A multi-angle dance video analysis system

Abstract

The present invention relates to the field of motion analysis technology, specifically to a multi-angle dance video analysis system, the system includes a motion capture feature extraction module, a motion data multi-scale analysis module, a motion analysis and mapping module, and a style classification and prediction module. In the present invention, through multi-angle video recording and Vietoris-Rips complex method, the spatial connection pattern of the dance is effectively identified and extracted, the geometric structure of the action is captured, and three-dimensional motion data analysis is provided. Multi-scale persistent homology analysis enhances the analysis of the time series and spatial layout of the action, accurately distinguishes dance styles and motion patterns, brings a high sensitivity to subtle differences in style, and helps to accurately predict the evolution of dance styles. The motion data is mapped to a low-dimensional space through multi-dimensional scaling analysis, which significantly improves data visualization and analysis capabilities, shows the similarity and difference of the action, and helps dance education, choreography and therapy, realizes the full process optimization of data analysis, and provides a comprehensive and in-depth dance motion analysis tool.

Description

Technical Field

[0001] The present invention relates to the technical field of motion analysis, and in particular to a multi-angle dance video analysis system. Background Art

[0002] The field of motion analysis technology involves the use of sensors, cameras and image capture devices to record and analyze the movements of the human body or objects. In this field, specialized software and hardware systems are designed to parse motion data to provide detailed insights into physical movements, including motion capture, posture assessment, motion reproduction and motion efficiency analysis. Motion analysis technology is widely used in sports science, physical therapy, film production and video game development to assist professionals in optimizing performance, diagnosing problems and improving user experience. This technology can not only analyze single-angle recordings, but also process multi-angle and multi-dimensional data to obtain more comprehensive motion information.

[0003] Among them, the multi-angle dance video analysis system is a high-tech system specially designed to capture dance movements from multiple angles and conduct detailed analysis. The system uses multiple cameras to synchronously record the dancers' movements, and uses advanced image processing and machine learning algorithms to analyze the video to identify the characteristics and patterns of dance movements. The main uses of this system include providing feedback to dance teachers and students to improve performance skills, assisting dance choreography, and providing experimental data for dance research. It can also be applied to dance therapy to assist therapists in evaluating the effectiveness of treatment by analyzing the fluency, strength and rhythm of movements.

[0004] Existing technologies provide basic tools for dance motion research, but they rely on sensors and single or limited camera angles, resulting in limited dimensions and angles of motion data, making it difficult to fully capture the three-dimensional characteristics of motion. Traditional methods are insufficient in dealing with motion continuity and style diversity, and are unable to accurately distinguish complex motion patterns or capture subtle style differences. This technical limitation restricts its application in dance teaching and choreography, and cannot provide sufficient data support for high-precision technical analysis and style prediction. In sports science, there is a lack of technical support for in-depth analysis of motion continuity. Traditional motion analysis technology is insufficient in multi-dimensional analysis and visualization of data, which limits the comprehensive analysis and application of complex motion data. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a multi-angle dance video analysis system.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: A multi-angle dance video analysis system comprises:

[0007] The motion capture feature extraction module is based on a multi-angle dance video dataset. It uses the Vietoris-Rips complex method to connect data points, identify spatial connection patterns in dance, extract spatial relationship data, and obtain topological network features.

[0008] The motion data multi-scale analysis module applies multi-scale persistent homology analysis based on the topological network features to determine the spatial and temporal continuity of the dance movements and obtain spatial and temporal continuity results. Through the spatial and temporal continuity results, the motion patterns appearing in the differentiated dance styles are identified to obtain a basic motion pattern set.

[0009] The action analysis and mapping module performs multi-dimensional scaling analysis on the actions in the multi-angle dance video based on the basic action pattern set, maps the action data to a low-dimensional space, and obtains an action mapping diagram. The action mapping diagram is used to display the similarities and differences between the dance actions, and obtain a dance action relationship diagram;

[0010] The style classification and prediction module analyzes the changes and trends of dance styles based on the dance movement relationship diagram, performs state analysis on the dance movement sequence, generates a dance state transition diagram, predicts dance style changes through the dance state transition diagram, and obtains dance video style prediction analysis results.

[0011] As a further solution of the present invention, the step of acquiring the topological network characteristics is specifically as follows:

[0012] Based on the multi-angle dance video data set, each dance movement is analyzed and key coordinate points are extracted, and the changes of key coordinate points in time series are recorded to generate dance movement time series data;

[0013] Applying the Vietoris-Rips complex method to the dance movement time series data, connecting the data points, and establishing a spatial connection graph according to a distance threshold;

[0014] According to the spatial connection graph, the key connection patterns are extracted and the connectivity of multiple connection patterns is calculated using the formula:

[0015] ;

[0016] Get topological network features, where: represents a spatial connection graph, is the weight adjusted according to the geometric characteristics and dynamic changes of the connection points. Represents the distance between connection points. represents the attenuation coefficient, Indicators of key connectivity patterns in the spatial connectivity graph representing dance movements, Indicates the total number of connection points.

[0017] As a further solution of the present invention, the steps of obtaining the spatial-temporal continuity result are specifically as follows:

[0018] Based on the characteristics of the topological network, the multi-scale persistent homology analysis method is used to evaluate the topological stability of dance movements. By calculating the Betti numbers at multiple scales, the continuity of the spatial structure is identified and preliminary topological analysis results are generated.

[0019] Based on the preliminary topological analysis results, the time series data of dance movements at multiple scales are analyzed, and the temporal continuity of the movements is evaluated using temporal coherence technology to generate the results of the temporal stability analysis of the movements;

[0020] Combining the preliminary topological analysis results and the motion time stability analysis results, the scores of spatial and temporal data are calculated using the formula:

[0021] ;

[0022] Generates a spatial-temporal continuity result where Represents the score after spatial and temporal data analysis, Represents a topological complex of time and space, represents the spatial continuity score of the kth time window, represents the corresponding time continuity score, Indicates the total number of time windows.

[0023] As a further solution of the present invention, the steps of acquiring the basic action mode set are specifically as follows:

[0024] Analyzing dance data using the spatial-temporal continuity results, identifying action patterns of differentiated styles, classifying the actions through cluster analysis, and generating action pattern classification results;

[0025] Analyzing the action pattern classification results, using frequent item set mining technology to capture basic action combinations appearing in the style, and generating a frequent action pattern set;

[0026] Combined with the frequent action pattern set, the representativeness and difference of each pattern are calculated using the formula:

[0027] ;

[0028] Generate a set of basic action patterns, where Represents a set of basic action patterns, represents the frequency of occurrence of pattern z, represents the difference weight corresponding to mode z, Indicates the total number of modes.

[0029] As a further solution of the present invention, the step of obtaining the action map is specifically as follows:

[0030] Analyze the multi-angle dance videos recorded in the basic movement pattern set, identify key movement features in multiple videos, mark the multi-dimensional attributes of each movement through statistical methods, and generate a movement feature label list;

[0031] Based on the action feature label list, a multidimensional scaling analysis technique is applied to convert high-dimensional data into low-dimensional data representation to generate a low-dimensional feature mapping matrix;

[0032] Based on the low-dimensional feature mapping matrix, the vector distance calculation formula is used:

[0033] ;

[0034] Generate an action map, where represents the spatial geometric clarity of the mapping, represents the vector position of the i-th action feature in the low-dimensional space, Represents the total action feature quantity.

[0035] As a further solution of the present invention, the steps of obtaining the dance action relationship diagram are specifically as follows:

[0036] Based on the action map, by analyzing the distance and configuration between multiple points in the map, a multidimensional scaling analysis method is applied to determine the similarities and differences between the actions, and preliminary action relationship data is obtained;

[0037] The preliminary action relationship data is converted into visual coordinates through matrix transformation method, and the Euclidean distance between actions is calculated using the formula:

[0038] ;

[0039] Generate a distance matrix where represents the Euclidean distance between two action points in the dance action mapping graph, , represents the x and y coordinates of the action in low-dimensional space, , Represents the coordinates of another action;

[0040] According to the distance matrix, a graphical algorithm is used to convert the action relationship into a visual chart, identify key and secondary action nodes, connect key action nodes to highlight the main line of the dance process, connect auxiliary secondary action nodes to refine the details of the dance action, distinguish the criticality of differentiated action nodes by color and line thickness, adjust the layout of the visual chart and optimize the display clarity of information, and generate a dance action relationship diagram.

[0041] As a further solution of the present invention, the steps of obtaining the dance state transition diagram are specifically as follows:

[0042] Extracting each action node and association from the dance action relationship graph, calculating the in-degree and out-degree of each node, each node representing an independent dance action, identifying potential connections between connectivity data actions, and generating a node association matrix;

[0043] The node association matrix is ​​used to determine the transition probabilities between dance moves, and the transition probability calculation formula is applied:

[0044] ;

[0045] Generate an action transition probability matrix, where represents the node degree of the starting action, represents the node degree of the end action, k is the sensitivity adjustment coefficient, is the state transition probability, e represents the base of the natural logarithm;

[0046] According to the action transition probability matrix, a state transition diagram is drawn using a graph theory algorithm, and the transition paths and probability sizes between differentiated dance actions are displayed to form a dance state transition diagram.

[0047] As a further solution of the present invention, the steps for obtaining the dance video style prediction analysis results are specifically as follows:

[0048] Based on the dance state transition graph, data on key dance movements and transition frequencies are extracted, and graph network analysis technology is used to identify key style influencing factors and obtain movement relationship feature data;

[0049] Based on the action relationship feature data, the sequence analysis technology is applied to predict the transition path of dance style, using the formula:

[0050] ;

[0051] Get dance style change prediction data, where: represents the predicted value of the dance state transition diagram prediction, and Indicates the state of continuous dance movements. , , are model parameters;

[0052] Integrate the dance style change prediction data, combine multiple style indicators through data fusion technology, capture the movement data of multiple styles in the dance video, classify and organize the data, use time series analysis to identify the trend of dance style changes, correspond the trend data with style indicators, and present the analysis results as charts through visualization tools to generate dance video style prediction analysis results.

[0053] Compared with the prior art, the advantages and positive effects of the present invention are:

[0054] In the present invention, by using multi-angle video data and Vietoris-Rips complex method, the spatial connection pattern in the dance is effectively identified and extracted, which not only captures the geometric structure of the action, but also reveals the complex spatial relationship between the actions, thereby providing a richer three-dimensional action data analysis than the traditional single perspective or two-dimensional analysis. The analysis of the time series and spatial layout of the action is further enhanced by multi-scale persistent homology analysis, allowing the system to accurately distinguish and classify differentiated dance styles and repeated action patterns. This in-depth analysis brings a high sensitivity to the subtle differences in action styles, which helps to accurately predict and classify the evolution of dance styles. The action data is mapped to a low-dimensional space through multi-dimensional scaling analysis, which significantly improves the visualization and analysis capabilities of the action data, so that the similarities and differences between actions can be intuitively presented, which not only helps dance teachers and students to analyze and improve dance skills from multiple dimensions, but also provides scientific data support for dance choreography and therapy, realizes the optimization of the whole process from data collection to analysis, and provides a comprehensive and in-depth dance action analysis tool, thus surpassing most of the limitations of the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a system flow chart of the present invention;

[0056] Figure 2 A flow chart showing the topological network characteristics of the present invention;

[0057] Figure 3 It is a flow chart of the space-time continuity result in the present invention;

[0058] Figure 4 is a flow chart of a basic action mode set in the present invention;

[0059] Figure 5 is a flow chart of the action mapping diagram in the present invention;

[0060] Figure 6 A flow chart of a dance action relationship diagram in the present invention;

[0061] Figure 7 is a flow chart of a dance state transition diagram in the present invention;

[0062] Figure 8 Flow chart of the dance video style prediction analysis results in the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to analyze the present invention and are not used to limit the present invention.

[0064] In the description of the present invention, it is necessary to interpret that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be interpreted as a limitation on the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0065] See also Figure 1 , a multi-angle dance video analysis system comprises:

[0066] The motion capture feature extraction module is based on a multi-angle dance video dataset. It uses the Vietoris-Rips complex method to connect data points, identify spatial connection patterns in dance, extract spatial relationship data, and obtain topological network features.

[0067] The multi-scale analysis module of motion data uses multi-scale persistent homology analysis based on topological network characteristics to determine the spatial and temporal continuity of dance movements and obtain spatial and temporal continuity results. Through the spatial and temporal continuity results, the motion patterns that appear in differentiated dance styles are identified and the basic motion pattern set is obtained.

[0068] The action analysis and mapping module performs multi-dimensional scaling analysis on the actions in multi-angle dance videos based on the basic action pattern set, maps the action data into a low-dimensional space, and obtains an action mapping diagram. Through the action mapping diagram, the similarities and differences between dance actions are displayed, and a dance action relationship diagram is obtained;

[0069] The style classification and prediction module analyzes the changes and trends of dance styles based on the dance movement relationship diagram, performs state analysis on the dance movement sequence, generates a dance state transition diagram, predicts dance style changes through the dance state transition diagram, and obtains dance video style prediction analysis results.

[0070] Topological network features include node type, edge connectivity, and network topology. Spatial-temporal continuity results include time series, spatial configuration, and continuity indicators. The basic action pattern set includes action elements, frequency characteristics, and style conformity. The action mapping diagram includes mapping dimensions, spatial vectors, and mapping accuracy. The dance action relationship diagram includes action connection points, similarity classification, and action network. The dance state transition diagram includes state classification, transition frequency, and state path. The dance video style prediction analysis results include prediction model, style distribution, and recognition accuracy.

[0071] See also Figure 2 , the specific steps for obtaining topological network features are:

[0072] Based on the multi-angle dance video data set, each dance movement is analyzed and key coordinate points are extracted, and the changes of key coordinate points in time series are recorded to generate dance movement time series data;

[0073] The analysis based on the multi-angle dance video dataset aims to extract key coordinate points. The process first involves parsing the dance movements in each frame of the video and identifying the positions of key joints and body parts in the movements. The coordinate data of each key coordinate point changes over time. The data is recorded by dedicated software and encoded into a time series format. The time series data includes the specific time point when each movement is performed and its precise position in space. The continuous data stream provides the basis for subsequent analysis, and then the time series data is processed by algorithms to identify specific patterns in the dance. The patterns reflect the coherence and rhythm of the dance movements, and finally generate dance movement time series data.

[0074] The Vietoris-Rips complex method is applied to the dance movement time series data to connect the data points and establish a spatial connection graph based on the distance threshold.

[0075] The Vietoris-Rips complex method is applied to dance movement time series data. A multidimensional spatial network is constructed by calculating the distance between data points, in which each data point represents a key coordinate point of a dance movement. The connection distance threshold is determined by statistical analysis of the distance between all movement data points. The key to selecting this threshold is to determine which points will be regarded as adjacent and thus connected. The construction of the spatial connection diagram refers to the relative positions between joints and body parts in dance movements. By comparing the repetition of similar positions in differentiated dance clips, it is determined which movements are part of the repeated or characteristic movements in the dance, and a spatial connection diagram representing the dance structure is generated.

[0076] According to the spatial connection graph, the key connection patterns are extracted and the connectivity of multiple connection patterns is calculated using the formula:

[0077] ;

[0078] Get topological network features, where: Represents a spatial connection graph, which is a weight adjusted according to the geometric characteristics and dynamic changes of the connection points. Represents the distance between connection points. represents the attenuation coefficient, Indicators of key connectivity patterns in the spatial connectivity graph representing dance movements, Indicates the total number of connection points;

[0079] The benefit of the formula is that it can refer to the combined influence of the distance and connection strength of key connection points in the dance. Through this weighted average method, it can more accurately reflect the importance of the spatial connection pattern of dance movements;

[0080] set up A spatial connection diagram for dance, is the weight of the ith connection, which is adjusted based on the geometric characteristics and dynamic changes of the connection. is the distance between the i-th connection points, calculated using the Euclidean distance. is the attenuation coefficient, which is adjusted according to the speed and rhythm of the dance movements in order to better control the impact of distance factors on connection strength.

[0081] Taking the specific values ​​as an example, there are three connections with weights of 0.5, 1, and 0.5, respectively, and distances of 1m, 2m, and 1.5m, respectively. The attenuation coefficient Set to 1, the calculation process is as follows:

[0082] ;

[0083] The results show that after referring to the combined effects of connection strength and distance, the tightness of the spatial connection pattern of dance movements is approximately 0.223, and the value reflects the overall structure and tightness of the connection between key movements in dance.

[0084] See also Figure 3 , the steps to obtain the spatial and temporal continuity results are as follows:

[0085] Based on the topological network characteristics, the multi-scale persistent homology analysis method is used to evaluate the topological stability of dance movements. By calculating the Betti numbers at multiple scales, the continuity of the spatial structure is identified and preliminary topological analysis results are generated.

[0086] In the preliminary implementation of multi-scale persistent homology analysis based on topological network characteristics, dance movement data were collected and converted into a format that can be used for analysis. By calculating the Betti numbers at differentiated scales, the analysis assisted in identifying and quantifying the continuity of the spatial structure of dance movements. This method can identify spatial patterns that are stable under multi-scale monitoring by analyzing the topological structure at the differentiated level. This process involves not only data collection and preprocessing, but also complex mathematical modeling to ensure the accuracy and reliability of the results obtained, generating preliminary topological analysis results, which provide basic data and reference standards for further spatial and temporal continuity analysis.

[0087] Based on the preliminary topological analysis results, the time series data of dance movements at multiple scales are analyzed, and the temporal coherence technology is used to evaluate the temporal continuity of the movements and generate the results of the temporal stability analysis of the movements.

[0088] After obtaining the preliminary topological analysis results, the data is further combined to analyze the time series of dance movements. This stage mainly involves the application of time synchronization technology, which allows the evaluation of the continuity and change patterns of movements in time. This technology identifies the topological features that change over time through further analysis of the time series data. The process requires not only a high degree of data synchronization technology, but also precise processing of time data to ensure that the analysis results of time continuity accurately reflect the true characteristics of the movement. Through complex analysis, the time continuity analysis results are generated, which further enriches the analysis and interpretation of the continuity of dance movements.

[0089] Combined with the preliminary topological analysis results and the motion time stability analysis results, the scores of spatial and temporal data are calculated using the formula:

[0090] ;

[0091] Generates a spatial-temporal continuity result where Represents the score after spatial and temporal data analysis, Represents a topological complex of time and space, represents the spatial continuity score of the kth time window, represents the corresponding time continuity score, Indicates the total number of time windows;

[0092] The usefulness of the formula lies in that it integrates the results of the analysis of spatial and temporal dimensions, providing a new perspective to analyze the continuity and fluidity of dance movements;

[0093] set up is a topological complex of time and space, and Represent the spatial and temporal continuity scores of each time window k, respectively. For example, if there are three time windows, the spatial scores of each window are 0.8, 0.7, and 0.9, and the temporal scores are 0.5, 0.6, and 0.7, respectively. The calculation is as follows:

[0094] ;

[0095] The results show that, combining spatial and temporal analysis of multiple time windows, the overall spatial-temporal continuity score of the dance movements is 1.45. The score reflects that the dance movements maintain a high degree of continuity and fluidity within a given time window. This numerical result further confirms the effectiveness of the analysis method and the continuity characteristics of the dance movements.

[0096] See also Figure 4 , the specific steps for obtaining the basic action mode set are:

[0097] Using the spatial-temporal continuity results, we analyze the dance data, identify the movement patterns of different styles, classify the movements through cluster analysis, and generate movement pattern classification results;

[0098] By utilizing the spatial-temporal continuity results, the process begins with a detailed analysis of the spatial and temporal attributes of the dance styles, focusing primarily on identifying unique movement patterns between styles. The execution process involves not only in-depth mathematical and statistical analysis of the existing classified movement data, but also the use of machine learning techniques such as cluster analysis to distinguish and identify the movement features of differentiated styles, ensuring that the movement patterns of each dance style are effectively identified and classified. The classification results provide a scientific basis for the next step of pattern set extraction, and ensure that the information extracted from the movement data is maximized, laying a solid foundation for creating a more accurate movement pattern set.

[0099] Analyze the action pattern classification results, use frequent item set mining technology to capture the basic action combinations that appear in the style, and generate frequent action pattern sets;

[0100] Continuing from the preliminary action pattern classification results, we further apply frequent item set mining technology, which is a technology used to identify patterns and trends in large data sets. By counting the frequency of multiple action patterns in the database, we determine which action patterns are common. The analysis process not only focuses on the frequency of action, but also evaluates the distribution of frequently occurring action patterns in differentiated dance styles, thereby obtaining action combinations with high support and confidence. The combination reflects the most basic and common action characteristics of the dance style, providing key data support for the formation of the final basic action pattern set.

[0101] Combined with the set of frequent action patterns, the representativeness and difference of each pattern are calculated using the formula:

[0102] ;

[0103] Generate a set of basic action patterns, where Represents a set of basic action patterns, represents the frequency of occurrence of pattern z, represents the difference weight corresponding to mode z, Indicates the total number of modes;

[0104] The formula is beneficial in that it provides a comprehensive method of determining the basic movement pattern set with reference to frequency and importance by quantifying the frequency of occurrence of each movement pattern and its uniqueness in the dance;

[0105] set up Represents a set of basic action patterns, is the total number of reference action patterns, and They represent the occurrence frequency and difference weight of the z-th action mode respectively. For example, if there are three action modes with frequencies of 0.2, 0.3, and 0.5 and difference weights of 2, 5, and 3 respectively, the calculation process is as follows:

[0106] ;

[0107] The results show that the total score of the basic movement pattern set is 3.4, which reflects the combined influence of the importance and uniqueness of the basic patterns in the overall dance style, and further guides the focus of dance training and performance.

[0108] See also Figure 5 , the steps for obtaining the action map are as follows:

[0109] Analyze multi-angle dance videos that record basic movement patterns, identify key movement features in multiple videos, and use statistical methods to mark the multi-dimensional attributes of each movement to generate a list of movement feature labels;

[0110] We analyzed multi-angle dance videos that recorded basic movement patterns, identified key movement features in each video, and used statistical methods to label the multidimensional attributes of each movement in detail, including its speed, rhythm, spatial occupancy, and frequency of occurrence in differentiated dance styles. This process used advanced data classification and pattern recognition techniques to ensure the accuracy and comprehensiveness of the labeling, and ultimately generated a list of labels containing a detailed description of each movement feature. The labels will be used for further multidimensional analysis.

[0111] Based on the action feature label list, multidimensional scaling analysis technology is applied to transform high-dimensional data into low-dimensional data representation and generate a low-dimensional feature mapping matrix;

[0112] The list of action feature labels extracted from multi-angle dance videos is processed, and multidimensional scaling analysis technology is used to convert high-dimensional data into a low-dimensional space representation to ensure that the overall structure and correlation of the data are preserved. This conversion process relies on complex mathematical models and algorithms, including the construction of distance matrices and dimensionality compression, which provides a scientific basis for generating accurate low-dimensional feature mapping matrices. The matrix will directly affect the quality and interpretability of the action mapping diagram.

[0113] Based on the low-dimensional feature mapping matrix, the vector distance calculation formula is used:

[0114] ;

[0115] Generate an action map, where represents the spatial geometric clarity of the mapping, represents the vector position of the i-th action feature in the low-dimensional space, Represents the total action feature quantity;

[0116] The benefit of the formula is that by calculating the Euclidean distance of all action vectors in low-dimensional space, it can effectively measure and display the similarities and differences between dance movements, enhance the resolution and visual effect of the action mapping diagram, and enable researchers to more clearly identify and distinguish the basic action patterns in differentiated dance styles;

[0117] Low-dimensional feature data obtained from multidimensional scaling analysis , calculate the Euclidean distance of each action vector, and based on this, further calculate the spatial distance between all actions to provide a numerical basis for constructing the action mapping diagram. In the example, set , ,but The calculation is: ;

[0118] The results show that the motion maps calculated by this method can effectively express the key characteristics of the original high-dimensional motion data in two-dimensional or three-dimensional space, making the comparison and analysis of motions more intuitive and accurate.

[0119] See also Figure 6 , the specific steps for obtaining the dance action relationship diagram are:

[0120] Based on the action map, by analyzing the distance and configuration between multiple points in the map, a multidimensional scaling analysis method is applied to determine the similarities and differences between actions, and obtain preliminary action relationship data;

[0121] By analyzing the distance and configuration between each point in the mapping diagram, and using the known coordinate data in the action mapping diagram, multidimensional scaling analysis is performed. This analysis quantifies the Euclidean distance between each action point and reveals the similarity between actions, thereby preliminarily drawing the relative position relationship between dance actions. The data will serve as the basis for creating a relationship diagram in the subsequent steps, thereby providing the necessary input parameters for the next step of processing.

[0122] The preliminary action relationship data is converted into visual coordinates through matrix transformation method, and the Euclidean distance between actions is calculated using the formula:

[0123] ;

[0124] Generate a distance matrix where represents the Euclidean distance between two action points in the dance action mapping graph, , represents the x and y coordinates of the action in low-dimensional space, , Represents the coordinates of another action;

[0125] The benefit of the formula is that by calculating the Euclidean distance between each pair of movements, the similarities and differences between movements can be quantified, which helps to accurately represent the positional relationship of each dance movement in the movement relationship graph;

[0126] Each point in the dance map has coordinates and , the formula for calculating the distance between two points is The distance of each pair of actions can be obtained. For example, for the specific points (3, 4) and (0, 0), the distance is calculated as ;

[0127] This result reveals the relative spatial positions of the two dance movements in the mapping diagram, which demonstrates the accuracy of the action mapping diagram and the clear expression of the relationship between movements, further verifying the effectiveness of using this formula.

[0128] According to the distance matrix, the action relationship is converted into a visual diagram using a graphical algorithm, key and secondary action nodes are identified, key action nodes are connected to highlight the main line of the dance process, auxiliary secondary action nodes are connected to refine the details of the dance action, the key of the action nodes is distinguished by color and line thickness, the layout of the visual diagram is adjusted and the clarity of information display is optimized to generate a dance action relationship diagram;

[0129] The distance data between all movements are converted into a visual relationship diagram through a graphical algorithm. The implementation process of the algorithm is explained in detail, including the sorting of distance data, cluster analysis and the rendering method of the final graph. The process not only tests data processing skills, but also demonstrates how to convert theoretical data into graphics that users can directly view. The results show the connections and differences between dance movements, providing new perspectives and tools for dance research.

[0130] See also Figure 7 , the specific steps for obtaining the dance state transition diagram are:

[0131] Extract each action node and correlation from the dance action relationship graph, calculate the in-degree and out-degree of each node, each node represents an independent dance action, identify the potential connection between the connectivity data actions, and generate a node correlation matrix;

[0132] Node data is extracted from the dance movement relationship graph. The nodes represent differentiated dance movements. The connections between the movements are analyzed. By calculating the connection density and direction of each node, the similarities and differences between differentiated dance styles can be revealed. Based on the data, the popular trends and style evolution of dance movements are analyzed, and finally a dance style prediction model is generated. The model can be used to guide dance choreography and teaching.

[0133] Using the node association matrix, determine the transition probability between dance moves and apply the transition probability calculation formula:

[0134] ;

[0135] Generate an action transition probability matrix, where represents the node degree of the starting action, represents the node degree of the end action, k is the sensitivity adjustment coefficient, is the state transition probability, e represents the base of the natural logarithm;

[0136] The benefit of the formula is that it can be based on the difference in connection density of action nodes and To adjust the transition probability, k is used as the adjustment coefficient, which can be adjusted according to the specific needs of the dance style. This allows the model to not only reflect the probabilistic relationship between dance movements, but also adapt to the changes between differentiated dance styles.

[0137] First determine the connection density of each dance action node and , get the value through actual data measurement, and then substitute the value into the formula to calculate the transition probability between each pair of dance moves , for example, if , , and set , then calculate ;

[0138] This result shows that based on the given parameters and model settings, the transition probability from action x to action y is 62.2%, which reflects the possibility of transitioning from one specific dance move to another.

[0139] According to the action transition probability matrix, the graph theory algorithm is used to draw the state transition diagram, and the transition paths and probability sizes between differentiated dance actions are displayed to form a dance state transition diagram;

[0140] Based on the dance state transition diagram, state analysis technology is further used to identify and predict state changes in dance sequences. By collecting actual performance data of various dance movements, analyzing the patterns and frequencies in the data, it is determined which dance movements appear together and which are less common. This information helps dance teachers and choreographers better analyze the evolution trends of dance styles and provide data support for creation. Ultimately, a visual chart that can dynamically display the evolution of dance styles is obtained.

[0141] See also Figure 8 ,The specific steps for obtaining the dance video style prediction analysis results are:

[0142] Based on the dance state transition graph, data on key dance movements and transition frequencies are extracted, and graph network analysis technology is used to identify key style influencing factors and obtain movement relationship feature data;

[0143] Based on the dance state transition graph, data extraction is performed on key dance movements and transition frequencies. First, the timestamps and types of all dance movements are extracted from the video footage. The continuous movements are divided into discrete events using video analysis technology, and each movement is marked and classified. For the transition frequency of movements, the time interval between each pair of continuous movements is analyzed, and the number of occurrences and periodic changes of each type of transition are counted. By constructing an action transition graph, the nodes represent dance movements, the edges represent the transition relationship between movements, and the weight of the edges corresponds to the transition frequency. The correlation and sequence dependency between dance movements are further analyzed through the connection strength and pattern of the nodes. The data is used as the feature input of the network. Using graph network analysis technology, key style influencing factors are identified and action relationship feature data is obtained.

[0144] Based on the action relationship feature data, the sequence analysis technology is applied to predict the transition path of dance style, using the formula:

[0145] ;

[0146] Get dance style change prediction data, where: represents the predicted value predicted by the dance state transition graph, and Indicates the state of continuous dance movements. , , are model parameters;

[0147] The benefit of the formula is that it allows the trend of dance style changes to be predicted from the action relationship feature data, by using the weight parameter , , Adjust the impact of differentiated action states on style prediction to provide more flexible style transfer analysis;

[0148] Setting parameters , , to represent the weight of the differentiated dance action state, for example, if and Represent the characteristic values ​​of two main dance movement states, such as modern dance and classical dance. Through this formula, the current style prediction value can be calculated. For example, and , then the predicted value ;

[0149] This indicates that the current dance style prediction tends to be modern dance style. This result shows that by adjusting , , The value of can effectively control the contribution of differentiated dance movements to style prediction, thereby optimizing the accuracy and adaptability of the style prediction model.

[0150] Integrate dance style change prediction data, combine multiple style indicators through data fusion technology, capture multiple styles of movement data in dance videos, classify and organize the data, use time series analysis to identify the trend of dance style changes, correspond trend data with style indicators, and use visualization tools to present the analysis results as charts to generate dance video style prediction analysis results;

[0151] Integrate the dance style change prediction data, combine multiple style indicators through data fusion technology, first match the dance movement data with the historical database of dance styles, and identify the style labels corresponding to each dance movement. By clustering the dance movements in different time periods, identify the main and secondary style features, analyze the frequency of occurrence and change patterns of the features in the dance sequence, integrate the style data of a single dance movement through advanced data fusion methods, and form a continuous prediction of the style change of the entire dance sequence. Use time series analysis tools to perform trend analysis on the integrated style data, determine the evolution trend of the dance style, and correspond the trend data to the style indicators. Use visualization tools such as advanced charts and timeline views to display the analysis results of style changes in the form of charts, and generate dance video style prediction analysis results, which not only assists dance choreographers in analyzing style evolution, but also provides a scientific basis for dance teaching.

[0152] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A multi-angle dance video analysis system, characterized in that: The system comprises: The motion capture feature extraction module is based on a multi-angle dance video dataset. It uses the Vietoris-Rips complex method to connect data points, identify spatial connection patterns in dance, extract spatial relationship data, and obtain topological network features. The steps for obtaining the topological network features are specifically as follows: Based on the multi-angle dance video data set, each dance movement is analyzed and key coordinate points are extracted, and the changes of key coordinate points in time series are recorded to generate dance movement time series data; Applying the Vietoris-Rips complex method to the dance movement time series data, connecting the data points, and establishing a spatial connection graph according to a distance threshold; According to the spatial connection graph, the key connection patterns are extracted and the connectivity of multiple connection patterns is calculated using the formula: ； Get topological network features, where: represents a spatial connection graph, is the weight adjusted according to the geometric characteristics and dynamic changes of the connection points. Represents the distance between connection points. represents the attenuation coefficient, Indicators of key connectivity patterns in the spatial connectivity graph representing dance movements, Indicates the total number of connection points; The motion data multi-scale analysis module applies multi-scale persistent homology analysis based on the topological network features to determine the spatial and temporal continuity of the dance motions and obtain spatial and temporal continuity results. Through the spatial and temporal continuity results, the motion patterns appearing in the differentiated dance styles are identified to obtain a basic motion pattern set. The steps for obtaining the spatial-temporal continuity result are specifically as follows: Based on the topological network characteristics, the multi-scale persistent homology analysis method is used to evaluate the topological stability of dance movements. By calculating the Betti numbers at multiple scales, the continuity of the spatial structure is identified and preliminary topological analysis results are generated. Based on the preliminary topological analysis results, the time series data of dance movements at multiple scales are analyzed, and the temporal continuity of the movements is evaluated using temporal coherence technology to generate the results of the temporal stability analysis of the movements; Combining the preliminary topological analysis results and the motion time stability analysis results, the scores of spatial and temporal data are calculated using the formula: ； Generates a spatial-temporal continuity result where Represents the score after spatial and temporal data analysis, Represents a topological complex of time and space, Representative The spatial continuity score of the time window, represents the corresponding time continuity score, Indicates the total number of time windows; The action analysis and mapping module performs multi-dimensional scaling analysis on the actions in the multi-angle dance video based on the basic action pattern set, maps the action data to a low-dimensional space, and obtains an action mapping diagram. The action mapping diagram is used to display the similarities and differences between the dance actions, and obtain a dance action relationship diagram; The steps for obtaining the action map are specifically as follows: Analyze the multi-angle dance videos recorded in the basic movement pattern set, identify key movement features in multiple videos, mark the multi-dimensional attributes of each movement through statistical methods, and generate a movement feature label list; Based on the action feature label list, a multidimensional scaling analysis technique is applied to convert high-dimensional data into low-dimensional data representation to generate a low-dimensional feature mapping matrix; Based on the low-dimensional feature mapping matrix, the vector distance calculation formula is used: ； Generate an action map, where represents the spatial geometric clarity of the mapping, Indicates The vector position of the action feature in the low-dimensional space, Represents the total action feature quantity; The steps for obtaining the dance action relationship diagram are specifically as follows: Based on the action map, by analyzing the distance and configuration between multiple points in the map, a multidimensional scaling analysis method is applied to determine the similarities and differences between the actions, and preliminary action relationship data is obtained; The preliminary action relationship data is converted into visual coordinates through matrix transformation method, and the Euclidean distance between actions is calculated using the formula: ； Generate a distance matrix where represents the Euclidean distance between two action points in the dance action mapping graph, , represents the x and y coordinates of the action in low-dimensional space, , Represents the coordinates of another action; Using a graphical algorithm to convert the action relationship into a visual chart according to the distance matrix, identifying key and secondary action nodes, connecting key action nodes to highlight the main line of the dance process, connecting auxiliary secondary action nodes to refine the details of the dance action, distinguishing the criticality of differentiated action nodes by color and line thickness, adjusting the layout of the visual chart and optimizing the display clarity of information, and generating a dance action relationship chart; The style classification and prediction module analyzes the changes and trends of dance styles based on the dance action relationship diagram, performs state analysis on the dance action sequence, generates a dance state transition diagram, predicts dance style changes through the dance state transition diagram, and obtains dance video style prediction analysis results; The steps for obtaining the dance state transition diagram are specifically as follows: Extracting each action node and association from the dance action relationship graph, calculating the in-degree and out-degree of each node, each node representing an independent dance action, identifying potential connections between connectivity data actions, and generating a node association matrix; The node association matrix is ​​used to determine the transition probabilities between dance moves, and the transition probability calculation formula is applied: ； Generate an action transition probability matrix, where represents the node degree of the starting action, represents the node degree of the end action, is the sensitivity adjustment coefficient, is the state transition probability, represents the base of natural logarithms; According to the action transition probability matrix, a state transition diagram is drawn using a graph theory algorithm, and the transition paths and probability sizes between differentiated dance actions are displayed to form a dance state transition diagram; The steps for obtaining the dance video style prediction analysis results are specifically as follows: Based on the dance state transition graph, data on key dance movements and transition frequencies are extracted, and graph network analysis technology is used to identify key style influencing factors and obtain movement relationship feature data; Based on the action relationship feature data, the sequence analysis technology is applied to predict the transition path of dance style, using the formula: ； Get dance style change prediction data, where: represents the predicted value predicted by the dance state transition graph, and Indicates the state of continuous dance movements. , , are model parameters; Integrate the dance style change prediction data, combine multiple style indicators through data fusion technology, capture the movement data of multiple styles in the dance video, classify and organize the data, use time series analysis to identify the trend of dance style changes, correspond the trend data with style indicators, and present the analysis results as charts through visualization tools to generate dance video style prediction analysis results.

2. The multi-angle dance video analysis system according to claim 1, characterized in that: The steps for obtaining the basic action mode set are specifically as follows: Analyzing dance data using the spatial-temporal continuity results, identifying action patterns of differentiated styles, classifying the actions through cluster analysis, and generating action pattern classification results; Analyzing the action pattern classification results, using frequent item set mining technology to capture basic action combinations appearing in the style, and generating a frequent action pattern set; Combined with the frequent action pattern set, the representativeness and difference of each pattern are calculated using the formula: ； Generate a set of basic action patterns, where Represents a set of basic action patterns, Delegate Mode The frequency of occurrence, Representation Mode The corresponding difference weights are, Indicates the total number of modes.

Images