Music data intelligent clustering system and method based on audio feature analysis

Abstract

The application provides a music data intelligent clustering system and method based on audio feature analysis, comprising: collecting updated music audio data of music software, identifying beat cycle features and melody emotion features of the updated music audio data, identifying music categories corresponding to the updated music audio data according to the beat cycle features and the melody emotion features, constructing clustering adaptive parameters according to category features corresponding to each of the music categories, determining clustering similarity between the updated music audio data and each of the music categories according to the clustering adaptive parameters, and dividing the updated music audio data into corresponding target music categories, so that music is classified by using a clustering technique combined with a feature identification technique, the drawbacks of traditional techniques are solved, and the accuracy of music data classification is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intelligent clustering system and method for music data based on audio feature analysis. Background Technology

[0002] With the development of multimedia and internet information technology, people's demand for retrieving audio information resources is increasing, making the efficient classification of music information a current research hotspot. Simultaneously, with the development of the internet, the amount of music data has exploded, with an average of 35 million different songs on each music platform, and a large number of albums being released online every week. This makes it difficult for traditional statistical and computational methods to quickly classify the massive amount of music.

[0003] Currently, although cluster analysis is applied in many fields, existing music data clustering methods have many shortcomings, such as inaccurate feature extraction, poor clustering algorithm performance, inability to meet the needs of music recommendation and classification management, and inability to respond to the rapidly developing Internet.

[0004] Therefore, this invention provides an intelligent clustering system and method for music data based on audio feature analysis. Summary of the Invention

[0005] This invention relates to an intelligent music data clustering system and method based on audio feature analysis. It uses clustering technology combined with feature recognition technology to classify music, which solves the drawbacks of traditional technologies and improves the accuracy of music data classification.

[0006] This invention provides an intelligent clustering system for music data based on audio feature analysis, comprising:

[0007] The data acquisition module is used to collect updated music audio data from music software.

[0008] The feature extraction module is used to identify the beat cycle features and melody emotion features of the updated music audio data;

[0009] The clustering processing module is used to identify the music category corresponding to the updated music audio data based on the beat cycle feature and the melody emotion feature, and to construct clustering adaptation parameters based on the category features corresponding to each music category;

[0010] The clustering optimization module is used to determine the clustering similarity between the updated music audio data and each music category based on the clustering adaptation parameters, and to classify the updated music audio data into the corresponding target music category.

[0011] In one feasible approach

[0012] The data acquisition module includes:

[0013] A data capture unit is used to capture real-time, complete update data of the music software;

[0014] The data filtering unit is used to perform duration analysis and type analysis on each of the real-time complete update data, and filter the updated music and audio data that meet the specified duration and type.

[0015] In one feasible approach

[0016] The feature extraction module includes:

[0017] The data processing unit is used to perform Fourier transform on the updated music audio data to obtain several frequency components, count the frequency frequency corresponding to each frequency component, determine the main frequency component of the updated music audio data, and draw each frequency component in the frequency domain space to obtain the phase information corresponding to each frequency component.

[0018] A beat determination unit is used to determine several single beat features of the updated music audio data based on the frequency difference between each frequency component and the main frequency component, determine the main beat feature of the updated music audio data based on the main frequency component, and perform beat correction on the main beat feature using each single beat feature to obtain the beat information of the updated music audio data.

[0019] The period determination unit is used to generate an initial period based on the main phase information corresponding to the main frequency component, decompose the beat information using the initial period to obtain several decomposed beats, and adjust the period range of the initial period according to the beat difference between different decomposed beats.

[0020] The melody determination unit is used to perform melody decomposition on the updated music audio data to obtain the melody features corresponding to the updated music audio data under different melody dimensions, and to determine the core melody of the updated music audio data based on the feature value corresponding to each melody feature.

[0021] The sentiment determination unit is used to perform text recognition on the updated music audio data using natural language processing technology to obtain the corresponding data text, to perform sentiment annotation on the data text using a preset sentiment dictionary to determine the sentiment tendency of the updated music audio data, and to train the sentiment tendency using the random forest method to obtain the data sentiment of the updated music audio data.

[0022] The feature determination unit is used to generate the beat period feature of the updated music audio data based on the beat information and the period range, and to generate the melody emotion feature of the updated music audio data based on the core melody and the data emotion.

[0023] In one feasible approach

[0024] The melody determination unit includes:

[0025] The melody decomposition subunit is used to decompose the updated music audio data into melody to obtain the pitch features, volume features, speech rate features, duration features, and accent features of the updated music audio data.

[0026] The numerical analysis subunit is used to analyze the numerical level of the corresponding feature value according to the numerical grading standard corresponding to each melody dimension, combine the numerical levels, and generate the combined melody information of the updated music audio data.

[0027] The melody reconstruction subunit is used to construct pure audio based on the combined melody information, and to generate the core melody of the updated music audio data based on the pure melody of the pure audio.

[0028] In one feasible approach

[0029] The clustering processing module includes:

[0030] The preliminary classification unit is used to match the updated music audio data with a corresponding first music category set based on the beat cycle feature, match the updated music audio data with a corresponding second music category set based on the melody emotion feature, and determine the initial music category corresponding to the updated music audio data based on the intersection of the first music category set and the second music category set.

[0031] The clustering analysis unit is used to obtain the existing music audio data corresponding to each initial music category, construct a category attribute map corresponding to each initial music category based on the music tags corresponding to each existing music audio data, input the updated music audio data into the category attribute map for classification verification, obtain the matching nodes of each updated music audio data in each category attribute map, and filter music categories whose node distance between the matching node and the corresponding existing node is less than a specified distance.

[0032] The parameter generation unit is used to generate category features corresponding to the music category based on the category attribute map, fit and match the beat cycle feature with the category feature to generate the beat cycle parameter of the updated music audio data in the corresponding music category, fit and match the melody emotion feature with the category feature to generate the melody emotion parameter of the updated music audio data in the corresponding music category, and fuse the beat cycle parameter and the melody emotion parameter to generate the clustering adaptation parameter of the updated music audio data in each music category.

[0033] In one feasible approach

[0034] The preliminary classification unit is also used for:

[0035] When there is no intersection between the first music category set and the second music category set, it is determined that the updated music audio data is abnormal, and a re-acquisition instruction is generated.

[0036] The reacquisition command is used to control the data acquisition module to reacquire the updated music audio data.

[0037] In one feasible approach

[0038] The clustering optimization module includes:

[0039] The clustering verification unit is used to iteratively cluster the updated music audio data with the corresponding music category based on the clustering adaptation parameters, so as to obtain the data similarity features between the updated music audio data and the existing music audio data in each music category.

[0040] The clustering optimization unit is used to statistically analyze several data similarity features corresponding to each music category to determine the clustering similarity between the updated music audio data and each music category, and to determine several target music categories corresponding to each updated music audio data based on the clustering similarity.

[0041] The division execution unit is used to filter relevant music audio data with a cluster similarity higher than the standard similarity to the updated music audio data in each of the target music categories, set recommendation guidelines for the updated music audio data, and divide the updated music audio data into the corresponding target music categories.

[0042] In one feasible approach

[0043] Also includes:

[0044] After updating the music audio data to the corresponding target music category, the updated music audio data is regarded as existing music audio data.

[0045] This invention provides an intelligent clustering method for music data based on audio feature analysis, including:

[0046] Step 1: Collect updated music audio data from the music software;

[0047] Step 2: Identify the beat cycle characteristics and melodic emotional characteristics of the updated music audio data;

[0048] Step 3: Identify the music category corresponding to the updated music audio data based on the beat cycle feature and the melody emotion feature, and construct clustering adaptation parameters based on the category features corresponding to each music category;

[0049] Step 4: Determine the clustering similarity between the updated music audio data and each music category based on the clustering adaptation parameters, and classify the updated music audio data into the corresponding target music category.

[0050] In one feasible approach

[0051] Step 2 includes:

[0052] Step 21: Perform Fourier transform on the updated music audio data to obtain several frequency components, count the frequency frequency corresponding to each frequency component, determine the main frequency component of the updated music audio data, and draw each frequency component in the frequency domain space to obtain the phase information corresponding to each frequency component.

[0053] Step 22: Determine several single beat features of the updated music audio data based on the frequency difference between each frequency component and the main frequency component, determine the main beat feature of the updated music audio data based on the main frequency component, and use each single beat feature to perform beat correction on the main beat feature to obtain the beat information of the updated music audio data.

[0054] Step 23: Generate an initial period based on the principal phase information corresponding to the principal frequency component, decompose the beat information using the initial period to obtain several decomposed beats, and adjust the period range of the initial period according to the beat differences between different decomposed beats.

[0055] Step 24: Perform melody decomposition on the updated music audio data to obtain the melody features corresponding to the updated music audio data under different melody dimensions, and determine the core melody of the updated music audio data based on the feature value corresponding to each melody feature;

[0056] Step 25: Use natural language processing technology to perform text recognition on the updated music audio data to obtain the corresponding data text, use a preset sentiment dictionary to perform sentiment annotation on the data text to determine the sentiment tendency of the updated music audio data, and use the random forest method to train the sentiment tendency to obtain the data sentiment of the updated music audio data.

[0057] Step 26: Generate the beat period feature of the updated music audio data based on the beat information and the period range, and generate the melody emotion feature of the updated music audio data based on the core melody and the data emotion.

[0058] The beneficial effects of the above technical solution are as follows: When music audio data is updated in music software, the data is collected in a timely manner, and the beat cycle characteristics and melodic emotional characteristics of the data are identified. Then, based on the beat cycle characteristics and melodic emotional characteristics, the updated music audio data is matched with the corresponding classification, and corresponding clustering adaptation parameters are constructed. These parameters are then used to optimize the clustering of the updated music audio data, determining the target music category that the updated music audio data corresponds to. In this way, updated music audio data in music software can be captured in a timely manner, and its necessary characteristics can be identified and then clustered. To improve the accuracy of clustering, clustering adaptation parameters between updated music audio data and different music categories are determined. These parameters are then used to deeply analyze the clustering similarity between the two, thereby determining the target music category of the updated music audio data. This improves the accuracy of traditional clustering techniques, allows for the assignment of suitable categories to updated music audio data, makes it easier for listeners to select their favorite music categories, and increases listeners' liking for the music software.

[0059] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0060] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0061] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0062] Figure 1 This is a schematic diagram of the composition of the intelligent music data clustering system based on audio feature analysis in an embodiment of the present invention;

[0063] Figure 2 This is a schematic diagram illustrating the workflow of the intelligent clustering method for music data based on audio feature analysis in an embodiment of the present invention. Detailed Implementation

[0064] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0065] Example 1

[0066] This embodiment provides an intelligent music data clustering system based on audio feature analysis, such as... Figure 1 As shown, it includes:

[0067] The data acquisition module is used to collect updated music audio data from music software.

[0068] The feature extraction module is used to identify the beat cycle features and melody emotion features of the updated music audio data;

[0069] The clustering processing module is used to identify the music category corresponding to the updated music audio data based on the beat cycle feature and the melody emotion feature, and to construct clustering adaptation parameters based on the category features corresponding to each music category;

[0070] The clustering optimization module is used to determine the clustering similarity between the updated music audio data and each music category based on the clustering adaptation parameters, and to classify the updated music audio data into the corresponding target music category.

[0071] In this example, the beat cycle feature represents the rhythmic basis and repetition cycle of the updated music audio data;

[0072] In this example, the melodic emotion feature represents the melody composed of the notes that update the music audio data and the emotion it expresses;

[0073] In this example, the clustering adaptation parameter represents the parameter generated when adapting the music audio data to a music category after updating the clustering of the music audio data to that music category;

[0074] In this example, cluster similarity represents the similarity between the updated music audio data and existing music audio data in the music category;

[0075] In this example, the number of target music categories can be one or more.

[0076] The working principle and beneficial effects of the above technical solution are as follows: When music audio data is updated in the music software, the data is collected in a timely manner, and the beat cycle characteristics and melodic emotional characteristics of the data are identified. Then, based on the beat cycle characteristics and melodic emotional characteristics, the updated music audio data is matched with the corresponding classification, and corresponding clustering adaptation parameters are constructed. These parameters are then used to optimize the clustering of the updated music audio data, determining the target music category that the updated music audio data corresponds to. In this way, updated music audio data in the music software can be captured in a timely manner, and its necessary characteristics can be identified and then clustered. In order to improve the accuracy of clustering, clustering adaptation parameters between the updated music audio data and different music categories are determined. Then, these parameters are used to deeply analyze the clustering similarity between the two, thereby determining the target music category of the updated music audio data. This improves the accuracy of traditional clustering technology, allows for the assignment of suitable categories to updated music audio data, makes it easier for listeners to select their favorite music categories, and increases listeners' liking for the music software.

[0077] Example 2

[0078] Based on Example 1, the music data intelligent clustering system based on audio feature analysis includes a data acquisition module comprising:

[0079] A data capture unit is used to capture real-time, complete update data of the music software;

[0080] The data filtering unit is used to perform duration analysis and type analysis on each of the real-time complete update data, and filter the updated music and audio data that meet the specified duration and type.

[0081] In this example, real-time complete update data refers to complete data that has been updated.

[0082] In this example, duration analysis refers to analyzing the music duration of real-time complete updated data, and type analysis refers to identifying the data type of real-time complete updated data.

[0083] In this example, the specified duration is 20 seconds;

[0084] In this example, the specified type represents one or both of the music type and audio type.

[0085] The working principle and beneficial effects of the above technical solution are as follows: timely capture of updated data in music software, analysis of which, and screening of updated music audio data that meet the specified duration and type are selected for the next step of clustering and classification. In this way, real-time complete updated data can be initially screened to avoid irrelevant data from being mixed in and interfering with the clustering quality and efficiency.

[0086] Example 3

[0087] Based on Example 1, the intelligent clustering system for music data based on audio feature analysis includes a feature extraction module comprising:

[0088] The data processing unit is used to perform Fourier transform on the updated music audio data to obtain several frequency components, count the frequency frequency corresponding to each frequency component, determine the main frequency component of the updated music audio data, and draw each frequency component in the frequency domain space to obtain the phase information corresponding to each frequency component.

[0089] A beat determination unit is used to determine several single beat features of the updated music audio data based on the frequency difference between each frequency component and the main frequency component, determine the main beat feature of the updated music audio data based on the main frequency component, and perform beat correction on the main beat feature using each single beat feature to obtain the beat information of the updated music audio data.

[0090] The period determination unit is used to generate an initial period based on the main phase information corresponding to the main frequency component, decompose the beat information using the initial period to obtain several decomposed beats, and adjust the period range of the initial period according to the beat difference between different decomposed beats.

[0091] The melody determination unit is used to perform melody decomposition on the updated music audio data to obtain the melody features corresponding to the updated music audio data under different melody dimensions, and to determine the core melody of the updated music audio data based on the feature value corresponding to each melody feature.

[0092] The sentiment determination unit is used to perform text recognition on the updated music audio data using natural language processing technology to obtain the corresponding data text, to perform sentiment annotation on the data text using a preset sentiment dictionary to determine the sentiment tendency of the updated music audio data, and to train the sentiment tendency using the random forest method to obtain the data sentiment of the updated music audio data.

[0093] The feature determination unit is used to generate the beat period feature of the updated music audio data based on the beat information and the period range, and to generate the melody emotion feature of the updated music audio data based on the core melody and the data emotion.

[0094] In this example, frequency components represent the components related to different frequencies contained in the updated music audio data;

[0095] In this example, the frequency count represents the number of times a frequency component occurs;

[0096] In this example, the dominant frequency component represents the frequency component that appears most frequently in the updated music audio data;

[0097] In this example, the phase information represents the waveform presented by the frequency components;

[0098] In this example, the frequency difference represents the phase shift between the frequency components and the main frequency component;

[0099] In this example, a single beat feature represents the beat presented by a frequency component;

[0100] In this example, the main beat feature represents the beat presented by the main frequency component;

[0101] In this example, the beat information represents the continuous beats presented in the updated music audio data;

[0102] In this example, the process of rhythm correction for the main beat feature is as follows: the process of correcting the notes of the main beat feature using a single beat feature;

[0103] In this example, the initial period represents the period derived from the principal phase information;

[0104] In this example, the period range represents the range of the period length of the initial period;

[0105] In this example, the melody dimension includes pitch dimension, volume dimension, speech rate dimension, duration dimension, and accent dimension;

[0106] In this example, the core melody represents the melody presented multiple times when the music audio data is updated;

[0107] In this example, sentiment labeling is represented as the process of matching data text with corresponding sentiments;

[0108] In this example, the preset sentiment dictionary contains several words, each of which is assigned a sentiment value. The sentiment tendency of a text is identified by statistically analyzing the sentiment values ​​contained in the text.

[0109] In this example, the data sentiment representation reflects the sentiment conveyed by the updated music audio data.

[0110] The working principle and beneficial effects of the above technical solution are as follows: In order to identify the various features of the updated music audio data, Fourier transform is used to identify the frequency components of the updated music audio data. By statistically analyzing the frequency count of each frequency component, the dominant frequency component of the updated music audio data is determined. Simultaneously, the frequency components are plotted in the frequency domain to determine the phase information of each frequency component. Then, based on the frequency difference between the frequency components and the dominant frequency component, the single beat feature and the dominant beat feature of the updated music audio data are determined, thereby determining the beat information of the updated music audio data. Next, the initial period of the updated music audio data is identified, and the initial period is adjusted using beat differences. Then, the core melody of the updated music audio data is identified by analyzing the melody of the updated music audio data. Furthermore, the data text of the updated music audio data is identified, and sentiment annotation is performed on the data text to train the data sentiment of the updated music audio data. Finally, based on the known information, the beat period feature and melody sentiment feature of the updated music audio data are generated. In this way, the updated music audio data can be analyzed in depth, and all the information of the updated music audio data is analyzed in detail, generating accurate beat period features and melody sentiment features, ensuring the accuracy of subsequent cluster analysis.

[0111] Example 4

[0112] Based on Example 3, the music data intelligent clustering system based on audio feature analysis, wherein the melody determination unit includes:

[0113] The melody decomposition subunit is used to decompose the updated music audio data into melody to obtain the pitch features, volume features, speech rate features, duration features, and accent features of the updated music audio data.

[0114] The numerical analysis subunit is used to analyze the numerical level of the corresponding feature value according to the numerical grading standard corresponding to each melody dimension, combine the numerical levels, and generate the combined melody information of the updated music audio data.

[0115] The melody reconstruction subunit is used to construct pure audio based on the combined melody information, and to generate the core melody of the updated music audio data based on the pure melody of the pure audio.

[0116] In this example, the pitch feature represents the frequency feature that updates the music audio data;

[0117] In this example, the volume feature represents the intensity of updating the music audio data;

[0118] In this example, the speech rate feature represents the speed at which the music audio data is updated;

[0119] In this example, the duration feature represents the duration of each syllable, word, or sentence that updates the music audio data;

[0120] In this example, the accent feature represents the emphasized portion of the updated music audio data;

[0121] In this example, the numerical grading standard represents the grading standard corresponding to a melody dimension.

[0122] The working principle and beneficial effects of the above technical solution are as follows: When performing melody decomposition on the updated music audio data, numerical analysis is performed on its various dimensions to determine the combined melody information of the updated music audio data. Then, a pure audio is constructed to identify the core melody. In this way, the core melody of the updated music audio data is determined. The simplicity of the pure audio notes is utilized to achieve melody recognition and determine the core melody of the updated music audio data.

[0123] Example 5

[0124] Based on Example 1, the intelligent clustering system for music data based on audio feature analysis includes a clustering processing module comprising:

[0125] The preliminary classification unit is used to match the updated music audio data with a corresponding first music category set based on the beat cycle feature, match the updated music audio data with a corresponding second music category set based on the melody emotion feature, and determine the initial music category corresponding to the updated music audio data based on the intersection of the first music category set and the second music category set.

[0126] The clustering analysis unit is used to obtain the existing music audio data corresponding to each initial music category, construct a category attribute map corresponding to each initial music category based on the music tags corresponding to each existing music audio data, input the updated music audio data into the category attribute map for classification verification, obtain the matching nodes of each updated music audio data in each category attribute map, and filter music categories whose node distance between the matching node and the corresponding existing node is less than a specified distance.

[0127] The parameter generation unit is used to generate category features corresponding to the music category based on the category attribute map, fit and match the beat cycle feature with the category feature to generate the beat cycle parameter of the updated music audio data in the corresponding music category, fit and match the melody emotion feature with the category feature to generate the melody emotion parameter of the updated music audio data in the corresponding music category, and fuse the beat cycle parameter and the melody emotion parameter to generate the clustering adaptation parameter of the updated music audio data in each music category.

[0128] In this example, the initial music category represents the music category that matches the updated music audio data;

[0129] In this example, the category attribute graph represents a graph of attribute relationships among all music in a music category, constructed by using music tags from existing music audio data as nodes;

[0130] In this example, the distance is specified as 1.

[0131] The working principle and beneficial effects of the above technical solution are as follows: First, the updated music audio data is matched with a first music category set based solely on the beat cycle feature. Simultaneously, the updated music audio data is matched with a second music category set based solely on the melody emotion feature. The initial music category of the updated music audio data is determined based on the intersection of the two music category sets. Then, a category attribute map is generated for the music tags corresponding to the existing music audio data of each initial music category. This map is then used to classify and verify the updated music audio data, filtering out music categories that meet the node distance requirements. Furthermore, the beat cycle feature of the updated music audio data is matched with the category features of the music categories, and the melody emotion feature is matched with the category features to determine the beat cycle parameters and melody emotion parameters of the updated music audio data. In this way, clustering adaptation parameters for the updated music audio data are generated. During the parameter generation process, both the beat cycle feature and the melody emotion feature are matched with the category features of the music categories, which improves the accuracy of the matching, avoids randomness, and ensures the effectiveness of clustering.

[0132] Example 6

[0133] Based on Example 5, the preliminary classification unit of the music data intelligent clustering system based on audio feature analysis is further used for:

[0134] When there is no intersection between the first music category set and the second music category set, it is determined that the updated music audio data is abnormal, and a re-acquisition instruction is generated.

[0135] The reacquisition command is used to control the data acquisition module to reacquire the updated music audio data.

[0136] The working principle and beneficial effects of the above technical solution are as follows: When the initial music category cannot be generated, the control data acquisition module reacquires and updates the music audio data, ensuring the integrity and accuracy of the updated music audio data.

[0137] Example 7

[0138] Based on Example 1, the intelligent clustering system for music data based on audio feature analysis includes a clustering optimization module comprising:

[0139] The clustering verification unit is used to iteratively cluster the updated music audio data with the corresponding music category based on the clustering adaptation parameters, so as to obtain the data similarity features between the updated music audio data and the existing music audio data in each music category.

[0140] The clustering optimization unit is used to statistically analyze several data similarity features corresponding to each music category to determine the clustering similarity between the updated music audio data and each music category, and to determine several target music categories corresponding to each updated music audio data based on the clustering similarity.

[0141] The division execution unit is used to filter relevant music audio data with a cluster similarity higher than the standard similarity to the updated music audio data in each of the target music categories, set recommendation guidelines for the updated music audio data, and divide the updated music audio data into the corresponding target music categories.

[0142] In this example, iterative clustering represents the process of adjusting the characteristics of updated music audio data using clustering adaptation parameters, and then matching it multiple times with music categories;

[0143] In this example, the data similarity feature represents the data similarity characteristics between existing music audio data and updated music audio data.

[0144] The working principle and beneficial effects of the above technical solution are as follows: By iteratively clustering the updated music audio data and music categories, the data similarity between the updated music audio data and each existing music in the music category is determined. This determines the cluster similarity between the updated music audio data and each music category, thereby identifying the target music category. Recommendations are then made to the updated music audio data within the target music category, improving the accurate exposure of the updated music audio data and enhancing the up-to-date performance of the music software.

[0145] Example 8

[0146] Based on Example 7, the intelligent clustering system for music data based on audio feature analysis further includes:

[0147] After updating the music audio data to the corresponding target music category, the updated music audio data is regarded as existing music audio data.

[0148] The working principle and beneficial effects of the above technical solution are as follows: After updating the target music category, the updated music audio data is regarded as existing music audio data, which facilitates the next clustering.

[0149] Example 9

[0150] This embodiment provides an intelligent clustering method for music data based on audio feature analysis, such as... Figure 2 As shown, it includes:

[0151] Step 1: Collect updated music audio data from the music software;

[0152] Step 2: Identify the beat cycle characteristics and melodic emotional characteristics of the updated music audio data;

[0153] Step 3: Identify the music category corresponding to the updated music audio data based on the beat cycle feature and the melody emotion feature, and construct clustering adaptation parameters based on the category features corresponding to each music category;

[0154] Step 4: Determine the clustering similarity between the updated music audio data and each music category based on the clustering adaptation parameters, and classify the updated music audio data into the corresponding target music category.

[0155] In this example, the beat cycle feature represents the rhythmic basis and repetition cycle of the updated music audio data;

[0156] In this example, the melodic emotion feature represents the melody composed of the notes that update the music audio data and the emotion it expresses;

[0157] In this example, the clustering adaptation parameter represents the parameter generated when adapting the music audio data to a music category after updating the clustering of the music audio data to that music category;

[0158] In this example, cluster similarity represents the similarity between the updated music audio data and existing music audio data in the music category;

[0159] In this example, the number of target music categories can be one or more.

[0160] The working principle and beneficial effects of the above technical solution are as follows: When music audio data is updated in the music software, the data is collected in a timely manner, and the beat cycle characteristics and melodic emotional characteristics of the data are identified. Then, based on the beat cycle characteristics and melodic emotional characteristics, the updated music audio data is matched with the corresponding classification, and corresponding clustering adaptation parameters are constructed. These parameters are then used to optimize the clustering of the updated music audio data, determining the target music category that the updated music audio data corresponds to. In this way, updated music audio data in the music software can be captured in a timely manner, and its necessary characteristics can be identified and then clustered. In order to improve the accuracy of clustering, clustering adaptation parameters between the updated music audio data and different music categories are determined. Then, these parameters are used to deeply analyze the clustering similarity between the two, thereby determining the target music category of the updated music audio data. This improves the accuracy of traditional clustering technology, allows for the assignment of suitable categories to updated music audio data, makes it easier for listeners to select their favorite music categories, and increases listeners' liking for the music software.

[0161] Example 10

[0162] Based on Example 9, the intelligent clustering method for music data based on audio feature analysis, step 2 includes:

[0163] Step 21: Perform Fourier transform on the updated music audio data to obtain several frequency components, count the frequency frequency corresponding to each frequency component, determine the main frequency component of the updated music audio data, and draw each frequency component in the frequency domain space to obtain the phase information corresponding to each frequency component.

[0164] Step 22: Determine several single beat features of the updated music audio data based on the frequency difference between each frequency component and the main frequency component, determine the main beat feature of the updated music audio data based on the main frequency component, and use each single beat feature to perform beat correction on the main beat feature to obtain the beat information of the updated music audio data.

[0165] Step 23: Generate an initial period based on the principal phase information corresponding to the principal frequency component, decompose the beat information using the initial period to obtain several decomposed beats, and adjust the period range of the initial period according to the beat differences between different decomposed beats.

[0166] Step 24: Perform melody decomposition on the updated music audio data to obtain the melody features corresponding to the updated music audio data under different melody dimensions, and determine the core melody of the updated music audio data based on the feature value corresponding to each melody feature;

[0167] Step 25: Use natural language processing technology to perform text recognition on the updated music audio data to obtain the corresponding data text, use a preset sentiment dictionary to perform sentiment annotation on the data text to determine the sentiment tendency of the updated music audio data, and use the random forest method to train the sentiment tendency to obtain the data sentiment of the updated music audio data.

[0168] Step 26: Generate the beat period feature of the updated music audio data based on the beat information and the period range, and generate the melody emotion feature of the updated music audio data based on the core melody and the data emotion.

[0169] In this example, frequency components represent the components related to different frequencies contained in the updated music audio data;

[0170] In this example, the frequency count represents the number of times a frequency component occurs;

[0171] In this example, the dominant frequency component represents the frequency component that appears most frequently in the updated music audio data;

[0172] In this example, the phase information represents the waveform presented by the frequency components;

[0173] In this example, the frequency difference represents the phase shift between the frequency components and the main frequency component;

[0174] In this example, a single beat feature represents the beat presented by a frequency component;

[0175] In this example, the main beat feature represents the beat presented by the main frequency component;

[0176] In this example, the beat information represents the continuous beats presented in the updated music audio data;

[0177] In this example, the process of rhythm correction for the main beat feature is as follows: the process of correcting the notes of the main beat feature using a single beat feature;

[0178] In this example, the initial period represents the period derived from the principal phase information;

[0179] In this example, the period range represents the range of the period length of the initial period;

[0180] In this example, the melody dimension includes pitch dimension, volume dimension, speech rate dimension, duration dimension, and accent dimension;

[0181] In this example, the core melody represents the melody presented multiple times when the music audio data is updated;

[0182] In this example, sentiment labeling is represented as the process of matching data text with corresponding sentiments;

[0183] In this example, the preset sentiment dictionary contains several words, each of which is assigned a sentiment value. The sentiment tendency of a text is identified by statistically analyzing the sentiment values ​​contained in the text.

[0184] In this example, the data sentiment representation reflects the sentiment conveyed by the updated music audio data.

[0185] The working principle and beneficial effects of the above technical solution are as follows: In order to identify the various features of the updated music audio data, Fourier transform is used to identify the frequency components of the updated music audio data. By statistically analyzing the frequency count of each frequency component, the dominant frequency component of the updated music audio data is determined. Simultaneously, the frequency components are plotted in the frequency domain to determine the phase information of each frequency component. Then, based on the frequency difference between the frequency components and the dominant frequency component, the single beat feature and the dominant beat feature of the updated music audio data are determined, thereby determining the beat information of the updated music audio data. Next, the initial period of the updated music audio data is identified, and the initial period is adjusted using beat differences. Then, the core melody of the updated music audio data is identified by analyzing the melody of the updated music audio data. Furthermore, the data text of the updated music audio data is identified, and sentiment annotation is performed on the data text to train the data sentiment of the updated music audio data. Finally, based on the known information, the beat period feature and melody sentiment feature of the updated music audio data are generated. In this way, the updated music audio data can be analyzed in depth, and all the information of the updated music audio data is analyzed in detail, generating accurate beat period features and melody sentiment features, ensuring the accuracy of subsequent cluster analysis.

[0186] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A music data intelligent clustering system based on audio feature analysis, characterized in that, include: The data acquisition module is used to collect updated music audio data from music software. The feature extraction module is used to identify the beat cycle features and melody emotion features of the updated music audio data; The clustering processing module is used to identify the music category corresponding to the updated music audio data based on the beat cycle feature and the melody emotion feature, and to construct clustering adaptation parameters based on the category features corresponding to each music category; The clustering optimization module is used to determine the clustering similarity between the updated music audio data and each music category based on the clustering adaptation parameters, and to classify the updated music audio data into the corresponding target music category; The clustering processing module includes: The preliminary classification unit is used to match the updated music audio data with a corresponding first music category set based on the beat cycle feature, match the updated music audio data with a corresponding second music category set based on the melody emotion feature, and determine the initial music category corresponding to the updated music audio data based on the intersection of the first music category set and the second music category set. The clustering analysis unit is used to obtain the existing music audio data corresponding to each initial music category, construct a category attribute map corresponding to each initial music category based on the music tags corresponding to each existing music audio data, input the updated music audio data into the category attribute map for classification verification, obtain the matching nodes of each updated music audio data in each category attribute map, and filter music categories whose node distance between the matching node and the corresponding existing node is less than a specified distance. The parameter generation unit is used to generate category features corresponding to the music category based on the category attribute map, fit and match the beat cycle feature with the category feature to generate the beat cycle parameter of the updated music audio data in the corresponding music category, fit and match the melody emotion feature with the category feature to generate the melody emotion parameter of the updated music audio data in the corresponding music category, and fuse the beat cycle parameter and the melody emotion parameter to generate the clustering adaptation parameter of the updated music audio data in each music category. The clustering optimization module includes: The clustering verification unit is used to iteratively cluster the updated music audio data with the corresponding music category based on the clustering adaptation parameters, so as to obtain the data similarity features between the updated music audio data and the existing music audio data in each music category. The clustering optimization unit is used to statistically analyze several data similarity features corresponding to each music category to determine the clustering similarity between the updated music audio data and each music category, and to determine several target music categories corresponding to each updated music audio data based on the clustering similarity. The division execution unit is used to filter relevant music audio data with a cluster similarity higher than the standard similarity to the updated music audio data in each of the target music categories, set recommendation guidelines for the updated music audio data, and divide the updated music audio data into the corresponding target music categories.

2. The intelligent music data clustering system based on audio feature analysis as described in claim 1, characterized in that, The data acquisition module includes: A data capture unit is used to capture real-time, complete update data of the music software; The data filtering unit is used to perform duration analysis and type analysis on each of the real-time complete update data, and filter the updated music and audio data that meet the specified duration and type.

3. The intelligent music data clustering system based on audio feature analysis as described in claim 1, characterized in that, The feature extraction module includes: The data processing unit is used to perform Fourier transform on the updated music audio data to obtain several frequency components, count the frequency frequency corresponding to each frequency component, determine the main frequency component of the updated music audio data, and draw each frequency component in the frequency domain space to obtain the phase information corresponding to each frequency component. A beat determination unit is used to determine several single beat features of the updated music audio data based on the frequency difference between each frequency component and the main frequency component, determine the main beat feature of the updated music audio data based on the main frequency component, and perform beat correction on the main beat feature using each single beat feature to obtain the beat information of the updated music audio data. The period determination unit is used to generate an initial period based on the main phase information corresponding to the main frequency component, decompose the beat information using the initial period to obtain several decomposed beats, and adjust the period range of the initial period according to the beat difference between different decomposed beats. The melody determination unit is used to perform melody decomposition on the updated music audio data to obtain the melody features corresponding to the updated music audio data under different melody dimensions, and to determine the core melody of the updated music audio data based on the feature value corresponding to each melody feature. The sentiment determination unit is used to perform text recognition on the updated music audio data using natural language processing technology to obtain the corresponding data text, to perform sentiment annotation on the data text using a preset sentiment dictionary to determine the sentiment tendency of the updated music audio data, and to train the sentiment tendency using the random forest method to obtain the data sentiment of the updated music audio data. The feature determination unit is used to generate the beat period feature of the updated music audio data based on the beat information and the period range, and to generate the melody emotion feature of the updated music audio data based on the core melody and the data emotion.

4. The intelligent music data clustering system based on audio feature analysis as described in claim 3, characterized in that, The melody determination unit includes: The melody decomposition subunit is used to decompose the updated music audio data into melody to obtain the pitch features, volume features, speech rate features, duration features, and accent features of the updated music audio data. The numerical analysis subunit is used to analyze the numerical level of the corresponding feature value according to the numerical grading standard corresponding to each melody dimension, combine the numerical levels, and generate the combined melody information of the updated music audio data. The melody reconstruction subunit is used to construct pure audio based on the combined melody information, and to generate the core melody of the updated music audio data based on the pure melody of the pure audio.

5. The intelligent music data clustering system based on audio feature analysis as described in claim 1, characterized in that, The preliminary classification unit is also used for: When there is no intersection between the first music category set and the second music category set, it is determined that the updated music audio data is abnormal, and a re-acquisition instruction is generated. The reacquisition command is used to control the data acquisition module to reacquire the updated music audio data.

6. The intelligent music data clustering system based on audio feature analysis as described in claim 1, characterized in that, Also includes: After updating the music audio data to the corresponding target music category, the updated music audio data is regarded as existing music audio data.

7. A music data intelligent clustering method based on audio feature analysis, characterized in that, include: Step 1: Collect updated music audio data from the music software; Step 2: Identify the beat cycle characteristics and melodic emotional characteristics of the updated music audio data; Step 3: Identify the music category corresponding to the updated music audio data based on the beat cycle feature and the melody emotion feature, and construct clustering adaptation parameters based on the category features corresponding to each music category; Step 4: Determine the clustering similarity between the updated music audio data and each music category based on the clustering adaptation parameters, and classify the updated music audio data into the corresponding target music category; The working process of step 3 includes: The updated music audio data is matched with a first music category set based on the beat cycle feature, and a second music category set is matched with the updated music audio data based on the melody emotion feature. The initial music category corresponding to the updated music audio data is determined based on the intersection of the first music category set and the second music category set. Obtain the existing music audio data corresponding to each initial music category, construct a category attribute map corresponding to each initial music category based on the music tags corresponding to each existing music audio data, input the updated music audio data into the category attribute map for classification verification, obtain the matching nodes of each updated music audio data in each category attribute map, and filter music categories whose node distance between the matching node and the corresponding existing node is less than a specified distance. Based on the category attribute map, category features corresponding to the music category are generated. The beat cycle feature is fitted and matched with the category feature to generate the beat cycle parameter of the updated music audio data in the corresponding music category. The melody emotion feature is fitted and matched with the category feature to generate the melody emotion parameter of the updated music audio data in the corresponding music category. The beat cycle parameter and the melody emotion parameter are fused to generate the clustering adaptation parameter of the updated music audio data in each music category. The working process of step 4 includes: Based on the clustering adaptation parameters, the updated music audio data and the corresponding music categories are iteratively clustered to obtain the data similarity features between the updated music audio data and the existing music audio data in each music category. The clustering similarity between the updated music audio data and each music category is determined by statistically analyzing several data similarity features corresponding to each music category, and several target music categories corresponding to each updated music audio data are determined based on the clustering similarity. In each of the target music categories, relevant music audio data with cluster similarity higher than the standard similarity to the updated music audio data are selected, a recommendation guide is set for the updated music audio data, and the updated music audio data is divided into the corresponding target music categories.

8. The intelligent clustering method for music data based on audio feature analysis as described in claim 7, characterized in that, Step 2 includes: Step 21: Perform Fourier transform on the updated music audio data to obtain several frequency components, count the frequency frequency corresponding to each frequency component, determine the main frequency component of the updated music audio data, and draw each frequency component in the frequency domain space to obtain the phase information corresponding to each frequency component. Step 22: Determine several single beat features of the updated music audio data based on the frequency difference between each frequency component and the main frequency component, determine the main beat feature of the updated music audio data based on the main frequency component, and use each single beat feature to perform beat correction on the main beat feature to obtain the beat information of the updated music audio data. Step 23: Generate an initial period based on the principal phase information corresponding to the principal frequency component, decompose the beat information using the initial period to obtain several decomposed beats, and adjust the period range of the initial period according to the beat differences between different decomposed beats. Step 24: Perform melody decomposition on the updated music audio data to obtain the melody features corresponding to the updated music audio data under different melody dimensions, and determine the core melody of the updated music audio data based on the feature value corresponding to each melody feature; Step 25: Use natural language processing technology to perform text recognition on the updated music audio data to obtain the corresponding data text, use a preset sentiment dictionary to perform sentiment annotation on the data text to determine the sentiment tendency of the updated music audio data, and use the random forest method to train the sentiment tendency to obtain the data sentiment of the updated music audio data. Step 26: Generate the beat period feature of the updated music audio data based on the beat information and the period range, and generate the melody emotion feature of the updated music audio data based on the core melody and the data emotion.

Images