CN110147463A - A music pushing method, system, device and storage medium - Google Patents

Abstract

The invention discloses a music push method, device and storage medium. The method includes acquiring first music playing history information and first age information from a first music player, acquiring second music playing history information from a second music player, and calculating the first music player and the second music playing history information. interest similarity between players, calculate the age similarity between the first music player and the second music player, calculate user similarity, use an unsupervised clustering algorithm to score and play the second music The part with the highest score in each piece of music corresponding to the historical information is pushed to the first music player and other steps. The present invention screens out the second music player by calculating interest similarity and age similarity, and uses an unsupervised clustering algorithm to score each piece of music, which can achieve a more accurate push effect and improve user experience. The invention is widely used in the technical field of Internet.

Description

A music pushing method, system, device and storage medium

technical field

The invention relates to the technical field of the Internet, in particular to a music push method, device and storage medium.

Background technique

When playing music, users of music playing software hope to be pushed to suitable music in addition to selecting the music in the playlist set by themselves. The existing music push technology is based on the analysis of the user's playback history, and analyzes the user's preferences by counting the genres and singers of the songs that the user often plays, so as to push the user similar preferences. song. The main disadvantages of the existing music push technology are as follows: First, when the user often plays a certain type of song, the same or similar type of song will be pushed, and only the user actively plays other types of songs. Only then can it be pushed to other types of songs, but this has violated the meaning of music push; second, the existing music push technology is completely based on the analysis of the user's play history, but the user's play history is the user's play history. The playback records of a certain period of time in the past, and the emotions of the same user at different time points in a period of time are slightly different, which leads to the user's preference for music and the type of music played at different time points. Therefore, the analysis results of the existing music push technology are actually the "average value" of users' preferences in the past period of time, but people's emotions are always in a state of fluctuation, and users' preferences and needs for music are also changing. Therefore, the existing music push technology cannot accurately promote the current user's preferences and needs for music. To sum up, the existing music push technology brings poor user experience.

Contents of the invention

In order to solve the above technical problems, the object of the present invention is to provide a music pushing method, device and storage medium.

On the one hand, the embodiment of the present invention includes a music push method, comprising the following steps:

Obtain the first music playing history information and the first age information from the first music player; The first music playing history information is the record of the music played by the first music player under the first weather condition; the first The age information is the age of the user of the first music player;

Obtain the second music playing history information from the second music player; The second music playing history information is the record of the music played by the second music player under the first weather conditions; The second age information is the second the age of the user of the music player;

Calculate the interest similarity between the first music player and the second music player according to the first music playing history information and the second music playing history information;

calculating the age similarity between the first music player and the second music player according to the first age information and the second age information;

Calculate the user similarity according to the interest similarity and age similarity; the user similarity is a weighted average of the interest similarity and age similarity;

When it is detected that the user similarity meets a preset threshold condition, an unsupervised clustering algorithm is used to score each piece of music corresponding to the second music playback history information;

Pushing the part with the highest score among the pieces of music corresponding to the second music playing history information to the first music player.

Further, the step of calculating the interest similarity between the first music player and the second music player according to the first music playing history information and the second music playing history information specifically includes:

Using an unsupervised clustering algorithm to classify each piece of music corresponding to the first music playback history information and the second music playback history information; the classification is used to classify each piece of music into a corresponding category;

Generate a first interest vector; the elements in the first interest vector are respectively the ratio of the number of music classified as each category in the first music playback history information to the total number of music corresponding to the first music playback history information;

Generate a second interest vector; the elements in the second interest vector are respectively the ratio of the number of music classified as each category in the second music playback history information to the total number of music corresponding to the second music playback history information;

The interest similarity is calculated according to the first interest vector, the second interest vector and the total number of categories.

Further, the formula used in the step of calculating the interest similarity according to the first interest vector, the second interest vector and the total number of categories is:

In the formula, A represents the first music player, X represents the second music player, Sim _IC (A, X) is the interest similarity between the first music player and the second music player, v _A is the first Interest vector, v _X is the second interest vector, distance(v _A , v _X ) is the distance between the first interest vector and the second interest vector, and K is the total number of categories.

Further, the formula used in the step of calculating the age similarity between the first music player and the second music player according to the first age information and the second age information is:

In the formula, A represents the first music player, X represents the second music player, Sim _SC (A, X) is the age similarity between the first music player and the second music player, α _A is the first Age information, α _X is the second age information, and α is a preset threshold.

Further, the step of using an unsupervised clustering algorithm to score each piece of music corresponding to the second music playback history information specifically includes:

Analyzing each piece of music corresponding to the second music playback history information, thereby obtaining a feature vector of each piece of music;

Using an unsupervised clustering algorithm to calculate a central feature vector according to the feature vectors of each piece of music, and classify the feature vectors of each piece of music into the corresponding central feature vector;

Euclidean distances between the eigenvectors of the respective pieces of music and the corresponding center eigenvectors are calculated respectively; the Euclidean distances are used as the score of the corresponding music.

Further, the step of using an unsupervised clustering algorithm to score each piece of music corresponding to the second music playback history information specifically includes:

Analyzing each piece of music corresponding to the second music playback history information, thereby obtaining a feature vector of each piece of music;

Using an unsupervised clustering algorithm to calculate a central feature vector according to the feature vectors of each piece of music, and classify the feature vectors of each piece of music into the corresponding central feature vector;

Calculate the Euclidean distance between the eigenvectors of each piece of music and the corresponding center eigenvectors respectively;

Statistically count the accumulative times played of each piece of music;

Calculate the quotient of the Euclidean distance corresponding to each piece of music and the cumulative number of played times as the score of the corresponding music.

Further, the elements of the feature vector include zero crossing rate, unit power, low energy rate, beat, spectrum center and tone type.

On the other hand, the embodiment of the present invention includes a music push system, comprising:

The first information acquisition module is used to obtain the first music playing history information and the first age information from the first music player; the first music playing history information is played by the first music player under the first weather conditions Music records; the first age information is the age of the user of the first music player;

The second information acquisition module is used to obtain the second music playing history information from the second music player; the second music playing history information is a record of the music played by the second music player under the first weather condition; the The second age information is the age of the user of the second music player;

An interest similarity calculation module, configured to calculate the interest similarity between the first music player and the second music player according to the first music playback history information and the second music playback history information;

an age similarity calculation module, configured to calculate the age similarity between the first music player and the second music player according to the first age information and the second age information;

The user similarity calculation module is used to calculate the user similarity according to the interest similarity and age similarity; the user similarity is a weighted average of the interest similarity and age similarity;

A scoring module, configured to use an unsupervised clustering algorithm to score each piece of music corresponding to the second music playback history information when it is detected that the user similarity meets a preset threshold condition;

A push module, configured to push the part with the highest score among the pieces of music corresponding to the second music playback history information to the first music player.

On the other hand, the embodiment of the present invention also includes a music push device, including a memory and a processor, the memory is used to store at least one program, and the processor is used to load the at least one program to execute the music push method of the present invention .

On the other hand, the embodiment of the present invention further includes a storage medium, which stores processor-executable instructions, and the processor-executable instructions are used to execute the music push method of the present invention when executed by the processor.

The beneficial effects of the present invention are: by calculating the interest similarity and age similarity between the first music player and the second music player, the first music player can be screened out for the first music player with the same music preference and music preference under the same weather conditions. The second music player at the same age stage; by using an unsupervised clustering algorithm to score each piece of music, each piece of music corresponding to the second music playing history information in the screened out second music player can be further screened ; By pushing the further filtered music to the first music player, a push effect with higher accuracy can be obtained and user experience can be improved.

Description of drawings

Fig. 1 is a flowchart of an embodiment of the music push method of the present invention.

Detailed ways

Example 1

The present embodiment comprises a kind of music pushing method, with reference to Fig. 1, described method comprises the following steps:

S1. Obtain the first music playing history information and the first age information from the first music player; the first music playing history information is a record of music played by the first music player under the first weather conditions; the The first age information is the age of the user of the first music player;

S2. Obtain the second music playback history information from the second music player; the second music playback history information is a record of the music played by the second music player under the first weather conditions; the second age information is the age of the user of the second music player;

S3. Calculate the interest similarity between the first music player and the second music player according to the first music playing history information and the second music playing history information;

S4. Calculate the age similarity between the first music player and the second music player according to the first age information and the second age information;

S5. Calculate the user similarity according to the interest similarity and age similarity; the user similarity is a weighted average of the interest similarity and age similarity;

S6. When it is detected that the user similarity meets the preset threshold condition, use an unsupervised clustering algorithm to score each piece of music corresponding to the second music playback history information;

S7. Push the part with the highest score among the pieces of music corresponding to the second music playback history information to the first music player.

In steps S1 and S2, the music player refers to a device such as a desktop computer, a notebook computer, a tablet computer or a mobile phone that can run music playing software, read music data and decode it, and drive a speaker to play music; , the music player may refer to not only the device hardware, but also the program software running on the device hardware, or the whole composed of the device hardware and the program software running on it.

In steps S1 and S2, the user of the music player refers to the person who operates the music player. The user can log in to the music player software run by the music player through the account number and password, so as to control the music player. In this embodiment, there is a one-to-one correspondence between music players and their users, that is, one music player is only used by one user, and one user only uses one music player. users to distinguish.

In this embodiment, the first music player is the target of the music to be pushed, and the second music player is the source of the pushed music, that is, the purpose of steps S1-S7 in this embodiment is to push the music that meets the requirements. The music played by the second music player is pushed to the first music player. Said playing may refer to running music playing software to drive the speaker to emit sound, or it may refer to only downloading music data files from the Internet to the music player.

Steps S1-S7 in this embodiment are executed by a server program running on the server. Both the first music player and the second music player are connected to the Internet, so that the music player software running on the first music player and the second music player can access the Internet and communicate with the server program on the server. communication. The first music player and the second music player open access rights to the music player software, so that the music player program can read any data stored on the first music player and the second music player and upload it to the server program.

In this embodiment, the first music player and the second music player also obtain the real-time weather information of the user's location through the weather forecast module built in the operating system or the installed third-party weather forecast software.

In this embodiment, when a trigger condition is detected, step S1 starts to be executed. The trigger condition may refer to a specific operation performed by the user of the first music player, for example, the user of the first music player sends a request to obtain a music push through the music player software running on the first music player, or The user of the first music player starts the music player software running on the first music player, or the weather forecast software updates the current weather conditions in real time (such as switching from sunny to rainy) and the like.

In this embodiment, the music playing software running on the music player will record the title of the played music, the start and end time of playing, the cumulative number of times of playing, and the weather conditions at that time every time the music is played, thereby forming a music playing software. history information. The music playing history information can be stored and displayed in the form of list or table. The weather conditions are obtained by the music player software from the weather forecast software.

In step S1, the first music playing history information refers to records of music played by the first music player under the first weather condition. The first weather condition can be set by the user, such as rainy day, sunny day or cloudy day; the first weather condition can also be obtained from real-time weather condition information from weather forecast software.

The first age information is a set of data, which stores the age of the user of the first music player. The first age information can be registered when the user registers the account of the music player software. When playing the software, the music playing software can read the first age information.

The second music player in step S2 may be a collective concept, that is, the second music player refers to one or more music players.

In step S3, by analyzing the songs contained in the first music playing history information and the second music playing history information, the interest similarity _SimIC ( A, X). The higher the similarity of interest, the more similar the music played by the first music player and the second music player is, that is, the music preference between the user of the first music player and the user of the second music player. The higher the degree of similarity. Since the first music playback history information and the second music playback history information are records of music played under the same weather conditions, the interest similarity calculated in step S3 takes into account the impact of weather conditions on user emotions. The impact on the user's music preference can make music push more accurate.

The age similarity Sim _SC (A,X) calculated in step S4 reflects the degree of age difference between the users of the first music player and the users of the second music player. The higher the age similarity, the smaller the age difference between the users of the first music player and the users of the second music player. Since the user's preference for music is related to his age, the higher the age similarity, the higher the similarity in music preference between the user of the first music player and the user of the second music player.

The user similarity calculated in step S5 is the weighted average of interest similarity and age similarity, that is Sim(A,X)=W _IC Sim _IC (A,X)+W _SC Sim _SC (A,X), Among them, the weights W _IC and W _SC can be customized. The user similarity comprehensively considers the influence of interest similarity and age similarity, that is, comprehensively considers the user's preference for music reflected by song playing history, weather conditions and age.

In step S6, the user similarity Sim(A,X) between the first music player and the second music player is compared with a preset threshold λ, and when Sim(A,X)>λ, the first The second music player is set as a neighbor of the first music player. The neighbor refers to an online connection relationship. In step S7, the music that meets the conditions in the second music playing history information of each neighbor of the first music player is pushed to the first music player. For example, if there are N second music players satisfying the condition of Sim(A,X)>λ, these N second music players are all regarded as neighbors of the first music player, and the N second music players are played The qualified music in the second music playing history information of the player is pushed to the first music player.

In step S6, an unsupervised clustering algorithm is also used to score each piece of music corresponding to the second music playing history information, and then the one or more songs with the highest score are selected, and pushed to the first music playing in step S7 device. In this embodiment, the number of music pushed in step S7 can be customized. For example, when the number of music pushed is set to 3, step S7 will push the 3 songs with the highest scores calculated in step S6. .

Pushing in step S7 refers to sending information such as the song title, singer or serial number of the song to the first music player, or sending the data file of the song to the first music player.

By performing steps S1-S5, the second music player with the same music preference and the same age group under the same weather conditions can be screened out for the first music player; by performing step S6, the screened out second music player can Each piece of music corresponding to the second music playback history information in the player is further screened; by performing step S7, the further screened music is pushed to the first music player, which can achieve a more accurate push effect and improve user experience .

Further as a preferred embodiment, the step S3 is to calculate the interest similarity between the first music player and the second music player according to the first music playback history information and the second music playback history information. One step, including:

S301. Use an unsupervised clustering algorithm to classify the pieces of music corresponding to the first music playback history information and the second music playback history information; the classification is used to classify each piece of music into a corresponding category;

S302. Generate a first interest vector; the elements in the first interest vector are the ratio of the number of music classified into each category in the first music playback history information to the total number of music corresponding to the first music playback history information;

S303. Generate a second interest vector; the elements in the second interest vector are the ratio of the number of music classified into each category in the second music playback history information to the total number of music corresponding to the second music playback history information;

S304. Calculate the interest similarity according to the first interest vector, the second interest vector and the total number of categories.

In this embodiment, step S3 is specifically composed of steps S301-S304.

In step S301, use an unsupervised clustering algorithm to classify the pieces of music corresponding to the first music playing history information and the second music playing history information. By setting the unsupervised clustering algorithm, you can set the total number of categories to be divided into. The unsupervised clustering algorithm used in the present embodiment is the K-means algorithm, by setting the K-means algorithm, music can be classified into one of category 1, category 2, category 3...category K, namely At this point the total number of categories is K.

Through step S301, each piece of music corresponding to the first music playback history information and the second music playback history information is classified into one of category 1, category 2, category 3...category K, that is, each piece of music is marked A label corresponding to one of category 1, category 2, category 3...category K is attached.

In step S302, a first interest vector v _A is generated. The first interest vector v _A is composed of a plurality of elements, and each element is the ratio of the number of music classified into various categories in the first music playback history information to the total number of music corresponding to the first music playback history information, namely Wherein, s1 _{category K} is the number of music classified as category K in the first music playback history information, and total1 is the total number of music corresponding to the first music playback history information.

In step S303, a second interest vector v _X is generated. The second interest vector v _X is composed of a plurality of elements, and each element is respectively the ratio of the number of music classified into various categories in the second music playback history information to the total number of music corresponding to the second music playback history information, namely Wherein, s2 _{category K} is the number of music classified as category K in the second music playback history information, and total2 is the total number of music corresponding to the first music playback history information.

The formula used in step S304 is:

In the formula, A represents the first music player, X represents the second music player, Sim _IC (A, X) is the interest similarity between the first music player and the second music player, v _A is the first Interest vector, v _X is the second interest vector, distance(v _A , v _X ) is the distance between the first interest vector and the second interest vector, and K is the total number of categories.

As a further preferred embodiment, the formula used in the step of calculating the age similarity between the first music player and the second music player according to the first age information and the second age information is:

In the formula, A represents the first music player, X represents the second music player, Sim _SC (A, X) is the age similarity between the first music player and the second music player, α _A is the first The age information is the age of the user of the first music player, α _X is the second age information, that is the age of the user of the second music player, and α is a preset threshold.

Further as a preferred embodiment, the step S6, that is, the step of scoring each piece of music corresponding to the second music playback history information using an unsupervised clustering algorithm, specifically includes:

S601A. Analyze each piece of music corresponding to the second music playback history information, so as to obtain the feature vector of each piece of music;

S602A. Using an unsupervised clustering algorithm to calculate a central feature vector according to the feature vectors of each piece of music, and classify the feature vectors of each piece of music into the corresponding central feature vector;

S603A. Calculate the Euclidean distance between the eigenvectors of each piece of music and the corresponding central eigenvector; the Euclidean distance is used as the score of the corresponding music.

Steps S601A-S603A are the first specific implementation method of step S6.

In step S601A, analyze the music array of each piece of music corresponding to the second music playback history information, and obtain a 6-dimensional data set composed of parameters such as zero-crossing rate, unit power, low energy rate, tempo, spectrum center, and tone type as elements. Feature vector.

The music array refers to the data format converted after the sound wave file of the music is loaded, and each element in the music array represents the sound amplitude value of the music at a certain moment, because a sampling operation may be performed during the loading process of the sound wave file , the subscript of each element in the music array is not equal to the actual moment of the audio. The music array is a floating-point time series. In this embodiment, each element of the floating-point value in the music array is of type float32, and the storage occupies 4Byte and a total of 32 bits.

The formula for calculating the zero-crossing rate is In the formula, x[n] is the nth element in the music array x, and N is the length of the music array.

The formula for calculating the unit power is

The formula for calculating the low energy rate is In the formula, M is the number of elements x[k] in the music array x[n] that satisfy the condition of x[k] ² < unit power.

The calculation process of the beat includes:

(1) Use the following formula to take the absolute value of each element of the music array and perform downsampling with a sampling rate of 32;

y=downsample(|x|,32)

(2) For the downsampled array, calculate its autocorrelation and limit its minimum value to 0:

(3) Then, based on the above calculation results, take a partition of about 4 seconds, where fs is the original linear interval, and the element subscript interval:

(4) Use the peak selection method for the data, with the minimum peak value of 0.5 and the minimum peak distance as a constraint. Converts the current scale to a beats-per-minute scale using a scaling factor, with the first peak of the beats-per-minute scale being the requested beat.

The calculation process of the spectrum center includes:

(1) Use the following formula to transform the music array from the time domain to the frequency domain, perform Fourier transform on the music, and obtain the amplitude spectrum:

Among them, in the formula, x _n is the nth element in the music array x, and N is the length of the music array. N=2 ^t , t∈Z, 2 ^t-1 <x array length ≤2 ^t ; calculate the center of the unilateral amplitude spectrum by taking the absolute value of the first half of the amplitude spectrum:

(2) Update linear intervals to frequency intervals:

(3) Calculate the spectral center using the following formula:

The calculation process of the tone type includes:

(1) Use the following formula to cut the frequency range below 20Hz in the music array, and map the frequency dimension of all the remaining ranges to the MIDI number dimension:

Among them, 69 is the MIDI number of A on middle C (440 Hz);

(2) Map MIDI numbers to MIDI pitch categories (MPC) by taking modulo 12:

MPC _k = mod(M _k ,12)

(3) Superimpose MPCs of the same category to obtain a timbre map; correspondingly, corresponding to the three tone types of Major, Minor and None:

major[k]=chroma[k]+chroma[(k+2)%12]+chroma[(k+4)%12]+chroma[(k+5)%12]+chroma[(k+7) %12]+chroma[(k+9)%12]+chroma[(k+11)%12]

minor[k]=chroma[k]+chroma[(k+2)%12]+chroma[(k+3)%12]+chroma[(k+5)%12]+chroma[(k+7) %12]+chroma[(k+8)%12]+chroma[(k+10)%12]

Wherein, chroma is a timbre map, which satisfies major[(k+3)%12]=minor[k]. When chroma[(k+3)%12]>chroma[k], the tone type is Major; when chroma[(k+3)%12]<chroma[k], the tone type is Minor; otherwise, the tone type is None.

In step S602A, execute an unsupervised clustering algorithm, select a cluster center from each piece of music corresponding to the second music playback history information and perform clustering, that is, calculate the center feature vector according to the feature vector of each piece of music, And classify the eigenvectors of each piece of music into the corresponding central eigenvectors. By executing step S602A, a total of K cluster centers are selected from the pieces of music corresponding to the second music playing history information, wherein the feature vector corresponding to the kth cluster center, that is, the center feature vector is f _kc . The rest of the music not selected as the cluster center is classified into each cluster center, and the feature vector of the i-th music classified into the k-th cluster center is f _ki .

In step S603A, the Euclidean distances between the feature vectors of the respective pieces of music and the corresponding center feature vectors are respectively calculated; the Euclidean distances are used as the scores of the corresponding music. That is, in step S603A, the score of the i-th piece of music classified into the k-th cluster center is: S _ki =S _k ·distance(f _kc ,f _ki ), k∈[1,K], where distance(f _kc ,f _ki ) is the distance between the center feature vector f _kc corresponding to the kth cluster center and the feature vector _fki of the i-th music classified to the kth cluster center, and S _k is a custom coefficient .

Further as a preferred embodiment, the step S6, that is, the step of scoring each piece of music corresponding to the second music playback history information using an unsupervised clustering algorithm, specifically includes:

S601B. Analyze each piece of music corresponding to the second music playing history information, so as to obtain the feature vector of each piece of music;

S602B. Using an unsupervised clustering algorithm to calculate the central feature vector according to the feature vectors of each piece of music, and classify the feature vectors of each piece of music into the corresponding central feature vector;

S603B respectively calculates the Euclidean distance between the feature vectors of each piece of music and the corresponding central feature vector;

S604B respectively counts the accumulative times played by each piece of music;

S605B. Calculate the quotient of the Euclidean distance corresponding to each piece of music and the cumulative number of played times as the score of the corresponding music.

Steps S601B-S605B are the second specific implementation method of step S6.

The effect of steps S601B-S603B is the same as that of steps S601A-S603A, that is, the center feature vector f _kc corresponding to the k-th cluster center and the feature vector of the i-th music classified to the k-th cluster center are calculated The distance between f _ki is distance(f _kc ,f _ki ).

In the second specific implementation method of step S6, step S604B is also performed, that is, counting the cumulative number of times each piece of music corresponding to the second music playing history information is played. By executing step S604B, the accumulative number of times N _ki of the i-th music classified into the k-th cluster center can be obtained.

In the second specific implementation method of step S6, step S605B is also performed, that is, calculating the quotient of the Euclidean distance corresponding to each piece of music and the cumulative number of times played, and the quotient calculated by step S605B is the score of the corresponding music. That is, in step S605B, the score of the i-th piece of music classified into the k-th cluster center is: k∈[1,K].

By executing the second specific implementation method of step S6, that is, steps S601B-S605B, on the basis of considering the classification results of the unsupervised clustering algorithm, the influence of the cumulative number of times each song is played is also considered, that is, the cumulative number of times the song is played The smaller the number of plays, the higher the score of the song, and the greater the possibility of being pushed to the first music player when other conditions remain unchanged, so as to avoid that some songs in the second music playing history information are repeatedly pushed, thereby Guarantee the novelty and coverage of the push results.

In this embodiment, after steps S301-S304 are performed, each piece of music corresponding to the second music playing history information is classified into one of category 1, category 2, category 3...category K. After step S6 is executed, each piece of music corresponding to the second music playback history information has a corresponding score. When step S7 is executed, the music corresponding to the second music playback history information can be sorted internally in the same category, and then according to the setting, the music classified into category 1, category 2, category 3...category K The one or more songs with the highest score in the category are pushed to the first music player.

Example 2

A kind of music pushing system in the present embodiment, comprises:

The first information acquisition module is used to obtain the first music playing history information and the first age information from the first music player; the first music playing history information is played by the first music player under the first weather conditions Music records; the first age information is the age of the user of the first music player;

The second information acquisition module is used to obtain the second music playing history information from the second music player; the second music playing history information is a record of the music played by the second music player under the first weather condition; the The second age information is the age of the user of the second music player;

An interest similarity calculation module, configured to calculate the interest similarity between the first music player and the second music player according to the first music playback history information and the second music playback history information;

an age similarity calculation module, configured to calculate the age similarity between the first music player and the second music player according to the first age information and the second age information;

The user similarity calculation module is used to calculate the user similarity according to the interest similarity and age similarity; the user similarity is a weighted average of the interest similarity and age similarity;

A scoring module, configured to use an unsupervised clustering algorithm to score each piece of music corresponding to the second music playback history information when it is detected that the user similarity meets a preset threshold condition;

A push module, configured to push the part with the highest score among the pieces of music corresponding to the second music playback history information to the first music player.

In this embodiment, the first information acquisition module, the second information acquisition module, the interest similarity calculation module, the age similarity calculation module, the user similarity calculation module, the scoring module and the push module can be connected to the Internet server with corresponding Functional software or hardware modules.

This embodiment also includes a music push device, including a memory and a processor, the memory is used to store at least one program, and the processor is used to load the at least one program to execute the method described in Embodiment 1.

This embodiment also includes a storage medium, which stores processor-executable instructions, and the processor-executable instructions are used to execute the method described in Embodiment 1 when executed by a processor.

The music push system, device, and storage medium in this embodiment can execute the music push method of the present invention, can execute any combination of implementation steps of the method embodiments, and have the corresponding functions and beneficial effects of the method.

The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the described embodiments, and those skilled in the art can also make various equivalent deformations or replacements without violating the spirit of the present invention. , these equivalent modifications or replacements are all within the scope defined by the claims of the present application.

Claims (10)

1. a kind of music method for pushing, which comprises the following steps:

First music play history information and the first age information are obtained from first music player；The first music broadcasting is gone through History information is the record for the music that first music player played under the first weather conditions；First age information is the The age of the user of one music player；

The second music historical information is obtained from the second music player；The second music historical information is the second sound The record for the music that happy player played under the first weather conditions；Second age information is the second music player The age of user；

The first music player is calculated according to the first music play history information and the second music historical information With the Interest Similarity between the second music player；

The first music player and the second music player are calculated according to first age information and the second age information Between age similarity；

According to the Interest Similarity and age similarity calculation user's similarity；User's similarity is that the interest is similar The weighted average of degree and age similarity；

When detecting that user's similarity meets preset threshold condition, using Unsupervised clustering algorithm to second music The corresponding each song of play history information scores；

The part of highest scoring in the corresponding each song of the second music historical information is pushed to first music to broadcast Put device.

2. a kind of music method for pushing according to claim 1, which is characterized in that described to be played according to the first music Historical information and the second music historical information calculate emerging between the first music player and the second music player The step for interesting similarity, specifically includes:

Using Unsupervised clustering algorithm respectively to the first music play history information and the second music historical information pair Each song answered is classified；The classification for being classified as corresponding classification for each song respectively；

Generate the first interest vector；Element in first interest vector is respectively to be divided in first music play history information Class is the ratio that music number of all categories accounts for the corresponding music sum of first music play history information；

Generate the second interest vector；Element in second interest vector is respectively to be divided in the second music historical information Class is the ratio that music number of all categories accounts for the corresponding music sum of the second music historical information；

The Interest Similarity is calculated according to first interest vector, the second interest vector and classification sum.

3. a kind of music method for pushing according to claim 2, which is characterized in that it is described according to first interest to The step for amount, the second interest vector and classification sum calculate Interest Similarity formula used are as follows:

In formula, A indicates that first music player, X indicate the second music player, Sim_IC(A, X) be first music player with Interest Similarity between second music player, v_AFor the first interest vector, v_XFor the second interest vector, distance (v_A, v_X) it is the distance between the first interest vector and the second interest vector, K is classification sum.

4. a kind of music method for pushing according to claim 1, which is characterized in that described according to first age information And the second age information the step for calculating the age similarity between the first music player and the second music player Formula used are as follows:

In formula, A indicates that first music player, X indicate the second music player, Sim_SC(A, X) be first music player with Age similarity between second music player, α_AFor the first age information, α_XFor the second age information, α is preset threshold Value.

5. a kind of music method for pushing according to claim 1, which is characterized in that described to use Unsupervised clustering algorithm pair The step for corresponding each song of the second music historical information is scored, specifically includes:

The corresponding each song of second music historical information is parsed, thus obtain the feature of each song to Amount；

Central feature vector is calculated according to the feature vector of each song using Unsupervised clustering algorithm, and by each first sound Happy feature vector is sorted out to corresponding central feature vector；

Calculate separately the Euclidean distance between the feature vector of each song and corresponding central feature vector；The Euclidean away from From the score as corresponding music.

6. a kind of music method for pushing according to claim 1, which is characterized in that described to use Unsupervised clustering algorithm pair The step for corresponding each song of the second music historical information is scored, specifically includes:

The corresponding each song of second music historical information is parsed, thus obtain the feature of each song to Amount；

Central feature vector is calculated according to the feature vector of each song using Unsupervised clustering algorithm, and by each first sound Happy feature vector is sorted out to corresponding central feature vector；

Calculate separately the Euclidean distance between the feature vector of each song and corresponding central feature vector；

The accumulative played number of each song is counted respectively；

Calculate score of the corresponding Euclidean distance of each song to the quotient of accumulative played number as corresponding music.

7. a kind of music method for pushing according to claim 5 or 6, which is characterized in that the element packet of described eigenvector Include zero-crossing rate, unit power, low energy rate, bat, spectral centroid and tone type.

8. a kind of music supplying system characterized by comprising

The first information obtains module, for obtaining first music play history information and the first age letter from first music player Breath；The first music play history information is the note for the music that first music player played under the first weather conditions Record；First age information is the age of the user of first music player；

Second data obtaining module, for obtaining the second music historical information from the second music player；Second sound Happy play history information is the record for the music that the second music player played under the first weather conditions；Second age Information is the age of the user of the second music player；

Interest Similarity computing module, for according to the first music play history information and the second music historical information Calculate the Interest Similarity between the first music player and the second music player；

Age similarity calculation module, for calculating the first music according to first age information and the second age information Age similarity between player and the second music player；

User's similarity calculation module, for according to the Interest Similarity and age similarity calculation user's similarity；It is described User's similarity is the weighted average of the Interest Similarity and age similarity；

Grading module, for using Unsupervised clustering algorithm when detecting that user's similarity meets preset threshold condition It scores the corresponding each song of the second music historical information；

Pushing module, for pushing the part of highest scoring in the corresponding each song of the second music historical information To first music player.

9. a kind of music driving means, which is characterized in that including memory and processor, the memory is for storing at least one A program, the processor require any one of 1-7 the method for loading at least one described program with perform claim.

10. a kind of storage medium, wherein being stored with the executable instruction of processor, which is characterized in that the processor is executable Instruction be used to execute such as any one of claim 1-7 the method when executed by the processor.