A dance education assessment method and system

Abstract

The application discloses to the technical field of dance education, specifically a dance education evaluation method and system, first collecting student dance basic video, human body key points, physiological and emotional multi-dimensional data, synchronously recording student, track and scene correlation information; then integrating and standardizing the collected data, extracting action, rhythm, emotion, physiological adaptation core features and assigning weights; then evaluating the extracted features in single dimension and multi-dimension, generating personalized evaluation labels, benchmark comparison and progress tracking reports, and completing layered quantitative evaluation; based on this, the application has the advantages of realizing efficient dance education evaluation process, accurate action detail capture, greatly improving evaluation efficiency and execution consistency, and adapting to large-scale, normalized and multi-scene dance education evaluation.

Description

Technical Field

[0001] This invention relates to the field of dance education technology, specifically to a dance education assessment method and system. Background Technology

[0002] As a core branch of arts education, the scientific and objective assessment of dance education's teaching quality and learning outcomes directly impacts the accuracy of dance talent cultivation. Currently, dance education assessments largely rely on teachers' subjective experience and lack a standardized, systematic evaluation system, making it difficult to comprehensively and accurately reflect students' dance skills, artistic expression, and learning weaknesses. With the deep integration of artificial intelligence technology into dance education, there is an urgent need for an assessment method that is creative, logical, and practical, breaking through the limitations of traditional assessments and achieving digitalization, quantification, and personalization of the assessment process. This would provide a scientific basis for optimizing dance teaching and targeted improvement for students, promoting the transformation of dance education towards precision and intelligence.

[0003] However, existing dance education assessment methods have many shortcomings and are unable to meet the assessment needs of modern dance education. Specific problems include outdated assessment methods, lack of digital support, and low efficiency: traditional assessments often rely on manual observation and scoring, which cannot accurately quantify and analyze dance movements. Furthermore, the assessment process is time-consuming, labor-intensive, and inefficient, making it difficult to adapt to the needs of large-scale, routine dance education assessments. For example, in a dance program at an art college, conducting a final assessment for 50 students requires three teachers to watch students' dance videos and manually record scores for three consecutive days. This is not only time-consuming, but manual observation also fails to capture subtle deviations in movements (such as joint angle errors or incorrect force application), resulting in insufficient assessment accuracy and an inability to provide precise error correction guidance for students. Simultaneously, existing assessments do not utilize digital technologies such as motion capture and AI analysis, making it difficult to achieve long-term tracking and dynamic updates of assessment data. Summary of the Invention

[0004] To address the technical problems of outdated evaluation methods, lack of digital support, and low efficiency, this invention provides the following technical solution:

[0005] A method for assessing dance education, comprising the following specific steps:

[0006] S1, Multimodal assessment data collection: Collect students' basic dance videos, human body key points, physiological and emotional multidimensional data, and simultaneously record information related to students, songs and scenes, and output multimodal assessment dataset;

[0007] S2, Data Preprocessing and Feature Extraction: The multimodal evaluation dataset output from S1 is cleaned, integrated and standardized, and core features of action, rhythm, emotion and physiological adaptation are extracted and weighted to output a weighted core evaluation feature set.

[0008] S3, Stratified Quantitative Assessment:

[0009] S31, Single-dimensional quantitative scoring: Based on the weighted core evaluation feature set output by S2, perform single-dimensional quantitative scoring on each type of core evaluation feature and output a single-dimensional quantitative scoring table.

[0010] S32, Multi-dimensional Comprehensive Scoring: Based on the single-dimensional quantitative scoring table output by S31, the multi-dimensional comprehensive evaluation score of the trainee is calculated using a weighted summation formula. At the same time, the evaluation level is divided according to the comprehensive score, and the deviation of each dimension score from the standard score is calculated simultaneously. The comprehensive evaluation score, evaluation level and dimension deviation analysis report are output.

[0011] S33: Personalized assessment tag generation. Based on the comprehensive assessment score, assessment level, and dimensional deviation analysis report output in S32, and combined with the student's basic information, personalized assessment tags are generated for the student. At the same time, combined with the assessment data of similar students, student ranking and progress analysis are generated, and a set of personalized assessment tags and a ranking analysis report are output.

[0012] S34, Benchmarking and Progress Tracking: Based on the personalized evaluation tag set and ranking analysis report output in S33, the evaluation data of outstanding students in the same dance style and age group are introduced as industry benchmarks to compare the core characteristic differences between students and benchmarks; at the same time, the historical evaluation data of students are retrieved to track the changing trends of scores in each dimension, and a benchmarking report and progress tracking curve are output.

[0013] S4, Evaluation Feedback and Dynamic Optimization: Based on the benchmark comparison report and progress tracking curve output by S34, generate personalized improvement plans and verify their feasibility, while providing teaching optimization suggestions, and then dynamically optimize the evaluation system parameters and feature library.

[0014] As a preferred embodiment of the dance education assessment method described in this invention, the specific steps of step S1 are as follows:

[0015] S11, Basic Movement Video Acquisition: Set up a multi-camera acquisition scene to simultaneously acquire video data of complete dance segments of students; at the same time, record acquisition scene information, student basic information and dance music information, and output student basic dance video data.

[0016] S12, Human Key Point Data Acquisition: Based on the basic video data output in S11, the Movenet human pose estimation algorithm is used to extract multiple human key node data in the student's dance process in real time. At the same time, the spatial coordinates, motion trajectory and angle change data of each key node are calculated, and abnormal data corresponding to blurred frames and occluded frames are removed. The effective data is denoised and a standardized human key point motion dataset is output.

[0017] S13, Physiological and Emotional Data Acquisition: Based on the standardized human key point motion dataset output in S12, combined with the basic video data output in S11, the physiological and emotional data of the trainees are collected simultaneously, and a multimodal assessment dataset is output.

[0018] As a preferred embodiment of the dance education assessment method described in this invention, the specific steps of step S2 are as follows:

[0019] S21, Multimodal data cleaning and integration: Based on the multimodal evaluation dataset output by S1, the video data, human key point data, physiological data, and emotional data are first cleaned, and then the video data, human key point data, physiological data, and emotional data are aligned on the time axis and integrated into a unified evaluation data matrix, outputting a standardized integrated dataset.

[0020] S22, Core Evaluation Feature Extraction: Based on the standardized integrated dataset output in S21, extract four categories of core evaluation features: action features, rhythm features, emotion features, and physiological adaptation features, which correspond to the key dimensions of dance evaluation; and output the core evaluation feature vector.

[0021] S23, Feature weight allocation: Based on the core evaluation feature vector output in S22, and combined with the evaluation focus of different dance styles, the weight of each feature is allocated using the analytic hierarchy process; at the same time, the weights are dynamically fine-tuned by combining the student's basic information, and a weighted core evaluation feature set is output.

[0022] As a preferred embodiment of the dance education assessment method described in this invention, the specific steps of step S4 are as follows:

[0023] S41, Personalized Improvement Plan Generation: Based on the benchmark comparison report and progress tracking curve output in S34, a targeted personalized improvement plan is generated and output;

[0024] S42, Feasibility verification of the improvement plan: Based on the targeted and personalized improvement plan output in S41, verify the feasibility and adaptability of the plan; at the same time, combine the students' learning time and acceptance ability, fine-tune the practice content and cycle, and output the optimized personalized improvement plan and feasibility verification report.

[0025] S43, Generating Teaching Optimization Suggestions: Based on the optimized personalized improvement plan and feasibility verification report output in S42, and combined with the overall class evaluation data, generating teaching optimization suggestions for the instructors and outputting them;

[0026] S44, Dynamic Optimization of the Evaluation System: Based on the teaching optimization suggestions output in S43, combined with the optimized personalized improvement plan and feasibility verification report output in S42, the core parameters of the evaluation are dynamically optimized. At the same time, in combination with new dance styles and new teaching needs, the evaluation dimensions and feature types are expanded, and the optimized evaluation parameters and feature library update report are output.

[0027] A dance education assessment system, comprising:

[0028] The multimodal assessment data acquisition module collects students' basic dance videos, key points of the human body, and multidimensional data on physiology and emotion. It also records information related to students, songs, and scenes, and outputs a multimodal assessment dataset.

[0029] The data preprocessing and feature extraction module cleans, integrates, and standardizes the multimodal evaluation dataset output by the multimodal evaluation data acquisition module, extracts core features of action, rhythm, emotion, and physiological adaptation, assigns weights, and outputs a weighted core evaluation feature set.

[0030] The tiered quantitative assessment module includes a single-dimensional quantitative scoring unit, a multi-dimensional comprehensive scoring unit, a personalized assessment label generation unit, and a benchmarking and progress tracking unit.

[0031] The single-dimensional quantitative scoring unit performs single-dimensional quantitative scoring on each type of core evaluation feature based on the weighted core evaluation feature set output by the data preprocessing and feature extraction module, and outputs a single-dimensional quantitative scoring table.

[0032] The multi-dimensional comprehensive scoring unit, based on the single-dimensional quantitative scoring table output by the single-dimensional quantitative scoring unit, uses a weighted summation formula to calculate the trainee's multi-dimensional comprehensive evaluation score, and simultaneously classifies the evaluation level according to the comprehensive score; and simultaneously calculates the deviation between the score of each dimension and the standard score, and outputs a comprehensive evaluation score, evaluation level and dimension deviation analysis report.

[0033] The personalized assessment tag generation unit generates personalized assessment tags for students based on the comprehensive assessment score, assessment level, and dimensional deviation analysis report output by the multi-dimensional comprehensive scoring unit, combined with the students' basic information; at the same time, it generates student ranking and progress analysis by combining assessment data of similar students, and outputs a set of personalized assessment tags and a ranking analysis report.

[0034] The benchmarking and progress tracking unit, based on the personalized evaluation tag set and ranking analysis report output by the personalized evaluation tag generation unit, introduces the evaluation data of outstanding students in the same dance style and age group as industry benchmarks, and compares the core characteristic differences between students and benchmarks; at the same time, it retrieves the student's historical evaluation data, tracks the changing trends of scores in each dimension, and outputs a benchmarking report and progress tracking curve.

[0035] The evaluation feedback and dynamic optimization module, based on the benchmark comparison report and progress tracking curve output by the benchmark comparison and progress tracking unit, generates personalized improvement plans and verifies their feasibility, while providing teaching optimization suggestions, thereby dynamically optimizing the evaluation system parameters and feature library.

[0036] As a preferred embodiment of the dance education assessment system described in this invention, the multimodal assessment data acquisition module includes:

[0037] The basic movement video capture unit sets up a multi-camera capture scene to simultaneously capture video data of complete dance segments of students; at the same time, it records the capture scene information, student basic information and dance song information, and outputs the student's basic dance video data.

[0038] The human body key point data acquisition unit, based on the basic video data output by the basic motion video acquisition unit, uses the Movenet human pose estimation algorithm to extract multiple human body key node data in real time during the student's dance process. At the same time, it calculates the spatial coordinates, motion trajectory and angle change data of each key node, removes abnormal data corresponding to blurred frames and occluded frames, performs noise reduction processing on the effective data, and outputs a standardized human body key point motion dataset.

[0039] The physiological and emotional data acquisition unit, based on the standardized human key point motion dataset output by the human key point data acquisition unit and combined with the basic video data output by the basic motion video acquisition unit, simultaneously acquires the physiological and emotional data of the trainees and outputs a multimodal assessment dataset.

[0040] As a preferred embodiment of the dance education assessment system described in this invention, the data preprocessing and feature extraction module includes:

[0041] The multimodal data cleaning and integration unit, based on the multimodal evaluation dataset output by the multimodal evaluation data acquisition module, first cleans the video data, human key point data, physiological data, and emotional data, then aligns the video data, human key point data, physiological data, and emotional data along the time axis, integrates them into a unified evaluation data matrix, and outputs a standardized integrated dataset.

[0042] The core evaluation feature extraction unit extracts four categories of core evaluation features—action features, rhythm features, emotion features, and physiological adaptation features—based on the standardized integrated dataset output by the multimodal data cleaning and integration unit. These features correspond to the key dimensions of dance evaluation, and the core evaluation feature vector is output.

[0043] The feature weight allocation unit, based on the core evaluation feature vector output by the core evaluation feature extraction unit, and in combination with the evaluation focus of different dance styles, uses the analytic hierarchy process to allocate the weight of each feature; at the same time, it dynamically fine-tunes the weights by combining the student's basic information, and outputs a weighted core evaluation feature set.

[0044] As a preferred embodiment of the dance education assessment system described in this invention, the assessment feedback and dynamic optimization module includes:

[0045] The personalized improvement plan generation unit generates and outputs targeted personalized improvement plans based on the benchmark comparison report and progress tracking curve output by the benchmark comparison and progress tracking unit.

[0046] The improvement scheme feasibility verification unit verifies the feasibility and adaptability of the personalized improvement scheme output by the personalized improvement scheme generation unit; at the same time, it fine-tunes the practice content and cycle based on the student's learning time and acceptance ability, and outputs the optimized personalized improvement scheme and feasibility verification report.

[0047] The teaching optimization suggestion generation unit generates and outputs teaching optimization suggestions for teachers based on the optimized personalized improvement plan and feasibility verification report output by the improvement plan feasibility verification unit, combined with the overall class evaluation data.

[0048] The dynamic optimization unit of the evaluation system generates teaching optimization suggestions based on the teaching optimization suggestions, combines the optimized personalized improvement scheme and feasibility verification report output by the improvement scheme feasibility verification unit, dynamically optimizes the core parameters of the evaluation, and expands the evaluation dimensions and feature types in combination with new dance styles and new teaching needs, outputting optimized evaluation parameters and feature library update reports.

[0049] Compared with existing technologies:

[0050] 1. By adopting digital technologies such as multi-camera high-definition video acquisition, automatic extraction of human body key points, AI posture analysis, and automated batch data processing, this method replaces the lagging methods of traditional manual observation, subjective scoring, and manual recording. It reduces human intervention and repetitive operations throughout the process, achieving high efficiency in the dance education assessment process, accurate capture of movement details, and significantly improving assessment efficiency and consistency. It is also suitable for large-scale, routine, and multi-scenario dance education assessments.

[0051] 2. By collecting multimodal quantitative data, standardizing and comparing AI algorithms, and cross-validating multidimensional data, it breaks away from the reliance on teachers' subjective experience in evaluation, and achieves unified and standardized evaluation standards, objective and fair results, significantly reduces human judgment bias, and comprehensively enhances the credibility of evaluation results.

[0052] 3. By extracting features and applying tiered weighted assessments across four core dimensions—movement standardization, rhythm control, emotional expression, and physiological adaptation—it breaks through the limitations of traditional single-movement standard assessments. This approach achieves comprehensive coverage of dance professional skills and artistic literacy, and provides a three-dimensional reflection of students' overall abilities and development potential.

[0053] 4. Through a progressive closed-loop design that integrates data collection, processing, evaluation, and optimization, data is reused at each stage and logically connected. This enables precise feedback of evaluation results to students' personalized learning improvement and teachers' teaching plans to be optimized, truly empowering the entire dance teaching process. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the overall framework of the present invention;

[0055] Figure 2 This is a schematic diagram of the framework of the multimodal evaluation data acquisition module of the present invention;

[0056] Figure 3 This is a schematic diagram of the data preprocessing and feature extraction module framework of the present invention;

[0057] Figure 4 This is a schematic diagram of the hierarchical quantitative evaluation module framework of the present invention;

[0058] Figure 5 This is a schematic diagram of the evaluation feedback and dynamic optimization module framework of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0060] This invention provides a method for assessing dance education, comprising the following specific steps:

[0061] S1, Multimodal assessment data collection: Collect students' basic dance videos, human body key points, physiological and emotional multidimensional data, and simultaneously record information related to students, songs and scenes, and output multimodal assessment dataset;

[0062] The specific steps of S1 are as follows:

[0063] S11, Basic Movement Video Acquisition: A multi-camera acquisition scene is set up, using 3 high-definition cameras (located at the front, side, and back respectively) to simultaneously acquire video data of the student's complete dance segment. The acquisition parameters are set to: 1080P resolution, 30 frames per second. During the acquisition process, it is ensured that the student's entire body is in the shot, without obstruction or blurring. At the same time, the acquisition scene information (such as venue size, light intensity), the student's basic information (such as age, dance style, and learning time) and dance music information (such as music name and tempo) are recorded, and the student's basic dance video data is output.

[0064] S12, Human Key Point Data Acquisition: Based on the basic video data output in S11, the Movenet human pose estimation algorithm is used to extract multiple human key node data (including head, neck, shoulder, elbow, wrist, hip, knee, ankle, etc.) in real time during the student's dance process. At the same time, the spatial coordinates, motion trajectory and angle change data of each key node are calculated, and abnormal data corresponding to blurred frames and occluded frames are removed. The effective data is denoised and a standardized human key point motion dataset is output.

[0065] S13, Physiological and Emotional Data Acquisition: Based on the standardized human key point motion dataset output in S12, combined with the basic video data output in S11, the physiological and emotional data of the trainees are collected simultaneously: Physiological data is collected through wearable devices, including heart rate and muscle strength (focusing on arm, waist, and leg muscles), and the physiological data time-series curve is output; Emotional data is obtained through facial expression recognition in video frames, extracting the trainees' facial feature points, and combining the amplitude of dance movements and rhythmic fit to determine the trainees' emotional expression state (such as joy, sadness, excitement), outputting emotional state labels and corresponding confidence levels, and outputting a multimodal evaluation dataset.

[0066] S2, Data Preprocessing and Feature Extraction: The multimodal evaluation dataset output from S1 is cleaned, integrated and standardized, and core features of action, rhythm, emotion and physiological adaptation are extracted and weighted to output a weighted core evaluation feature set.

[0067] The specific steps of S2 are as follows:

[0068] S21, Multimodal Data Cleaning and Integration: Based on the multimodal evaluation dataset output in S1, the video data, human keypoint data, physiological data, and emotional data are first cleaned. Specifically, the video data undergoes frame filtering, retaining valid frames with clear movements and no abnormalities, while deleting invalid frames that are blurry, occluded, or have interrupted movements. Outlier removal is performed on the human keypoint data, and interpolation is used to supplement missing data to ensure data continuity. Physiological and emotional data are standardized, converting data from different dimensions into standardized scores of 0-100. Subsequently, the video data, human keypoint data, physiological data, and emotional data are time-axis aligned and integrated into a unified evaluation data matrix, outputting a standardized integrated dataset.

[0069] S22, Core Evaluation Feature Extraction: Based on the standardized integrated dataset output in S21, four categories of core evaluation features are extracted: motion features, rhythm features, emotional features, and physiological adaptation features. These correspond to the key dimensions of dance evaluation: ① Motion features: Based on human keypoint data, motion standardization (deviation from standard motion), motion fluency (continuity of motion trajectory at key nodes), and force application rationality (muscle strength and motion matching); ② Rhythm features: Based on the comparison between video data and music beats, rhythm fit (synchronization rate between motion and beat) and rhythm change adaptability (ability to adjust motion to changes in music rhythm); ③ Emotional features: Based on emotional state labels and facial expression data, emotional transmission (match between emotional labels and the emotion of the dance piece) and emotional expressiveness (coordination between facial expressions and motion amplitude); ④ Physiological adaptation features: Based on the time-series curve of physiological data, physiological load adaptation (match between heart rate, muscle strength, and dance difficulty); and output the core evaluation feature vector.

[0070] S23, Feature Weight Allocation: Based on the core evaluation feature vector output in S22, and combined with the evaluation focus of different dance styles (such as classical dance emphasizing movement standardization and modern dance emphasizing emotional expression), the Analytic Hierarchy Process (AHP) is used to allocate the weights of each feature: movement feature weight 40%, rhythm feature weight 25%, emotional feature weight 25%, and physiological adaptation feature weight 10%. At the same time, the weights are dynamically fine-tuned by combining basic information such as the student's learning time and age (such as appropriately reducing the weight of emotional features and increasing the weight of movement features for children), and the weighted core evaluation feature set is output.

[0071] S3, Layered Quantitative Evaluation: Based on the weighted core evaluation feature set output by S2, conduct single-dimensional and multi-dimensional comprehensive scoring, generate personalized evaluation labels, benchmark comparisons and progress tracking reports, and complete the layered quantitative evaluation;

[0072] The specific steps of S3 are as follows:

[0073] S31, Single-dimensional quantitative scoring: Based on the weighted core evaluation feature set output by S2, a single-dimensional quantitative scoring is performed on each type of core evaluation feature: The Dynamic Time Warping (DTW) algorithm is used to compare the student's movement features with the standard movement features of the corresponding dance style, and calculate the scores (0-100 points) of indicators such as movement standard and fluency; the rhythm feature score is calculated through rhythm synchronization rate, the emotional feature score is calculated through emotional matching degree, and the physiological adaptation feature score is calculated through physiological load adaptation degree, and a single-dimensional quantitative scoring table is output;

[0074] S32, Multi-dimensional Comprehensive Scoring: Based on the single-dimensional quantitative scoring table output in S31, a weighted summation formula is used to calculate the trainee's multi-dimensional comprehensive evaluation score: Comprehensive Score = Σ (Single-dimensional score × corresponding weight), with a score range of 0-100 points. Simultaneously, evaluation levels are assigned based on the comprehensive score (Excellent: 90-100 points, Good: 80-89 points, Pass: 60-79 points, Fail: 0-59 points). The deviation between each dimension's score and the standard score is calculated concurrently to identify the trainee's strengths and weaknesses, and to output a comprehensive evaluation score, evaluation level, and dimension deviation analysis report.

[0075] S33: Personalized assessment tag generation. Based on the comprehensive assessment score, assessment level, and dimensional deviation analysis report output in S32, and combined with the student's basic information, personalized assessment tags are generated for the student. These tags include strength tags (such as "outstanding movement standard" and "high rhythm fit"), weakness tags (such as "insufficient emotional transmission" and "incorrect force application"), and potential tags (such as "strong physiological load adaptability, which can improve high-difficulty movements"). At the same time, combined with the assessment data of similar students, student ranking and progress analysis are generated, and a personalized assessment tag set and ranking analysis report are output.

[0076] S34, Benchmarking and Progress Tracking: Based on the personalized evaluation tag set and ranking analysis report output in S33, the evaluation data of outstanding students in the same dance style and age group are introduced as industry benchmarks. The core characteristics of the students and the benchmarks are compared to identify the key gaps. At the same time, the historical evaluation data of the students (if available) is retrieved to track the changing trends of scores in each dimension, analyze the highlights of progress and the shortcomings that have not been improved, and output a benchmarking report and progress tracking curve.

[0077] S4, Evaluation Feedback and Dynamic Optimization: Based on the benchmark comparison report and progress tracking curve output by S34, generate personalized improvement plans and verify their feasibility, while providing teaching optimization suggestions, and then dynamically optimize the evaluation system parameters and feature library.

[0078] The specific steps of S4 are as follows:

[0079] S41, Personalized Improvement Plan Generation: Based on the benchmark comparison report and progress tracking curve output in S34, a targeted personalized improvement plan is generated and output for each student's weakness dimension. For example, for the weakness of "insufficient movement standard", specific incorrect movements (such as knee bending angle deviation) are identified by combining human body key point data, and specific correction methods (such as daily targeted stretching, decomposed movement practice) and practice duration are given; for the weakness of "insufficient emotional transmission", suggestions for emotional expression training are given (such as understanding emotions in conjunction with the background of the music, and strengthening the transmission through facial expression practice).

[0080] S42, Feasibility Verification of the Improvement Plan: Based on the targeted and personalized improvement plan output in S41, verify the feasibility and adaptability of the plan: By simulating the effect of students implementing the improvement plan through AI, determine whether the practice intensity and frequency are suitable for the students' physiological load, and whether there are problems of excessive or insufficient difficulty; at the same time, combine the students' learning time and acceptance ability to fine-tune the practice content and cycle to ensure that the improvement plan is scientific, feasible and easy to implement, and output the optimized personalized improvement plan and feasibility verification report.

[0081] S43, Generating Teaching Optimization Suggestions: Based on the optimized personalized improvement plan and feasibility verification report output in S42, combined with the overall class evaluation data, generating teaching optimization suggestions for the instructor and outputting them: For example, if most students in the class have the problem of "insufficient rhythm matching", it is recommended that the instructor add rhythm training sessions; if some students have "incorrect force application points", it is recommended that the instructor conduct targeted explanations of force application techniques.

[0082] S44, Dynamic Optimization of the Evaluation System: Based on the teaching optimization suggestions output in S43, and combined with the optimized personalized improvement plan and feasibility verification report output in S42, the core parameters of the evaluation are dynamically optimized: the weight allocation of each evaluation feature is adjusted, the movement standard feature library is optimized, and the quantitative scoring standard is updated. At the same time, in combination with new dance styles and new teaching needs, the evaluation dimensions and feature types are expanded to ensure the adaptability and accuracy of the evaluation method, forming a dynamically optimized evaluation system, and outputting the optimized evaluation parameters and feature library update report.

[0083] A dance education assessment system, please refer to Figure 1 ,include:

[0084] The multimodal assessment data acquisition module collects students' basic dance videos, key points of the human body, and multidimensional data on physiology and emotion. It also records information related to students, songs, and scenes, and outputs a multimodal assessment dataset.

[0085] The data preprocessing and feature extraction module cleans, integrates, and standardizes the multimodal evaluation dataset output by the multimodal evaluation data acquisition module, extracts core features of action, rhythm, emotion, and physiological adaptation, assigns weights, and outputs a weighted core evaluation feature set.

[0086] The hierarchical quantitative evaluation module, based on the weighted core evaluation feature set output by the data preprocessing and feature extraction module, conducts single-dimensional and multi-dimensional comprehensive scoring, generates personalized evaluation labels, benchmark comparisons and progress tracking reports, and completes the hierarchical quantitative evaluation.

[0087] The evaluation feedback and dynamic optimization module, based on the benchmark comparison report and progress tracking curve output by the benchmark comparison and progress tracking unit, generates personalized improvement plans and verifies their feasibility, while providing teaching optimization suggestions, thereby dynamically optimizing the evaluation system parameters and feature library.

[0088] Please see Figure 2 The multimodal evaluation data acquisition module includes:

[0089] The basic movement video capture unit sets up a multi-camera capture scene, using three high-definition cameras (located at the front, side, and back respectively) to simultaneously capture video data of the student's complete dance segments. The capture parameters are set to: 1080P resolution and 30 frames per second. During the capture process, it is ensured that the student's entire body is in the shot, without obstruction or blur. At the same time, it records the capture scene information (such as venue size and light intensity), the student's basic information (such as age, dance style, and learning time), and the dance music information (such as music name and tempo), and outputs the student's basic dance video data.

[0090] The human body key point data acquisition unit, based on the basic video data output by the basic motion video acquisition unit, uses the Movenet human pose estimation algorithm to extract multiple human body key node data (including head, neck, shoulder, elbow, wrist, hip, knee, ankle, etc.) in real time during the student's dance process. At the same time, it calculates the spatial coordinates, motion trajectory and angle change data of each key node, removes abnormal data corresponding to blurred frames and occluded frames, performs noise reduction processing on the effective data, and outputs a standardized human body key point motion dataset.

[0091] The physiological and emotional data acquisition unit, based on the standardized human key point motion dataset output by the human key point data acquisition unit and combined with the basic video data output by the basic movement video acquisition unit, simultaneously collects the student's physiological and emotional data: physiological data is collected through wearable devices, including heart rate and muscle strength (focusing on arm, waist, and leg muscles), and outputs a physiological data time-series curve; emotional data is obtained through facial expression recognition in video frames, extracting the student's facial feature points, and combining the amplitude and rhythmic fit of dance movements to determine the student's emotional expression state (such as joy, sadness, excitement), outputting emotional state labels and corresponding confidence levels, and outputting a multimodal evaluation dataset.

[0092] Please see Figure 3 The data preprocessing and feature extraction module includes:

[0093] The multimodal data cleaning and integration unit, based on the multimodal assessment dataset output by the multimodal assessment data acquisition module, first cleans the video data, human key point data, physiological data, and emotional data. Specifically, the video data undergoes frame filtering, retaining valid frames with clear movements and no abnormalities, while deleting invalid frames that are blurry, occluded, or have interrupted movements. Outlier removal is performed on the human key point data, and interpolation is used to supplement missing data to ensure data continuity. The physiological and emotional data undergo standardization processing, converting data from different dimensions into standardized scores of 0-100. Subsequently, the video data, human key point data, physiological data, and emotional data are time-axis aligned and integrated into a unified assessment data matrix, outputting a standardized integrated dataset.

[0094] The core evaluation feature extraction unit, based on the standardized integrated dataset output by the multimodal data cleaning and integration unit, extracts four categories of core evaluation features: action features, rhythm features, emotional features, and physiological adaptation features. These correspond to the key dimensions of dance evaluation: ① Action features: Based on human keypoint data, extracting action standardization (deviation from standard action), action fluency (continuity of movement trajectory at key nodes), and force application rationality (muscle strength and action matching degree); ② Rhythm features: Based on the comparison between video data and music beats, extracting rhythm fit (synchronization rate between action and beat) and rhythm change adaptability (ability to adjust actions to changes in music rhythm); ③ Emotional features: Based on emotional state labels and facial expression data, extracting emotional transmission (matching degree between emotional labels and dance piece emotion) and emotional expressiveness (coordination between facial expression and movement amplitude); ④ Physiological adaptation features: Based on physiological data time-series curves, extracting physiological load adaptation (matching degree between heart rate, muscle strength, and dance difficulty); and outputting the core evaluation feature vector.

[0095] The feature weight allocation unit, based on the core evaluation feature vector output by the core evaluation feature extraction unit, and combined with the evaluation focus of different dance styles (such as classical dance emphasizing movement standardization and modern dance emphasizing emotional expression), uses the analytic hierarchy process (AHP) to allocate the weights of each feature: movement feature weight 40%, rhythm feature weight 25%, emotional feature weight 25%, and physiological adaptation feature weight 10%. At the same time, it dynamically fine-tunes the weights by combining basic information such as the student's learning time and age (such as appropriately reducing the weight of emotional features and increasing the weight of movement features for children), and outputs a weighted core evaluation feature set.

[0096] Please see Figure 4 The hierarchical quantitative evaluation module includes:

[0097] The single-dimensional quantitative scoring unit, based on the weighted core evaluation feature set output by the data preprocessing and feature extraction module, performs single-dimensional quantitative scoring on each type of core evaluation feature: using the Dynamic Time Warping (DTW) algorithm, it compares the student's movement features with the standard movement features of the corresponding dance style, and calculates scores (0-100 points) for indicators such as movement standardization and fluency; it calculates rhythm feature scores through rhythm synchronization rate, emotion feature scores through emotion matching degree, and physiological adaptation feature scores through physiological load adaptation degree, and outputs a single-dimensional quantitative scoring table;

[0098] The multi-dimensional comprehensive scoring unit, based on the single-dimensional quantitative scoring table output by the single-dimensional quantitative scoring unit, uses a weighted summation formula to calculate the trainee's multi-dimensional comprehensive evaluation score: Comprehensive Score = Σ (Single-dimensional score × corresponding weight), with a score range of 0-100 points. Simultaneously, the evaluation level is divided according to the comprehensive score (Excellent: 90-100 points, Good: 80-89 points, Pass: 60-79 points, Fail: 0-59 points). It also simultaneously calculates the deviation between each dimension's score and the standard score, identifies the trainee's strengths and weaknesses, and outputs a comprehensive evaluation score, evaluation level, and dimension deviation analysis report.

[0099] The personalized assessment tag generation unit, based on the comprehensive assessment score, assessment level, and dimensional deviation analysis report output by the multi-dimensional comprehensive scoring unit, and combined with the student's basic information, generates personalized assessment tags for the student. These tags include strength tags (such as "outstanding movement standard" and "high rhythm fit"), weakness tags (such as "insufficient emotional transmission" and "incorrect force application"), and potential tags (such as "strong physiological load adaptability, capable of improving high-difficulty movements"). Simultaneously, it combines assessment data from similar students to generate student rankings and progress analysis, outputting a personalized assessment tag set and ranking analysis report.

[0100] The benchmarking and progress tracking unit, based on the personalized evaluation tag set and ranking analysis report output by the personalized evaluation tag generation unit, introduces evaluation data of outstanding students in the same dance style and age group as industry benchmarks, compares the core characteristic differences between students and benchmarks, and clarifies the key gaps; at the same time, it retrieves students' historical evaluation data (if available), tracks the changing trends of scores in each dimension, analyzes the highlights of progress and the shortcomings that have not been improved, and outputs a benchmarking comparison report and progress tracking curve.

[0101] Please see Figure 5 The evaluation feedback and dynamic optimization module includes:

[0102] The personalized improvement plan generation unit, based on the benchmark comparison report and progress tracking curve output by the benchmark comparison and progress tracking unit, generates and outputs targeted personalized improvement plans for the trainee's weaknesses. For example, for the weakness of "insufficient movement standard," it identifies specific incorrect movements (such as knee bending angle deviation) by combining human key point data, and provides specific correction methods (such as daily targeted stretching and decomposed movement practice) and practice duration; for the weakness of "insufficient emotional transmission," it provides suggestions for emotional expression training (such as understanding emotions in conjunction with the background of the music and strengthening transmission through facial expression practice).

[0103] The improvement scheme feasibility verification unit verifies the feasibility and adaptability of the personalized improvement scheme output by the personalized improvement scheme generation unit: by simulating the effect of students implementing the improvement scheme through AI, it judges whether the practice intensity and frequency are suitable for the students' physiological load and whether there are problems of excessive or insufficient difficulty; at the same time, it fine-tunes the practice content and cycle based on the students' learning time and acceptance ability to ensure that the improvement scheme is scientific, feasible and easy to implement, and outputs the optimized personalized improvement scheme and feasibility verification report.

[0104] The teaching optimization suggestion generation unit, based on the optimized personalized improvement plan and feasibility verification report output by the improvement plan feasibility verification unit, and combined with the overall class evaluation data, generates and outputs teaching optimization suggestions for the instructors. For example, if most students in the class have the problem of "insufficient rhythm matching", it is recommended that the instructor add rhythm training sessions; if some students have "incorrect force application", it is recommended that the instructor conduct targeted explanations of force application techniques.

[0105] The dynamic optimization unit of the evaluation system generates teaching optimization suggestions based on the teaching optimization suggestions, and combines them with the optimized personalized improvement schemes and feasibility verification reports output by the improvement scheme feasibility verification unit. This dynamically optimizes the core parameters of the evaluation: adjusting the weight allocation of each evaluation feature, optimizing the movement standard feature library, and updating the quantitative scoring standards. At the same time, it expands the evaluation dimensions and feature types to meet new dance styles and new teaching needs, ensuring the adaptability and accuracy of the evaluation method. This forms a dynamically optimized evaluation system, and outputs optimized evaluation parameters and a feature library update report.

[0106] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for assessing dance education, characterized in that, The specific steps are as follows: S1, Multimodal assessment data collection: Collect students' basic dance videos, human body key points, physiological and emotional multidimensional data, and simultaneously record information related to students, songs and scenes, and output multimodal assessment dataset; S2, Data Preprocessing and Feature Extraction: The multimodal evaluation dataset output from S1 is cleaned, integrated and standardized, and core features of action, rhythm, emotion and physiological adaptation are extracted and weighted to output a weighted core evaluation feature set. S3, Stratified Quantitative Assessment: S31, Single-dimensional quantitative scoring: Based on the weighted core evaluation feature set output by S2, perform single-dimensional quantitative scoring on each type of core evaluation feature and output a single-dimensional quantitative scoring table. S32, Multi-dimensional Comprehensive Scoring: Based on the single-dimensional quantitative scoring table output by S31, the multi-dimensional comprehensive evaluation score of the trainee is calculated using a weighted summation formula. At the same time, the evaluation level is divided according to the comprehensive score, and the deviation of each dimension score from the standard score is calculated simultaneously. The comprehensive evaluation score, evaluation level and dimension deviation analysis report are output. S33: Personalized assessment tag generation. Based on the comprehensive assessment score, assessment level, and dimensional deviation analysis report output in S32, and combined with the student's basic information, personalized assessment tags are generated for the student. At the same time, combined with the assessment data of similar students, student ranking and progress analysis are generated, and a set of personalized assessment tags and a ranking analysis report are output. S34, Benchmarking and Progress Tracking: Based on the personalized evaluation tag set and ranking analysis report output in S33, the evaluation data of outstanding students in the same dance style and age group are introduced as industry benchmarks to compare the core characteristic differences between students and benchmarks; at the same time, the historical evaluation data of students are retrieved to track the changing trends of scores in each dimension, and a benchmarking report and progress tracking curve are output. S4, Evaluation Feedback and Dynamic Optimization: Based on the benchmark comparison report and progress tracking curve output by S34, generate personalized improvement plans and verify their feasibility, while providing teaching optimization suggestions, and then dynamically optimize the evaluation system parameters and feature library.

2. The dance education assessment method according to claim 1, characterized in that, The specific steps of S1 are as follows: S11, Basic Movement Video Acquisition: Set up a multi-camera acquisition scene to simultaneously acquire video data of complete dance segments of students; at the same time, record acquisition scene information, student basic information and dance music information, and output student basic dance video data. S12, Human Key Point Data Acquisition: Based on the basic video data output in S11, the Movenet human pose estimation algorithm is used to extract multiple human key node data in the student's dance process in real time. At the same time, the spatial coordinates, motion trajectory and angle change data of each key node are calculated, and abnormal data corresponding to blurred frames and occluded frames are removed. The effective data is denoised and a standardized human key point motion dataset is output. S13, Physiological and Emotional Data Acquisition: Based on the standardized human key point motion dataset output in S12, combined with the basic video data output in S11, the physiological and emotional data of the trainees are collected simultaneously, and a multimodal assessment dataset is output.

3. The dance education assessment method according to claim 1, characterized in that, The specific steps of S2 are as follows: S21, Multimodal data cleaning and integration: Based on the multimodal evaluation dataset output by S1, the video data, human key point data, physiological data, and emotional data are first cleaned, and then the video data, human key point data, physiological data, and emotional data are aligned on the time axis and integrated into a unified evaluation data matrix, outputting a standardized integrated dataset. S22, Core Evaluation Feature Extraction: Based on the standardized integrated dataset output in S21, extract four categories of core evaluation features: action features, rhythm features, emotion features, and physiological adaptation features, which correspond to the key dimensions of dance evaluation; and output the core evaluation feature vector. S23, Feature weight allocation: Based on the core evaluation feature vector output in S22, and combined with the evaluation focus of different dance styles, the weight of each feature is allocated using the analytic hierarchy process; at the same time, the weights are dynamically fine-tuned by combining the student's basic information, and a weighted core evaluation feature set is output.

4. The dance education assessment method according to claim 1, characterized in that, The specific steps of S4 are as follows: S41, Personalized Improvement Plan Generation: Based on the benchmark comparison report and progress tracking curve output in S34, a targeted personalized improvement plan is generated and output; S42, Feasibility verification of the improvement plan: Based on the targeted and personalized improvement plan output in S41, verify the feasibility and adaptability of the plan; at the same time, combine the students' learning time and acceptance ability, fine-tune the practice content and cycle, and output the optimized personalized improvement plan and feasibility verification report. S43, Generating Teaching Optimization Suggestions: Based on the optimized personalized improvement plan and feasibility verification report output in S42, and combined with the overall class evaluation data, generating teaching optimization suggestions for the instructors and outputting them; S44, Dynamic Optimization of the Evaluation System: Based on the teaching optimization suggestions output in S43, combined with the optimized personalized improvement plan and feasibility verification report output in S42, the core parameters of the evaluation are dynamically optimized. At the same time, in combination with new dance styles and new teaching needs, the evaluation dimensions and feature types are expanded, and the optimized evaluation parameters and feature library update report are output.

5. A dance education assessment system, characterized in that, include: The multimodal assessment data acquisition module collects students' basic dance videos, key points of the human body, and multidimensional data on physiology and emotion. It also records information related to students, songs, and scenes, and outputs a multimodal assessment dataset. The data preprocessing and feature extraction module cleans, integrates, and standardizes the multimodal evaluation dataset output by the multimodal evaluation data acquisition module, extracts core features of action, rhythm, emotion, and physiological adaptation, assigns weights, and outputs a weighted core evaluation feature set. The tiered quantitative assessment module includes a single-dimensional quantitative scoring unit, a multi-dimensional comprehensive scoring unit, a personalized assessment label generation unit, and a benchmarking and progress tracking unit. The single-dimensional quantitative scoring unit performs single-dimensional quantitative scoring on each type of core evaluation feature based on the weighted core evaluation feature set output by the data preprocessing and feature extraction module, and outputs a single-dimensional quantitative scoring table. The multi-dimensional comprehensive scoring unit, based on the single-dimensional quantitative scoring table output by the single-dimensional quantitative scoring unit, uses a weighted summation formula to calculate the trainee's multi-dimensional comprehensive evaluation score, and simultaneously classifies the evaluation level according to the comprehensive score; and simultaneously calculates the deviation between the score of each dimension and the standard score, and outputs a comprehensive evaluation score, evaluation level and dimension deviation analysis report. The personalized assessment tag generation unit generates personalized assessment tags for students based on the comprehensive assessment score, assessment level, and dimensional deviation analysis report output by the multi-dimensional comprehensive scoring unit, combined with the students' basic information; at the same time, it generates student ranking and progress analysis by combining assessment data of similar students, and outputs a set of personalized assessment tags and a ranking analysis report. The benchmarking and progress tracking unit, based on the personalized evaluation tag set and ranking analysis report output by the personalized evaluation tag generation unit, introduces the evaluation data of outstanding students in the same dance style and age group as industry benchmarks, and compares the core characteristic differences between students and benchmarks; at the same time, it retrieves the student's historical evaluation data, tracks the changing trends of scores in each dimension, and outputs a benchmarking report and progress tracking curve. The evaluation feedback and dynamic optimization module, based on the benchmark comparison report and progress tracking curve output by the benchmark comparison and progress tracking unit, generates personalized improvement plans and verifies their feasibility, while providing teaching optimization suggestions, thereby dynamically optimizing the evaluation system parameters and feature library.

6. A dance education assessment system according to claim 5, characterized in that, The multimodal evaluation data acquisition module includes: The basic movement video capture unit sets up a multi-camera capture scene to simultaneously capture video data of complete dance segments of students; at the same time, it records the capture scene information, student basic information and dance song information, and outputs the student's basic dance video data. The human body key point data acquisition unit, based on the basic video data output by the basic motion video acquisition unit, uses the Movenet human pose estimation algorithm to extract multiple human body key node data in real time during the student's dance process. At the same time, it calculates the spatial coordinates, motion trajectory and angle change data of each key node, removes abnormal data corresponding to blurred frames and occluded frames, performs noise reduction processing on the effective data, and outputs a standardized human body key point motion dataset. The physiological and emotional data acquisition unit, based on the standardized human key point motion dataset output by the human key point data acquisition unit and combined with the basic video data output by the basic motion video acquisition unit, simultaneously acquires the physiological and emotional data of the trainees and outputs a multimodal assessment dataset.

7. A dance education assessment system according to claim 5, characterized in that, The data preprocessing and feature extraction module includes: The multimodal data cleaning and integration unit, based on the multimodal evaluation dataset output by the multimodal evaluation data acquisition module, first cleans the video data, human key point data, physiological data, and emotional data, then aligns the video data, human key point data, physiological data, and emotional data along the time axis, integrates them into a unified evaluation data matrix, and outputs a standardized integrated dataset. The core evaluation feature extraction unit extracts four categories of core evaluation features—action features, rhythm features, emotion features, and physiological adaptation features—based on the standardized integrated dataset output by the multimodal data cleaning and integration unit. These features correspond to the key dimensions of dance evaluation, and the core evaluation feature vector is output. The feature weight allocation unit, based on the core evaluation feature vector output by the core evaluation feature extraction unit, and in combination with the evaluation focus of different dance styles, uses the analytic hierarchy process to allocate the weight of each feature; at the same time, it dynamically fine-tunes the weights by combining the student's basic information, and outputs a weighted core evaluation feature set.

8. A dance education assessment system according to claim 5, characterized in that, The evaluation feedback and dynamic optimization module includes: The personalized improvement plan generation unit generates and outputs targeted personalized improvement plans based on the benchmark comparison report and progress tracking curve output by the benchmark comparison and progress tracking unit. The improvement scheme feasibility verification unit verifies the feasibility and adaptability of the personalized improvement scheme output by the personalized improvement scheme generation unit; at the same time, it fine-tunes the practice content and cycle based on the student's learning time and acceptance ability, and outputs the optimized personalized improvement scheme and feasibility verification report. The teaching optimization suggestion generation unit generates and outputs teaching optimization suggestions for teachers based on the optimized personalized improvement plan and feasibility verification report output by the improvement plan feasibility verification unit, combined with the overall class evaluation data. The dynamic optimization unit of the evaluation system generates teaching optimization suggestions based on the teaching optimization suggestions, combines the optimized personalized improvement scheme and feasibility verification report output by the improvement scheme feasibility verification unit, dynamically optimizes the core parameters of the evaluation, and expands the evaluation dimensions and feature types in combination with new dance styles and new teaching needs, outputting optimized evaluation parameters and feature library update reports.

Images