machine vision, characterized by: The system is mainly divided into three stages: input layer, computing module and output layer. In the input layer, the system uses convolutional neural network technology for face recognition. This technology can analyze and process images to identify faces and transmit them to the system. In the computing module, the system uses four different technologies to distinguish expressions, concentration, distance between faces and relative positions, etc. These technologies include image classification, target detection and tracking, posture estimation, and deep learning methods. Through these technologies, the system can determine the expression, concentration level and relative position of the audience (audience) with the speaker. Finally, in the output layer, the system marks the faces that meet the requirements with green boxes, those that do not meet the requirements with red boxes, and doubtful faces with yellow boxes, and displays the results on the screen or reminds participants in real time through a mobile phone APP. The system will also store the captured face recognition results as a time series data set so that the overall recognition results can be output by machine learning in the later stage.

2. The machine vision-based person concentration recognition and evaluation system according to claim 1, characterized in that: The initial stage of the particle swarm algorithm is to capture feature data and generate a particle set, which is the input variable of the algorithm and the Hungarian genetic algorithm optimization; Each example contains the following vector information [a1, a2, b1, b2, c1, c2, d1, d2]: The expression information collection: Expression recognition state (a1): used to determine whether the participant's expression meets the requirements; Whether it is in line with the general atmosphere (a2): used to judge whether the participant's expression is suitable for the current environment; The concentration information collection: pupil recognition status (b1): used to evaluate the concentration of the participant, judged by the size and changes of the pupil; Whether facing the camera direction (b2): used to determine whether the participant is looking directly at the camera to indicate his / her concentration; face distance: face recognition status (c1): used to detect the participant's face and determine whether it is correctly recognized; whether the distribution distances of adjacent faces are similar (c2): used to evaluate whether the distances between adjacent faces are within a reasonable range to determine whether they are too dense or too dispersed; face position information collection: face recognition status (d1): used to identify the participant's face and determine its position; whether the distance between the face and the shoulder or table is similar (d2): used to determine whether the distance between the participant's face and the shoulder or table is appropriate to ensure that the participant's posture meets the requirements.

3. The machine vision-based person concentration recognition and evaluation system according to claim 1, characterized in that: According to the principle of triangular fuzzy numbers, the semantic factors of the expert prior experience evaluation are graded in this paper, and fuzzy values ​​are used to represent different levels of evaluation; the evaluation levels from "very good" to "very bad" are represented by fuzzy values ​​1.0 to 0.2, where "very good" corresponds to a fuzzy value of 1.0, "good" corresponds to a fuzzy value of 0.8, "average" corresponds to a fuzzy value of 0.6, "poor" corresponds to a fuzzy value of 0.4, and "very bad" corresponds to a fuzzy value of 0.2; These fuzzy values ​​can help the system to make a more accurate fuzzy evaluation of the performance of the participants, and provide corresponding feedback and grade judgment accordingly. According to the changes in fuzzy values, the semantic factors in the expert's prior experience evaluation can be graded; according to the principle of triangular fuzzy numbers, we can map the fuzzy values ​​to the corresponding evaluation grades; for example, a fuzzy value of 1.0 corresponds to "very good", and a fuzzy value of 0.8 corresponds to "better". According to these fuzzy values ​​and evaluation grades, corresponding grade divisions are provided. The algorithm preliminarily constructs the pheromone matrix of the particle swarm algorithm based on the fuzzified information of face recognition, and uses the Hungarian genetic algorithm for multi-objective feedback optimization.

4. The machine vision-based person concentration recognition and evaluation system according to claim 1, characterized in that: The eigenvalue vector is bundled with the data processed by the algorithm and the corresponding captured image and stored in the system background, which can effectively associate the data and the image so that the system can access and use the information at any time. The system can also display the relevant information on the large screen on site or the participant's mobile phone APP according to the evaluation results of the fuzzy evaluation link, so as to remind the participants to pay attention to the adjustment of the assembly discipline and the gathering effect, so as to promote the quality and effect of the assembly. The system can also conduct individual and overall evaluation of the quality of the participants' assemblies throughout the semester based on the time series state data set of each individual. The system can also give instructions on the potential improvement details of the participants and predict the speed of the participants' progress and improvement. These evaluation results will constitute the evidence of the new digital evaluation of smart assemblies, and provide strong support for the management and improvement of assemblies. In this way, the system can realize real-time monitoring and evaluation of the quality of assemblies, and provide a scientific basis for the improvement and optimization of assemblies.

5. The machine vision-based person concentration recognition and evaluation system according to claim 2, characterized in that: In the initial stage of the particle swarm algorithm, feature data is captured and used to generate particle sets, which are used as input variables for the algorithm and Hungarian genetic algorithm optimization. Each example contains vector information such as expression information, concentration information and face position information. According to the status of these factors, the system divides the participants' situations into green boxes (meeting the requirements), red boxes (not meeting the requirements) and yellow boxes (doubtful or difficult to determine). The principle of triangular fuzzy numbers is used to grade the semantic factors evaluated by experts' prior experience, and fuzzy values ​​are used to represent different evaluation levels, such as "very good" corresponds to 1.0 and "very poor" corresponds to 0.

2. These fuzzy values ​​help the system accurately evaluate the performance of participants and provide feedback and grade discrimination. The new algorithm constructs the pheromone matrix of the particle swarm algorithm based on the fuzzified information of face recognition, and uses the Hungarian genetic algorithm for multi-objective feedback optimization to obtain the optimal solution. It has the advantages of fast speed and low power consumption, and can effectively evaluate the assembly state behavior.

6. The machine vision-based person concentration recognition and evaluation system according to claim 5, characterized in that: This paper adopts the principle of triangular fuzzy numbers to grade the semantic factors of experts' prior experience evaluation, and maps different evaluation levels from "very good" to "very poor" to fuzzy values ​​1.0 to 0.2, so as to more accurately evaluate the performance of participants and provide feedback. Based on this, the system constructs the pheromone matrix of the particle swarm algorithm through face recognition fuzzification information, and applies the Hungarian genetic algorithm for multi-objective feedback optimization to quickly obtain the optimal solution and effectively evaluate the assembly state behavior. The eigenvalue vector is bound and stored with the processed data and images to achieve information association and access at any time. At the same time, the evaluation results are displayed on a large screen or mobile phone APP to promote the adjustment of assembly discipline and effect. The system can also evaluate the quality of participants' gatherings individually and as a whole, give improvement suggestions and predict the speed of progress, support the digital evaluation of smart gatherings, provide scientific support for management and improvement, realize real-time monitoring and improvement, and provide a scientific basis.

Images