A method and system for automatically classifying music works

Abstract

The present invention relates to the field of music information retrieval technology, specifically to a method and system for automatically classifying music works, comprising the following steps: based on the classification standard of music genres, analyzing the hierarchical node relationship of music genres, extracting the corresponding hierarchical characteristics of genre major categories and minor categories, identifying the correlation ratio between each level, analyzing the hierarchical connection strength, and generating a music genre hierarchical weight table. In the present invention, through the in-depth analysis of the hierarchical structure of music genres, a more refined and dynamic classification of music works is achieved, and the accuracy and flexibility of classification are significantly improved. By extracting hierarchical features and revealing the association between levels, music classification not only breaks through the limitations of simple labeling, but also reflects the hierarchical relationship between genres, extracts the law and amplitude of chord changes, enhances the dynamic capture of music features, matches the association between chord changes and genre hierarchical features, dynamically adjusts the classification weight, accurately matches genre characteristics, and better adapts to the diversity and complexity of music creation.

Description

Technical Field

[0001] The present invention relates to the technical field of music information retrieval, and in particular to a method and system for automatically classifying musical works. Background Art

[0002] The field of music information retrieval technology aims to extract and retrieve information related to user needs from a large amount of music data. Its main goal is to enable people to effectively access and analyze music data through computer algorithms and models. The field includes multiple research directions, such as audio feature extraction, music signal processing, pattern recognition, machine learning and data mining. Audio analysis technology is used to identify rhythm, melody, harmony and other features in audio content, and then support music recommendation, similar song search, automatic tagging, sentiment analysis and other applications. Music information retrieval technology is also widely used in music copyright management, music creation assistance, music education and other fields.

[0003] Among them, the automatic classification method of music works is an important task in music information retrieval. Its purpose is to automatically divide music works into different categories or labels, such as music genres, emotions, playing styles, etc. through algorithms. The classification method relies on features extracted from audio signals, such as pitch, rhythm, spectral features, etc., combined with machine learning or deep learning models for training and prediction. Automatic classification can not only provide personalized recommendations for users, but also play a role in music management and big data analysis, helping music platforms and content providers to quickly process and manage massive music data.

[0004] When performing music work classification, existing technologies mainly rely on static feature extraction and a single classification rule, which is difficult to cope with complex music changes. Existing methods mostly rely on quantitative features of audio signals, such as rhythm, pitch and other simple statistical features for classification. This method ignores the dynamic changes and hierarchical relationships in music works. Traditional classification methods fail to fully consider the hierarchical relationships between different genres, and fail to perform in-depth analysis based on more detailed music elements such as chord changes. This static and single-dimensional analysis method has obvious limitations on the accuracy and scalability of classification in complex and diverse music data. Static analysis based on rhythm and pitch cannot accurately capture the complex chord changes and genre characteristics in different music styles, resulting in some music works with cross-genre characteristics being misclassified, thereby affecting the effectiveness of applications such as recommendation systems and copyright management. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method and system for automatically classifying musical works.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: a method for automatically classifying musical works, comprising the following steps:

[0007] S1: Based on the classification criteria of music genres, analyze the hierarchical node relationships of music genres, extract the corresponding hierarchical characteristics of genre categories and subcategories, identify the correlation ratio between each level, analyze the hierarchical connection strength, and generate a music genre hierarchical weight table;

[0008] S2: Based on the music genre level weight table, extract the chord structure information of the music work, divide the chord data segments according to the time series, compare the chord frequency and duration of the segments, extract the key characteristics of the chord changes, generate a music chord change amplitude table, match the frequency change and position relationship of the sequence segments, extract the chord characteristic dynamic change index, and generate a chord characteristic change sequence table;

[0009] S3: Based on the chord characteristic change sequence table, analyzing the correlation value between the chord change index and the genre level weight, extracting the distribution relationship between the chord change rule and the level node, generating a genre chord matching rule table, extracting and analyzing the dynamic weight change of the genre classification, and generating a music work classification path rule set;

[0010] S4: Based on the music work classification path rule set, dynamically match the music genre path value and adjust the weight change, compare and analyze the genre path value with the dynamic change data of the work chord, extract the path nodes matching the genre rules, combine the hierarchical order and node value, and generate a music work classification hierarchical structure table.

[0011] As a further solution of the present invention, the step of obtaining the music genre level weight table is specifically as follows:

[0012] S111: Based on the classification standard of music genres, extract the basic classification characteristics of the genres from the hierarchical node relationships of the music genres, analyze the subordinate major and minor categories of each genre node, and generate an initial set of node association ratios by analyzing the association ratio between each genre node and the upper level node;

[0013] S112: calling the initial set of node association ratios, analyzing the connection characteristics between differentiated levels, identifying the differences in the level association ratios of genre categories and subcategories, and setting a preliminary indicator of the level connection strength, using the formula:

[0014] ;

[0015] Calculate the hierarchical connection strength values ​​from the genre category to the subcategory and generate a hierarchical connection strength set;

[0016] in, Represents the hierarchical connection strength value from the main category to the subcategory of the genre, Represents the weight parameter between major categories and minor categories, Represents the correlation ratio between major categories and minor categories, is the total number of nodes;

[0017] S113: Using the hierarchical connection strength set and combining the hierarchical node quantity weights of differentiated genres, perform a hierarchical weight analysis between major and minor categories of each genre, optimize the hierarchical weight values ​​through recursive iteration, and generate a music genre hierarchical weight table.

[0018] As a further solution of the present invention, the step of obtaining the music chord variation amplitude table is specifically as follows:

[0019] S211: extracting chord information from a musical work based on the music genre level weight table, and dividing it into multiple chord segments according to time series, analyzing the chord type in each segment, calculating the occurrence frequency and duration of the chord, and generating a chord frequency and duration distribution table through basic statistics of the chord data in the segment;

[0020] S212: According to the chord frequency and duration distribution table, the chord frequency and duration of each segment are compared, and the variation range of the chords in the segment is analyzed, and the formula is used:

[0021] ;

[0022] Get the chord change amplitude data of the fragment;

[0023] in, Indicates the amplitude of the chord change in the fragment. and Indicates the chord frequency of the current segment and the previous segment, and Indicates the duration of the chords in the current and previous segments. , , are the adjustment coefficients for chord frequency and duration, is the additional weight coefficient for chord changes;

[0024] S213: calling the chord change amplitude data of the segment, analyzing the change trend of the chord change amplitude in the differentiated time periods, classifying the chord type and dynamic characteristics in the segment through the change amplitude data, adjusting the amplitude weight in combination with the change trend, and generating a music chord change amplitude table.

[0025] As a further solution of the present invention, the step of obtaining the chord characteristic change sequence table is specifically as follows:

[0026] S221: Based on the music chord variation amplitude table, extract the starting frequency and ending frequency of each chord, sort the frequency range of each chord in chronological order, compare the rising and falling trends of the chord frequencies according to the time series, record the frequency fluctuation amplitude and direction, and generate a frequency variation and position relationship table;

[0027] S222: Based on the frequency change and position relationship table, the time series is segmented according to the chord mark, the frequency change range in each segment is counted segment by segment, the frequency fluctuation start point and end point are divided into intervals, and the chord position mark and frequency change trend are matched segment by segment, the segment duration and span are recorded, and a chord segment frequency change table is generated;

[0028] S223: Based on the chord segment frequency change table, the chord change index of each segment is analyzed, the frequency change amplitude and direction within the segment are extracted, the continuity and transition characteristics between segments are recorded, and the dynamic characteristics of the segments are integrated in sequence to construct a chord characteristic change sequence table.

[0029] As a further solution of the present invention, the step of obtaining the genre chord matching rule table is specifically as follows:

[0030] S311: Based on the chord characteristic change sequence table, by analyzing the chord change index, extracting the relationship node between each chord characteristic and the corresponding genre level, performing correlation operation on the frequency value of the chord change index and the genre level weight, and generating a preliminary correlation matrix between the chord change index and the genre level weight;

[0031] S312: calling the preliminary correlation matrix of the chord change index and the genre level weight, analyzing the distribution relationship between the chord change rule and the level node, and analyzing the weighted distribution of the level node corresponding to the chord characteristic change, using the formula:

[0032] ;

[0033] Establish a correlation table between chord characteristics and hierarchical node distribution;

[0034] in, Indicates the level of genre The chord change rule value, represents the genre level weight, Indicates the time proportion of chord changes. and Indicates the value of the adjacent chord characteristic change index, is the adjustment coefficient of time proportion, is the weight parameter of the chord characteristic change index;

[0035] S313: calling the chord characteristics and hierarchical node distribution association table, analyzing the matching degree of chord characteristics in the genre hierarchy, classifying the chord change rules by the association values ​​of the hierarchical nodes, and generating a genre chord matching rule table.

[0036] As a further solution of the present invention, the steps of obtaining the music work classification path rule set are specifically as follows:

[0037] S321: extracting structural parameters of chord sequences based on the genre chord matching rule table, performing frequency statistics, and classifying parameters according to genre matching conditions, summarizing weight values ​​corresponding to chord combinations, integrating weight data and chord combination information, and obtaining a chord matching genre classification weight table;

[0038] S322: using the chord matching genre classification weight table, extracting the time series interval corresponding to the weight value, performing segmented statistics on the weight change trend, extracting the continuously changing features and performing labeling integration, combining the genre matching classification information to perform trend quantification, and obtaining a genre dynamic weight change trend set;

[0039] S323: Using the genre dynamic weight change trend set, matching the weight change features and chords in the trend data, and matching the chord combinations in the genre classification weight table, extracting the classification path rules based on the corresponding relationship between the weight intervals and the combinations, integrating the path data, and generating a music work classification path rule set.

[0040] As a further solution of the present invention, the steps of obtaining the music work classification hierarchical structure table are specifically as follows:

[0041] S411: based on the music work classification path rule set, extracting the chord dynamic change data of the music work, analyzing the frequency, duration and rhythm characteristics of the chords, and performing dynamic matching, calculating the feature vector of each group of chords, and generating a music work chord dynamic change feature set;

[0042] S412: calling the music work chord dynamic change feature set, and comparing and matching it with the path value in the predefined music genre path rule set, and extracting the path nodes and hierarchical information that conform to the path rule according to the matching result, using the formula:

[0043] ;

[0044] Calculate the matching scores between the path nodes and the chord changes to obtain the node set matching the path rule;

[0045] in, represents the matching score between the path node and the chord change, Represents the dynamic characteristics of chords. represents the chord weight, Represents the node feature value in the genre path rule set, represents the node weight, Represents the path dynamic adjustment parameters;

[0046] S413: Based on the node set matched by the path rule, the node set is arranged in hierarchical order and matching score, and the correlation and hierarchical relationship between the hierarchical nodes are analyzed to generate a hierarchical structure table for the classification of musical works.

[0047] A music work automatic classification system, the music work automatic classification system is used to execute the above-mentioned music work automatic classification method, the system comprises:

[0048] The music genre hierarchical analysis module extracts the weight value and correlation relationship of each genre node based on the classification standard of music genre, calculates the upper and lower connection strength values ​​between nodes, compares the distribution of node weight values, identifies the distribution rules and correlation ratios between levels, and generates a music genre hierarchical weight table;

[0049] The chord characteristic extraction module extracts the music chord structure information based on the music genre level weight table, divides the chord data into time segments, compares the chord frequency and duration in each segment, arranges the chord change characteristic values ​​into time series, and generates a chord characteristic change sequence table;

[0050] The genre chord rule analysis module extracts the corresponding relationship between the chord change index and the genre level weight node value based on the chord characteristic change sequence table, analyzes the chord dynamic change and the level distribution matching value, identifies the dynamic change law and the level matching rule association value, and generates the genre chord matching rule table;

[0051] The dynamic classification path generation module analyzes the distribution of dynamic changes in music chord values ​​based on the genre chord matching rule table, extracts the matching node values ​​from the rule table, analyzes the node weight distribution, adjusts the hierarchical order and weight relationship, identifies the dynamic path node hierarchical connection rules, and generates a music work classification path rule set;

[0052] The classification hierarchical structure output module extracts the path rule hierarchical node distribution relationship based on the music work classification path rule set, analyzes the correspondence between the node distribution value and the classification hierarchical structure, organizes the music genre classification hierarchical information into hierarchical structure data, and generates a music work classification hierarchical structure table.

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

[0054] In the present invention, through in-depth analysis of the hierarchical structure of music genres, music works can be effectively classified in a more refined and dynamic manner, significantly improving the accuracy and flexibility of the prior art. By analyzing the hierarchical node relationships between different genres, specific features of each hierarchy can be extracted, revealing the strength association between hierarchies, so that music classification is not limited to simple labeling, but can also reflect the subtle hierarchical structure between music genres, providing a more dimensional reference for genre identification. Combined with chord structure analysis, a detailed comparison of chord frequency and duration is made based on time series data, and the chord change rules and amplitude are extracted, further enhancing the dynamic capture capability of music features. By matching the association between chord change data and genre hierarchical features, the music genre can be accurately matched while dynamically adjusting the classification weight, making the classification process of music works not only more flexible, but also able to adapt to the diversity and complexity of music creation, greatly improving the accuracy and adaptability of classification. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0056] Figure 2 A flowchart of obtaining a music genre level weight table in the present invention;

[0057] Figure 3 A flowchart of obtaining a music chord variation amplitude table in the present invention;

[0058] Figure 4 is a flow chart for obtaining a chord characteristic change sequence table in the present invention;

[0059] Figure 5 A flowchart for obtaining a genre chord matching rule table in the present invention;

[0060] Figure 6 A flowchart of obtaining a music work classification path rule set in the present invention;

[0061] Figure 7 The present invention is a flowchart for obtaining a hierarchical structure table of music work classification. DETAILED DESCRIPTION

[0062] 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 herein are only used to explain the present invention and are not intended to limit the present invention.

[0063] In the description of the present invention, it should be understood 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 drawings, and 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 understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0064] Example 1: Please refer to Figure 1 The present invention provides a technical solution: a method for automatically classifying musical works, comprising the following steps:

[0065] S1: Based on the classification criteria of music genres, analyze the hierarchical node relationships of music genres, extract the corresponding hierarchical characteristics of genre categories and subcategories, identify the correlation ratio between each level, analyze the hierarchical connection strength, and generate a music genre hierarchical weight table;

[0066] S2: Based on the music genre level weight table, extract the chord structure information of the music work, segment the chord data segments according to the time series, compare the chord frequency and duration of the segments, extract the key characteristics of the chord changes, generate the music chord change amplitude table, match the frequency change and position relationship of the sequence segments, extract the chord characteristic dynamic change index, and generate the chord characteristic change sequence table;

[0067] S3: Based on the chord characteristic change sequence table, analyze the correlation value between the chord change index and the genre level weight, extract the distribution relationship between the chord change rule and the level node, generate the genre chord matching rule table, extract and analyze the dynamic weight change of the genre classification, and generate the music work classification path rule set;

[0068] S4: Based on the music work classification path rule set, dynamically match the music genre path value and adjust the weight change, compare and analyze the genre path value with the dynamic change data of the work chord, extract the path nodes that match the genre rules, combine the hierarchical order and node value, and generate a music work classification hierarchical structure table.

[0069] The music genre hierarchy weight table includes genre major category weights, minor category weights, hierarchy association ratios, and hierarchy connection strengths; the music chord change amplitude table includes chord frequency distribution, chord duration, and chord change amplitude characteristics; the chord characteristic change sequence table includes chord frequency change indicators, time series distribution characteristics, and dynamic change position relationships; the genre chord matching rule table includes chord change regularity characteristics, genre hierarchy node distribution, and hierarchy matching rule values; the music work classification path rule set includes classification path node values, genre dynamic weight distribution, and genre path matching rules; the music work classification hierarchy structure table includes classification hierarchy nodes, path weight value distribution, and hierarchy sequence relationships.

[0070] See also Figure 2 , the specific steps for obtaining the music genre level weight table are:

[0071] S111: Based on the classification standard of music genres, extract the basic classification characteristics of the genres from the hierarchical node relationships of the music genres, analyze the subordinate major and minor categories of each genre node, and generate an initial set of node association ratios by analyzing the association ratio between each genre node and the upper level node;

[0072] According to the hierarchical node relationship of music genres, the basic classification characteristics of each genre are extracted. By analyzing the hierarchical relationship between the major and minor categories to which each node belongs, the node information of all genres is disassembled. The attributes of each node need to be retrieved from the genre characteristic database. The classification characteristics of each genre include the genre name, characteristic dimensions (such as rhythm, melody, timbre, etc.) and the level to which it belongs. The classification characteristic data needs to be extracted through feature analysis of music samples. The spectral characteristics of music samples (such as band energy and short-time Fourier transform results) are called, and the rhythm characteristics of the genres are analyzed one by one. The nodes are classified into a set of basic classification characteristics (such as beats per minute, note duration distribution) and melody characteristics (such as pitch sequence distribution, melody center of gravity) and are sorted into a set of basic classification characteristics of the genre. Combined with the relationship between the major and minor categories, the associations between the nodes are processed through hierarchical feature mapping to establish a preliminary association relationship between the nodes. The association relationship is obtained by calculating the classification characteristic similarity index (such as cosine similarity) between the upper and lower nodes. The similarity index is calculated by the ratio of the intersection of the feature distribution between the major category and the minor category to the overall feature distribution. Finally, the calculation of the node association ratio is completed to generate an initial set of node association ratios.

[0073] S112: Call the initial set of node association ratios, analyze the connection characteristics between differentiated levels, identify the differences in the level association ratios of genre categories and subcategories, and set a preliminary indicator of the level connection strength using the formula:

[0074] ;

[0075] Calculate the hierarchical connection strength values ​​from the genre category to the subcategory and generate a hierarchical connection strength set;

[0076] in, Represents the hierarchical connection strength value from the main category to the subcategory of the genre, Represents the weight parameter between major categories and minor categories, Represents the correlation ratio between major categories and minor categories, is the total number of nodes;

[0077] The formula is useful because it introduces the weight parameter and the associated ratio , comprehensively considers the structural association and influence strength between levels, making the calculation results of hierarchical connection strength more comprehensive and quantifiable;

[0078] Set the weight parameters by calling the initial set of node association ratios The influence strength between the major and minor categories of each genre is obtained through the evaluation data of the importance of the hierarchy in the music genre classification standard. For example, the weight parameter can be obtained by extracting the hierarchical scoring data of the music samples. The scoring is quantified (ranging from 0 to 1) by multiple music classification experts based on the influence criteria (such as the distribution ratio of the genre in the major or minor categories, and the number of classified samples);

[0079] Correlation ratio It is calculated through the hierarchical node association matrix, and its value represents the intersection ratio of the feature distribution between a node and the upper node. This value is calculated by the overlap ratio of the spectral features in the music sample.

[0080] Total number of nodes is the total number of nodes at the level, obtained directly from the number of nodes;

[0081] set up , ;

[0082] Substitute into the formula: ;

[0083] Calculate the numerator part: ;

[0084] Calculate the denominator: ;

[0085] The denominator results in: ;

[0086] Final calculation: ;

[0087] The result shows that the hierarchical connection strength value is 0.885, indicating that the hierarchical association between large categories and small categories has strong connectivity and serves as the result item of the hierarchical connection strength set.

[0088] S113: using the hierarchical connection strength set and combining the hierarchical node quantity weights of the differentiated genres, performing hierarchical weight analysis between the major and minor categories of each genre, optimizing the hierarchical weight values ​​through recursive iteration, and generating a music genre hierarchical weight table;

[0089] The hierarchical connection strength value and the node quantity weight are called, and the hierarchical weight between each major category and minor category is calculated by analyzing the hierarchical connection characteristics one by one. The weight distribution analysis is performed on the hierarchical characteristics of each node. The distribution analysis is calculated by counting the proportion of nodes at each level in the major or minor category to which they belong. The node proportion calculation is based on the distribution ratio in the hierarchical structure of music genres. This ratio comes from the distribution characteristics of genre samples in the music feature database. The weight proportion of each node is adjusted through recursive iteration, and the adjusted weight set is called to reassign and optimize the weight of each level, and finally a music genre hierarchical weight table is generated.

[0090] See also Figure 3 , the specific steps for obtaining the music chord change amplitude table are:

[0091] S211: extracting chord information from a musical work based on a music genre level weight table, and dividing it into multiple chord segments according to a time series, analyzing the chord type in each segment, calculating the occurrence frequency and duration of the chord, and generating a chord frequency and duration distribution table through basic statistics of the chord data in the segment;

[0092] First, the audio data of the music work is parsed, and the time series data of the chords is extracted through parsing. The chords are divided into multiple segments according to the time axis, and each segment corresponds to the chord combination characteristics within a certain time period. Then, the frequency of occurrence of the chords is statistically calculated. By traversing the time axis of the chord segments, the number of times each chord appears in each segment is calculated and normalized to obtain the chord frequency ratio information. The duration is directly accumulated and calculated through the time series of each segment, and the duration array is called to gradually analyze the distribution of the existence duration of each chord. After converting the frequency and duration into quantifiable data forms, a comparative distribution result of the chord frequency and duration is formed. This distribution result is obtained by calculation. The frequency is calculated by the ratio of the total number of occurrences of the chord type in the time series to the total length of the time series, and the duration is the direct accumulation of the time length in the segment, forming a chord frequency and duration distribution table for subsequent calculations.

[0093] S212: According to the chord frequency and duration distribution table, the chord frequency and duration of each segment are compared, and the variation range of the chords in the segment is analyzed, and the formula is used:

[0094] ;

[0095] Get the chord change amplitude data of the fragment;

[0096] in, Indicates the amplitude of the chord change in the fragment. and Indicates the chord frequency of the current segment and the previous segment, and Indicates the duration of the chords in the current and previous segments. , , are the adjustment coefficients for chord frequency and duration, is the additional weight coefficient for chord changes;

[0097] The benefit of the formula is that, through the analysis of the ratio of frequency to duration, and the comprehensive calculation of absolute difference and weight coefficient, the sensitivity to the amplitude of chord changes can be effectively improved, making the formula suitable for the analysis of the change characteristics of different music clips, especially in clips with drastic or slight chord changes, relatively reasonable amplitude values ​​can be obtained;

[0098] Chord Frequency The data is obtained by calculating the number of occurrences of the chord in the current segment and then normalizing it. This data comes from the proportion of the chord in the segment in the time series, the frequency of the previous segment Obtained statistically in the same way;

[0099] The duration of the current chord seconds. This data is obtained by directly accumulating the duration of the segment. The duration of the previous segment Seconds are also obtained by accumulating the time series;

[0100] Weight Parameters It is set according to the music genre hierarchy weight table. This value reflects the importance of the music genre and is obtained through expert scoring method and change sensitivity adjustment method.

[0101] Adjustment factor , , They are determined by sensitivity analysis of changes in music clips, and are specifically adjusted by the normalization range and variation range of the data;

[0102] Substituting the above values ​​into the formula: ;

[0103] Substituting the above results into: ;

[0104] The result shows that the chord change amplitude of the current segment is 6.42, which is a comprehensive reflection of the frequency change and the duration change. The quantitative description of the intensity of the chord change is further optimized through weight adjustment, which can be used to generate the change amplitude model later.

[0105] S213: calling the chord change amplitude data of the segment, analyzing the change trend of the chord change amplitude in the differentiated time period, classifying the chord type and dynamic characteristics in the segment according to the change amplitude data, adjusting the amplitude weight according to the change trend, and generating a music chord change amplitude table;

[0106] The fluctuation trend of the chord change amplitude in the time dimension is analyzed. By calling the chord change amplitude value, the highest and lowest amplitudes of the chord changes in the fragment are gradually counted, and the values ​​are divided into intervals for classifying the dynamic changes of chords. Combined with the changing trends of chord frequency and duration, the data is further organized into quantifiable change characteristics. Through dynamic weight adjustment of the changing trend, the weight is optimized and allocated according to the significance of chord changes in different time periods. Each group of chord data introduced in the weight allocation process is dynamically classified and trend adjusted, and finally a music chord change amplitude table is formed, which records the chord change characteristics and classification indicators in all fragments.

[0107] See also Figure 4 , the specific steps for obtaining the chord characteristic change sequence table are:

[0108] S221: Based on the music chord change amplitude table, extract the starting frequency and ending frequency of each chord, sort the frequency range of each chord in chronological order, and compare the rising and falling trends of the chord frequencies according to the time series, record the frequency fluctuation amplitude and direction, and generate a frequency change and position relationship table;

[0109] First, each chord fragment in the music file is parsed, and its frequency components are band-pass filtered to screen the dominant frequency components in a specific interval. The frequency values ​​of the start time and end time of the chord fragment are recorded respectively. During the parsing process, the frequency range of each chord is sorted in chronological order, and the main frequency components in the frequency range are arranged in order to ensure that the recorded frequency can accurately correspond to the actual time point. Based on the time series, the frequency change trend of each chord is analyzed. By comparing the change amplitude and direction of the main frequency values ​​at adjacent time points, the frequency rise and fall trend is marked. If there is a continuous increase or decrease in frequency, it is classified separately and the directional characteristics of the change are marked. During the recording process, the frequency fluctuation amplitude and direction of each chord are classified into a time series feature group to ensure that the connection between each feature group can present the continuity characteristics of the frequency change. Finally, a frequency change and position relationship table is generated. The start time, end time, main frequency range, frequency rise and fall trend and its corresponding time segment position of each chord are marked one by one, providing a basic reference for the subsequent automatic classification of music works.

[0110] S222: Based on the frequency change and position relationship table, the time series is segmented according to the chord mark, the frequency change range in each segment is counted segment by segment, the frequency fluctuation start point and end point are divided into intervals, and the chord position mark and frequency change trend are matched segment by segment, the segment duration and span are recorded, and the chord segment frequency change table is generated;

[0111] The time axis in the relationship table is divided into multiple segments, each segment corresponds to an independent chord change mark, and the main frequency change in the chord segment is recorded. After segmentation, the frequency change range of each segment is counted, and the fluctuation range of the frequency within the segment is analyzed in turn. By recording the lowest and highest frequency values ​​of the segment, the frequency fluctuation amplitude of the segment is calculated, and the time position corresponding to the starting and ending points of the frequency fluctuation is marked. On this basis, each chord segment is divided into intervals, and the starting and ending points of the frequency fluctuation are mapped to specific intervals on the time axis respectively, and the main frequency change characteristics in the interval are marked. In the process of segment-by-segment matching, the directional characteristics and dominant frequency change types of the frequency fluctuations of each segment are confirmed by comparing the chord position mark and the frequency change trend, and the duration and frequency span of the segment are recorded. Through multiple retrievals and verifications, a chord segment frequency change table is generated, which records the time length, frequency span, start and end positions, fluctuation amplitude and trend characteristics of each chord segment in detail, providing a segmented feature description basis for automatic classification.

[0112] S223: Based on the chord segment frequency change table, analyzing the chord change index of each segment, extracting the frequency change amplitude and direction within the segment, recording the continuity and transition characteristics between segments, and sequentially integrating the segment dynamic characteristics to construct a chord characteristic change sequence table;

[0113] The frequency change amplitude and direction in each segment are extracted from the table, and the change characteristics of the chord segment are described in detail. The frequency change characteristics are analyzed segment by segment, and the difference range between the highest and lowest frequency values ​​in the segment is recorded. Combined with the mark of the frequency change direction, the frequency fluctuation characteristics of the segment are comprehensively evaluated. According to the continuity and transition characteristics between the segments, the main frequency change trend between adjacent segments is retrieved, and the connection point between the two segments is analyzed whether there is a sudden or gradual change in frequency. The smoothness of the frequency connection during the change process is recorded. In the analysis process, the dynamic characteristics of all segments are sequentially integrated to construct a chord characteristic change sequence table. The sequence table includes indicators such as the frequency fluctuation amplitude, directionality, segment connection characteristics, and dynamic change rate of the chord, which provides a quantitative feature dimension basis for automatic classification. Through the coherence analysis of the overall characteristics, the internal change rules and dynamic characteristics of the music work are clarified, thereby supporting the implementation of the classification task.

[0114] See also Figure 5 , the specific steps for obtaining the genre chord matching rule table are:

[0115] S311: Based on the chord characteristic change sequence table, by analyzing the chord change index, extracting the relationship node between each chord characteristic and the corresponding genre level, performing correlation operation on the frequency value of the chord change index and the genre level weight, and generating a preliminary correlation matrix between the chord change index and the genre level weight;

[0116] First, the time series data of the chord change index is called, and the index value of each chord characteristic change is analyzed. The time proportion of each index is obtained by normalization. The time proportion is calculated by ratio operation of the duration of the chord change in the current time window and the total duration. Then, the calculated chord time proportion is associated with the corresponding weight value in the genre level weight table item by item. The association operation method is to multiply the chord change time proportion value and the level weight value according to the corresponding relationship, and then accumulate them one by one. Then, the correlation results are hierarchically counted, and the chord changes of different genre levels are preliminarily analyzed according to the distribution range of the chord characteristic changes. The hierarchical statistical values ​​are organized into a matrix form. Each unit in the matrix represents the relationship between the weight value of a genre level and a chord change index. Finally, the preliminary correlation matrix of the chord change index and the genre level weight is obtained by generating the matrix and normalizing the distribution data.

[0117] S312: calling the preliminary correlation matrix of chord change index and genre level weight, analyzing the distribution relationship between chord change rules and level nodes, and analyzing the weighted distribution of chord characteristic changes corresponding to level nodes, using the formula:

[0118] ;

[0119] Establish a correlation table between chord characteristics and hierarchical node distribution;

[0120] in, Indicates the level of genre The chord change rule value, represents the genre level weight, Indicates the time proportion of chord changes. and Indicates the value of the adjacent chord characteristic change index, is the adjustment coefficient of time proportion, is the weight parameter of the chord characteristic change index;

[0121] The benefit of the formula is that it combines the genre-level weight value with the time proportion of the chord characteristic change index and its change trend. By introducing the adjustment coefficient and weight parameter, it can more carefully reflect the correlation between the chord characteristic change and the genre-level distribution, and provide more accurate data support for the generation of the rule table.

[0122] parameter Indicates the weight value of a certain genre level, which is directly called by the genre level weight table. The table determines the specific weight by counting the influence of the genre on the music characteristics;

[0123] parameter Indicates the time proportion of the chord change feature in the time window, and the normalized result is obtained by dividing the chord change duration in the time series by the current total duration;

[0124] parameter , It represents the chord change index value in two adjacent time windows, which is obtained by counting the chord change characteristic data;

[0125] Adjustment factor , The influence of time proportion and variation index on the result is controlled respectively, and the value of adjustment coefficient is set according to the distribution range of chord data;

[0126] Substitution , , , Add the value to the formula and calculate : ;

[0127] Calculate the time proportion weight square: ;

[0128] Calculate the power of the absolute value of the chord change indicator: ;

[0129] Comprehensive calculation: ;

[0130] The result shows that the chord characteristic change and index distribution value of a certain genre level is 0.356, which indicates the degree of response of the genre level to the current chord change law and provides basic data for rule table generation.

[0131] S313: calling the chord characteristics and hierarchical node distribution association table, analyzing the matching degree of the chord characteristics in the genre hierarchy, classifying the chord change rules by the association values ​​of the hierarchical nodes, and generating a genre chord matching rule table;

[0132] By analyzing the distribution of chord change regularity values ​​at the genre level, the chord change regularity values ​​at the genre level are first classified, and the distribution characteristics of each group of regularity values ​​are screened, and divided into multiple intervals according to the variation range of the regularity values. Then, the regularity values ​​in each interval are normalized, and all regularity values ​​in the interval are weighted averaged to form the average value and variation trend characteristics of the interval. Then, the regularity values ​​of the interval are relabeled and classified according to the weight sorting of the genre level. Finally, the regularity values ​​of different intervals are grouped and organized into a rule table based on the genre weight and the distribution trend of the chord change regularity values. The rule table records the variation law, distribution trend and classification information of the chord characteristics of each genre level to form a genre chord matching rule table.

[0133] See also Figure 6 , the specific steps for obtaining the music work classification path rule set are:

[0134] S321: extracting structural parameters of chord sequences based on the genre chord matching rule table, performing frequency statistics, and classifying parameters according to genre matching conditions, summarizing weight values ​​corresponding to chord combinations, integrating weight data and chord combination information, and obtaining a chord matching genre classification weight table;

[0135] Each chord sequence is parsed one by one, and the characteristic parameters of each chord are extracted, including the chord type, duration, frequency change amplitude, and connection characteristics with the previous and next chords. The frequency of each chord type in the sequence is counted. According to the extracted chord characteristics, each parameter is classified according to the genre matching in the rule table. Combined with the chord characteristic description of each genre in the rule table, the combination of chords is matched with the characteristics of a specific genre. For the matching results of each chord combination, the corresponding weight value is summarized. The weight value reflects the importance of the chord combination in the genre. Through the induction of the weight value, the chord combination and the weight value are integrated one by one to generate a new data structure, in which each chord combination establishes an accurate correspondence with the weight value, and the genre characteristics of the chord combination are further marked. In the integration process, by comparing the frequency distribution in the chord sequence, the weight distribution differences between genres are evaluated, and finally a chord matching genre classification weight table is generated, in which the structural characteristics, weight values, and correspondence with the genre of the chord combination are clearly marked, providing weight data support for subsequent classification operations.

[0136] S322: using the chord matching genre classification weight table, extracting the time series interval corresponding to the weight value, performing segmented statistics on the weight change trend, extracting the continuously changing features and performing labeling integration, combining the genre matching classification information to perform trend quantification, and obtaining a genre dynamic weight change trend set;

[0137] The chord combination information in the weight table is mapped to the music timeline one by one. For each chord combination, its start time and end time in the music sequence are recorded, and its corresponding weight value is marked. After the mapping is completed, the weight change trend on the timeline is segmented and counted. Based on the weight fluctuation in each time interval, the fluctuation pattern and continuity characteristics of the weight change are analyzed. In the segmented statistical process, the features of continuous rise or fall in the weight change trend are extracted, and the feature change interval is marked. Combined with the genre matching classification information, the trend characteristics of the weight change are quantified. In the quantification process, the weight change rate, fluctuation amplitude and consistency of the change direction are analyzed to ensure that the trend characteristics can reflect the distribution law of the chord combination in the time series. By integrating the marked features, a genre dynamic weight change trend set is generated, which records in detail the weight change interval, trend characteristics, genre-related characteristics and other information, providing a dynamic trend basis for the subsequent genre classification path law analysis.

[0138] S323: using the genre dynamic weight change trend set, matching the weight change characteristics and chords in the trend data, and matching the chord combination in the genre classification weight table, extracting the classification path rule according to the corresponding relationship between the weight interval and the combination, integrating the path data, and generating a music work classification path rule set;

[0139] The weight change characteristics recorded in the trend set are compared with the specific combination forms in the chord sequence. According to the change interval and directional characteristics in the trend data, it is confirmed whether the chord combination characteristics in each time interval are consistent with the chord combination description in the genre classification weight table. After the matching is completed, the chord combination that meets the weight change characteristics is classified with the genre classification characteristics, and the weight interval characteristics and the corresponding relationship between the chord combinations in the time series are further extracted. In the matching process, the regular characteristics of weight changes are extracted by performing interval analysis on the weight values ​​in the genre classification weight table, and the classification path rules of the chord sequence are extracted and sorted according to the characteristics. When integrating the path data, the weight interval characteristics and the time distribution rules of the chord combination are combined to generate a music work classification path rule set. In the rule set, each classification path contains complete weight interval characteristics, chord combination characteristics, genre correspondence and classification rule description, providing a clear path basis and rule framework for the automatic classification of music works.

[0140] See also Figure 7 , the specific steps for obtaining the music work classification hierarchical structure table are:

[0141] S411: based on the music work classification path rule set, extracting the chord dynamic change data of the music work, analyzing the frequency, duration and rhythm characteristics of the chords, and performing dynamic matching, calculating the feature vector of each group of chords, and generating the music work chord dynamic change feature set;

[0142] By extracting the dynamic change data of chords of musical works, the chord frequency distribution calculation model is first called to perform frequency sampling analysis on the chord sequence of the musical works, and the frequency change information of each chord on the time axis is extracted. The duration of the chord is further summarized and feature extracted in combination with the note duration distribution model. The above two features are stored as time series vectors in the form of feature vectors. Then, the dynamic change rhythm of the chords is quantified by using the dynamic rhythm distribution parameters. The strength contrast and change trend of the chords are corrected by calling the rhythm weight adjustment function to obtain the dynamic feature value of the chord rhythm. The combination vector of the three key features of the frequency, duration and dynamic rhythm of the chords is integrated. The unified dimension and comparability of the feature vector are ensured through the standardization of the combination vector. The above processing results are stored as a chord feature set. The chord feature set is further dynamically analyzed to extract the time node distribution characteristics and the overall change trend of each group of chords in the time series. The predefined chord change pattern template is combined for matching, and the dynamic change pattern of each group of chords is summarized. The dynamic change pattern is stored as a dynamic change feature set corresponding to the time series feature, and the dynamic change feature set of the chords of the musical works is generated.

[0143] S412: calling the music work chord dynamic change feature set, and comparing and matching it with the path value in the predefined music genre path rule set, and extracting the path node and level information that conforms to the path rule according to the matching result, using the formula:

[0144] ;

[0145] Calculate the matching scores between the path nodes and the chord changes to obtain the node set matching the path rule;

[0146] in, represents the matching score between the path node and the chord change, Represents the dynamic characteristics of chords. represents the chord weight, Represents the node feature value in the genre path rule set, represents the node weight, Represents the path dynamic adjustment parameters;

[0147] The benefit of the formula is that by introducing dynamic adjustment parameters and path weight coefficients, the matching degree calculation of the path nodes can more accurately reflect the degree of compliance of the dynamic changes of the chord with the path rules, thereby improving the precision and accuracy of the matching;

[0148] , which is obtained by calculating the frequency and rhythm distribution characteristics of the chord dynamic change data, mainly by normalizing the chord time series vector and the chord weight distribution after convolution operation;

[0149] , according to the quantification of the importance coefficient of the chord frequency in the music work, combined with the priority coefficient of the chord frequency in the preset weight matrix;

[0150] ,The node feature values ​​extracted from the predefined genre path rule set are determined by parsing the specific node feature parameters of the path rule;

[0151] ,Based on the weight value assigned to the importance of the path node in the overall genre rules, the importance of each node is quantitatively assigned through the weight analysis model;

[0152] ,It is measured through the dynamic adjustment characteristics of the path, which is mainly obtained by parametric analysis after summarizing the distribution trend of the chord change amplitude and the path rule adaptation;

[0153] Substituting into the formula: ;

[0154] The matching score of the path node is 0.304, which means that the node is 30.4% consistent with the dynamic changes of the chords in the current music work. Based on the matching score results, we can further extract the node set that meets the path rules. The matching score can be used for the final judgment and analysis of the classification level.

[0155] S413: Based on the node set matched by the path rule, the node set is arranged in hierarchical order and matching score, and the correlation and hierarchical relationship between the hierarchical nodes are analyzed to generate a hierarchical structure table of music work classification;

[0156] Based on the hierarchical order and matching score of the path node set, the hierarchical relationship template set in the path rule set is called to screen and sort the matching score of each node. By setting the matching score threshold, the path nodes below the threshold are eliminated, and the node set above the threshold is retained as the candidate set. Subsequently, the node hierarchical order and the logical relationship parameters between nodes in the candidate set are calculated, the correlation strength between adjacent hierarchical nodes and the node importance weight are analyzed, and the path dynamic adjustment weight factor model is called to normalize the importance of the node hierarchy, completing the preliminary determination of the candidate set hierarchical structure. Then, the matching score of each node and the weight distribution between the hierarchies are analyzed, and the arrangement order and logical hierarchy of the nodes are further optimized. By performing structured matching of the optimized hierarchical relationship with the original path rules, a hierarchical structure table for the classification of musical works is generated.

[0157] The automatic classification system for musical works is used to implement the automatic classification method for musical works, and the system comprises:

[0158] The music genre hierarchical analysis module extracts the weight value and correlation relationship of each genre node based on the classification standard of music genre, calculates the upper and lower connection strength values ​​between nodes, compares the distribution of node weight values, identifies the distribution rules and correlation ratios between levels, and generates a music genre hierarchical weight table;

[0159] The chord feature extraction module extracts music chord structure information based on the music genre hierarchical weight table, divides the chord data into time segments, compares the chord frequency and duration in each segment, arranges the chord change feature values ​​into time series, and generates a chord feature change sequence table;

[0160] The genre chord rule analysis module extracts the correspondence between the chord change index and the genre level weight node value based on the chord characteristic change sequence table, analyzes the chord dynamic change and level distribution matching value, identifies the dynamic change law and the level matching rule association value, and generates the genre chord matching rule table;

[0161] The dynamic classification path generation module is based on the genre chord matching rule table, analyzes the dynamic change value distribution of music chords, extracts the matching node values ​​in the rule table, analyzes the node weight distribution, adjusts the hierarchical order and weight relationship, identifies the dynamic path node hierarchical connection rules, and generates a music work classification path rule set;

[0162] The classification hierarchical structure output module is based on the music work classification path rule set, extracts the path rule hierarchical node distribution relationship, analyzes the correspondence between the node distribution value and the classification hierarchical structure, organizes the music genre classification hierarchical information into hierarchical structure data, and generates a music work classification hierarchical structure table.

[0163] 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 method for automatically classifying musical works, characterized in that: The following steps are involved: Based on the classification standards of music genres, analyze the hierarchical node relationships of music genres, extract the corresponding hierarchical characteristics of genre categories and subcategories, identify the correlation ratio between each level, analyze the hierarchical connection strength, and generate a music genre hierarchical weight table; Based on the music genre level weight table, extract the chord structure information of the music work, segment the chord data segments according to the time series, compare the chord frequency and duration of the segments, extract the key characteristics of the chord changes, generate a music chord change amplitude table, match the frequency change and position relationship of the sequence segments, extract the chord characteristic dynamic change index, and generate a chord characteristic change sequence table; Based on the chord characteristic change sequence table, the correlation value between the chord change index and the genre level weight is analyzed, the distribution relationship between the chord change rule and the level node is extracted, the genre chord matching rule table is generated, the dynamic weight change of the genre classification is extracted and analyzed, and a music work classification path rule set is generated; Based on the music work classification path rule set, the music genre path value is dynamically matched and the weight change is adjusted, the genre path value is compared and analyzed with the dynamic change data of the work chords, the path nodes matching the genre rules are extracted, and the hierarchical structure table of the music work classification is generated by combining the hierarchical order and the node value.

2. The method for automatically classifying musical works according to claim 1, characterized in that: The steps for obtaining the music genre level weight table are specifically as follows: Based on the classification standard of music genres, the basic classification characteristics of genres are extracted from the hierarchical node relationships of music genres, and the subordinate major and minor categories of each genre node are analyzed. By analyzing the correlation ratio between each genre node and the upper level node, an initial set of node correlation ratios is generated; The initial set of node association ratios is called to analyze the connection characteristics between differentiated levels, identify the differences in the level association ratios of genre categories and subcategories, and set a preliminary indicator of the level connection strength using the formula: ； Calculate the hierarchical connection strength values ​​from the genre category to the subcategory and generate a hierarchical connection strength set; in, Represents the hierarchical connection strength value from the main category to the subcategory of the genre, Represents the weight parameter between major categories and minor categories, Represents the correlation ratio between major categories and minor categories, is the total number of nodes; By using the hierarchical connection strength set and combining the hierarchical node quantity weights of differentiated genres, a hierarchical weight analysis is performed between major and minor categories of each genre. The hierarchical weight values ​​are optimized through recursive iteration to generate a music genre hierarchical weight table.

3. The method for automatically classifying musical works according to claim 2, characterized in that: The steps for obtaining the music chord variation range table are specifically as follows: Based on the music genre level weight table, extract the chord information in the music work, and divide it into multiple chord segments according to the time series, analyze the chord type in each segment, calculate the occurrence frequency and duration of the chord, and generate a chord frequency and duration distribution table through basic statistics of the chord data in the segment; According to the chord frequency and duration distribution table, the chord frequency and duration of each segment are compared, and the variation range of the chords in the segment is analyzed, and the formula is used: ； Get the chord change amplitude data of the fragment; in, Indicates the amplitude of the chord change in the fragment. and Indicates the chord frequency of the current segment and the previous segment, and Indicates the duration of the chords in the current and previous segments. , , are the adjustment coefficients for chord frequency and duration, is the additional weight coefficient for chord changes; The chord change amplitude data of the fragment is called, and the change trend of the chord change amplitude in the differentiated time periods is analyzed. The chord type and dynamic characteristics in the fragment are classified according to the change amplitude data, and the amplitude weight is adjusted in combination with the change trend to generate a music chord change amplitude table.

4. The method for automatically classifying musical works according to claim 3, characterized in that: The steps for obtaining the chord characteristic change sequence table are specifically as follows: Based on the music chord variation amplitude table, the starting frequency and ending frequency of each chord are extracted, the frequency range of each chord is sorted in chronological order, and the rising and falling trends of the chord frequencies are compared according to the time series, the frequency fluctuation amplitude and direction are recorded, and a frequency variation and position relationship table is generated; Based on the frequency change and position relationship table, the time series is segmented according to the chord mark, the frequency change range in each segment is counted segment by segment, the frequency fluctuation starting point and end point are divided into intervals, and the chord position mark and frequency change trend are matched segment by segment, the segment duration and span are recorded, and the chord segment frequency change table is generated; Based on the chord segment frequency change table, the chord change index of each segment is analyzed, the frequency change amplitude and direction within the segment are extracted, the continuity and transition characteristics between segments are recorded, and the dynamic characteristics of the segments are integrated in sequence to construct a chord characteristic change sequence table.

5. The method for automatically classifying musical works according to claim 4, characterized in that: The steps for obtaining the genre chord matching rule table are specifically as follows: Based on the chord characteristic change sequence table, by analyzing the chord change index, extracting the relationship node between each chord characteristic and the corresponding genre level, performing correlation operation on the frequency value of the chord change index and the genre level weight, and generating a preliminary correlation matrix between the chord change index and the genre level weight; The preliminary correlation matrix of the chord change index and the genre level weight is called to analyze the distribution relationship between the chord change law and the level nodes. By analyzing the weighted distribution of the level nodes corresponding to the chord characteristic changes, the formula is adopted: ； Establish a table associating chord characteristics with hierarchical node distribution; in, Indicates the level of genre The chord change rule value, represents the genre level weight, Indicates the time proportion of chord changes. and Indicates the value of the adjacent chord characteristic change index, is the adjustment coefficient of time proportion, is the weight parameter of the chord characteristic change index; The chord characteristics and hierarchical node distribution association table is called to analyze the matching degree of the chord characteristics in the genre hierarchy, and the chord change rules are classified according to the association values ​​of the hierarchical nodes to generate a genre chord matching rule table.

6. The method for automatically classifying musical works according to claim 5, characterized in that: The steps for obtaining the music work classification path rule set are specifically as follows: Based on the genre chord matching rule table, the structural parameters of the chord sequence are extracted, the frequency statistics are performed, and the parameters are classified according to the genre matching situation, the weight values ​​corresponding to the chord combinations are summarized, and the weight data and the chord combination information are integrated to obtain the chord matching genre classification weight table; The chord matching genre classification weight table is used to extract the time series interval corresponding to the weight value, perform segmented statistics on the weight change trend, extract the features of continuous changes and perform labeling integration, combine the genre matching classification information to perform trend quantification, and obtain the genre dynamic weight change trend set; The genre dynamic weight change trend set is adopted to match the weight change characteristics and chords in the trend data, and the chord combinations in the genre classification weight table are matched. The classification path rules are extracted according to the corresponding relationship between the weight interval and the combination, the path data is integrated, and a music work classification path rule set is generated.

7. The method for automatically classifying musical works according to claim 6, characterized in that: The steps for obtaining the music work classification hierarchical structure table are specifically as follows: Based on the music work classification path rule set, extract the chord dynamic change data of the music work, analyze the frequency, duration and rhythm characteristics of the chords, and perform dynamic matching, calculate the feature vector of each group of chords, and generate the music work chord dynamic change feature set; The music work chord dynamic change feature set is called and compared with the path value in the predefined music genre path rule set, and the path nodes and hierarchical information that meet the path rules are extracted according to the matching results, using the formula: ； Calculate the matching scores between the path nodes and the chord changes to obtain the node set matching the path rule; in, represents the matching score between the path node and the chord change, Represents the dynamic characteristics of chords. represents the chord weight, Represents the node feature value in the genre path rule set, represents the node weight, Represents the path dynamic adjustment parameters; Based on the node set matched by the path rule, the node set is arranged in hierarchical order and matching score, and the correlation and hierarchical relationship between the hierarchical nodes are analyzed to generate a hierarchical structure table for the classification of musical works.

8. A system for automatically classifying musical works, characterized in that: According to any one of claims 1 to 7, the method for automatically classifying musical works comprises: The music genre hierarchical analysis module extracts the weight value and correlation relationship of each genre node based on the classification standard of music genre, calculates the upper and lower connection strength values ​​between nodes, compares the distribution of node weight values, identifies the distribution rules and correlation ratios between levels, and generates a music genre hierarchical weight table; The chord characteristic extraction module extracts the music chord structure information based on the music genre level weight table, divides the chord data into time segments, compares the chord frequency and duration in each segment, arranges the chord change characteristic values ​​into time series, and generates a chord characteristic change sequence table; The genre chord rule analysis module extracts the corresponding relationship between the chord change index and the genre level weight node value based on the chord characteristic change sequence table, analyzes the chord dynamic change and the level distribution matching value, identifies the dynamic change law and the level matching rule association value, and generates the genre chord matching rule table; The dynamic classification path generation module analyzes the distribution of dynamic changes in music chord values ​​based on the genre chord matching rule table, extracts the matching node values ​​from the rule table, analyzes the node weight distribution, adjusts the hierarchical order and weight relationship, identifies the dynamic path node hierarchical connection rules, and generates a music work classification path rule set; The classification hierarchical structure output module extracts the path rule hierarchical node distribution relationship based on the music work classification path rule set, analyzes the correspondence between the node distribution value and the classification hierarchical structure, organizes the music genre classification hierarchical information into hierarchical structure data, and generates a music work classification hierarchical structure table.

Images