CN109726310A - A kind of determination method, apparatus and storage medium for recommending music track - Google Patents
A kind of determination method, apparatus and storage medium for recommending music track Download PDFInfo
- Publication number
- CN109726310A CN109726310A CN201811362181.7A CN201811362181A CN109726310A CN 109726310 A CN109726310 A CN 109726310A CN 201811362181 A CN201811362181 A CN 201811362181A CN 109726310 A CN109726310 A CN 109726310A
- Authority
- CN
- China
- Prior art keywords
- audio
- music
- music track
- information
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a kind of determination method, apparatus and storage medium for recommending music track, wherein the described method includes: obtaining audio file to be identified;According to preset audio feature extraction model, the audio feature information including at least audio generic of audio file to be identified is extracted;Search and the matched audio categories information of audio feature information in audio classification index database;According to audio categories information, the music track with audio categories information matches is searched in music assorting index database, and the music track matched is determined as to the recommendation music track for being used to recommend.The solution of the present invention improves the precision for recommending music track, improves user experience.
Description
Technical field
The present invention relates to field of computer technology, and in particular to it is a kind of recommend music track determination method, apparatus and deposit
Storage media.
Background technique
It is widely available with intelligent terminal, for example, smart phone, tablet computer have become it is required in people's life
Product.Moreover, the continuous maturation of domestic enterprise's technology with production of intelligent mobile phone, present smart phone is not only quality-high and inexpensive,
The function of smart phone is also stronger and stronger, is mainly used for making a phone call and send short messages from the past, also possesses broadcasting music till now
Function, play the function of video, payment function, camera function, the function of reading electronic book, function of surfing the Net.
In above-mentioned function, for music-lover, often to use this using smart phone as music player
Sample can not only appreciate interesting to listen to music track, additionally it is possible to focus on, avoid being interrupted sb's train of tought by the external world.
Have the function of music player based on smart phone, if active user hears a first good song at random, but works as
When the name of the song is not recorded, only with the tune of the song, the bitty information such as rhythm is to be difficult to restore
And the song is accurately found in library.
In addition, it is based on present expanding economy, either popular song or opera or rock music, hip-hop music,
How the increasingly diversification of present music finds the music track that user likes in the numerous music libraries of music track, will
It is the process taken time and effort, and the music track that user filters out often has randomness, the precision filtered out is not high, often
And the preference of user is not met.
The businessman of existing music portal, although also oriented target user recommends the service of music track, the recommendation
Service often has randomness, for example, according to the setting requirements of target user, if the user setting song of a certain singer,
The new album or history album of the singer are periodically pushed to target user, and there is randomness.In another example according to target user
Setting requirements, recommend rock music, then periodically to target user push newly go out rock music new album or classics go through
History album equally has randomness;In this way, the Experience Degree of user is very low, if what the service was also paid for, user can feel to prop up
Payment is unworthy very much.
The existing precision for recommending music track to user how is improved, the Experience Degree of user is improved, is skill to be solved
Art problem.
Summary of the invention
The embodiment of the present invention is designed to provide a kind of determination method, apparatus and storage medium for recommending music track,
To solve existing not high, the low problem of user experience to the precision of user's recommendation music track.
To achieve the above object, the embodiment of the present invention provides a kind of determination method for recommending music track, the method packet
It includes: obtaining audio file to be identified;According to preset audio feature extraction model, the audio file to be identified is extracted at least
Audio feature information including audio generic;It is searched in audio classification index database matched with the audio feature information
Audio categories information;According to the audio categories information, search and the audio categories information in music assorting index database
The music track matched, and the music track matched is determined as to the recommendation music track for being used to recommend.
Optionally, the method also includes: according to the preference information of user, the recommendation music track is pushed to pair
On the mobile terminal of the user answered.
Optionally, the method also includes: by convolutional neural networks model to the preset audio feature extraction mould
Type is constructed.
Optionally, the convolutional neural networks model is constructed using Keras framework.
Match with the above method, another aspect of the present invention provides a kind of determining device for recommending music track, the dress
Setting includes: acquiring unit, obtains audio file to be identified;Extraction unit extracts institute according to preset audio feature extraction model
State the audio feature information including at least audio generic of audio file to be identified;Matching unit is indexed in audio classification
The matched audio categories information of the audio feature information extracted with the extraction unit is searched in library;And according to described
Audio categories information searches for the music track with the audio categories information matches in music assorting index database;Determination unit,
The music track that the matching unit is matched is determined as the recommendation music track for being used to recommend.
Optionally, described device further includes push unit, the push unit, according to the preference information of user, by institute
The recommendation music track that determination unit is determined is stated to push on the mobile terminal of corresponding user.
Optionally, described device further includes model construction unit, and the model construction unit passes through convolutional neural networks mould
Type constructs the preset audio feature extraction model.
Optionally, the convolutional neural networks model is constructed using Keras framework.
Matching with above-mentioned apparatus, further aspect of the present invention provides one kind and recommends music destination device for determining,
It is characterized in that, includes memory and more than one program one of them or more than one a program storage
In memory, and be configured to execute the one or more programs by one or more than one processor include
Instruction for performing the following operation: audio file to be identified is obtained;According to preset audio feature extraction model, described in extraction
The audio feature information including at least audio generic of audio file to be identified;Search and institute in audio classification index database
State the matched audio categories information of audio feature information;According to the audio categories information, searched in music assorting index database
With the music track of the audio categories information matches, and the music track matched is determined as being used to recommend
Recommend music track.
Match with the above method, another aspect of the invention provides a kind of computer readable storage medium, is stored thereon with
Computer program, the step of any of the above-described the method is realized when described program is executed by processor.
The embodiment of the present invention has the advantages that a kind of determination side for recommending music track provided in an embodiment of the present invention
Method, device and storage medium can be improved the precision for recommending music track, improve user experience.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of the determination method for recommendation music track that the embodiment of the present invention 1 provides;
Fig. 2 is convolutional Neural used in a kind of determination method for recommendation music track that the embodiment of the present invention 1 provides
The circuit theory schematic diagram of network model;
Fig. 3 is a kind of structural schematic diagram of the determining device for recommendation music track that the embodiment of the present invention 2 provides.
In figure: 301- acquiring unit;302- extraction unit;303- matching unit;304- determination unit.
Specific embodiment
Embodiments of the present invention are illustrated by particular specific embodiment below, those skilled in the art can be by this explanation
Content disclosed by book is understood other advantages and efficacy of the present invention easily.
Embodiment 1
Embodiment according to the present invention 1 provides a kind of determination method for recommending music track, as shown in Figure 1, being this hair
A kind of flow diagram of the determination method for recommendation music track that bright embodiment 1 provides.This method at least includes the following steps:
S101 obtains audio file to be identified;
S102 extracts the including at least belonging to audio of audio file to be identified according to preset audio feature extraction model
The audio feature information of classification;
S103, search and the matched audio categories information of audio feature information in audio classification index database;
S104 searches for the music with audio categories information matches according to audio categories information in music assorting index database
Song, and the music track matched is determined as to the recommendation music track for being used to recommend;In this way, implementing through the invention
The determination method that example 1 provides can be improved the precision for recommending music track, improve user experience.
It should be noted that searching in above-mentioned steps 103 in audio classification index database matches with audio feature information
Audio categories information search process and information matches process used in technological means, be conventional technology,
Details are not described herein.
In an optional example, structure is carried out to preset audio feature extraction model by convolutional neural networks model
It builds.
Convolutional neural networks model: convolutional neural networks (Convoltional Neural Networks, CNN) are a kind of
Comprising convolution or relevant calculation and with depth structure feedforward neural network (Feedforward Neural Networks),
It is one of the representative algorithm of deep learning (deep learning).Convolutional neural networks copy the visual perception (visual of biology
Perception) mechanism construction, can exercise supervision study and unsupervised learning, the convolution kernel parameter sharing in hidden layer and
The sparsity of interlayer connection, enables convolutional neural networks to reveal (grid-like with lesser calculation amount plaid matching
Topology) feature, such as pixel and audio are learnt, there is stable effect and the Feature Engineering not additional to data is wanted
It asks.
Particle swarm algorithm, also referred to as particle swarm optimization algorithm or flock of birds foraging algorithm (Particle Swarm
Optimization), it is abbreviated as PSO.PSO algorithm belongs to one kind of evolution algorithm, and PSO algorithm is similar with simulated annealing,
It is also to find optimal solution from RANDOM SOLUTION by iteration, the quality of solution is evaluated by fitness, but it compares genetic algorithm
Rule is more simple, it does not have " intersection " (Crossover) of genetic algorithm and " variation " (Mutation) operation, it is by chasing after
With current search to optimal value find global optimum.This algorithm has the advantages that realize that easy, precision is high, convergence is fast.
Particle swarm algorithm is a kind of parallel algorithm.
It should be noted that the convolutional neural networks being related in the scheme for the determination method that the embodiment of the present invention 1 provides
Model is different from existing convolutional neural networks model, improves on the basis of existing convolutional neural networks model, this
The convolutional neural networks model that inventive embodiments 1 use is the convolutional neural networks with 8 layers to 9 layers, such convolutional Neural
Network model can learn different genres of music and (for example, having the song of different classifications style, or belong to different schools
Song) corresponding atlas image, and particle swarm algorithm is introduced, in this way, relative to the existing existing side for distinguishing music categories
For method, the precision of identification and the accuracy of recognition result can be improved.
In an optional example, convolutional neural networks model is constructed using Keras framework.
Keras be high level neural network an API, Keras by pure Python write into and base Tensorflow,
The rear end Theano and CNTK.
Kears has the advantages that
Advantage 1, Keras are people-oriented in design, reinforce rapid modeling, user can be rapidly by the mechanism of required model
It is mapped in Keras code, reduces the workload for writing code as far as possible, especially for mature types of models, thus plus
Fast development rate.
Advantage 2, support existing common structure, such as convolutional neural networks, time recurrent neural network etc., it is sufficient to cope with
A large amount of common application scenarios.
Advantage 3, high modularization, user almost can any combination modules construct required model.In Keras
In, any neural network model can be described as a graph model or series model, component therein be divided into
Lower module: neural net layer, loss function, activation primitive, initial method, regularization method, optimization engine.These modules can
By by arbitrarily reasonably in a manner of be put into graph model or series model and construct required model, user does not need to know every
The details of a module rear.This mode needs user to write a large amount of codes or described with language-specific compared to other software
The method efficiency of neural network structure is much higher, does not also allow error-prone.
Advantage 4 is based on Python, and user also can be used Python code and carry out descriptive model, therefore ease for use, expansible
Row is all very high.User can easily write the customized module of oneself, and either existing module is modified or expanded
Exhibition, therefore, can the easily new model and method of development and application, accelerate iteration speed.
Advantage 5, can between CPU and GPU seamless switching, suitable for different application environments.
It should be noted that setting two class deep learning models in Keras: one kind is series model, and one kind is logical
With model, difference is different topological structures.
Series model, series model belong to a subclass of universal model, are successively sequence between this each layer of model
Linear relationship can carry out constructing neural network plus various elements between kth layer and+1 layer of kth.These elements can pass through
One list class is formulated, and then generates corresponding model as parameter translation sequence model.
Universal model can be used to design extremely complex, randomly topologically structured neural network, such as directed acyclic net
Network, shared layer network etc..Similar to series model, universal model is by the application interface of function come Definition Model.Use letter
The application interface of numberization has multiple benefits, for example, the unique elements of determining function implementing result are its return values, and determines to return
The unique elements of value are then its parameters, this significantly reduces the workload of code tester;Because functional language is a form
System, as long as can with mathematical operation express as long as can be stated with this language, therefore, as long as be mathematically it is of equal value, that
Machine can replace the code of low efficiency without influencing result using still more efficient code of equal value.This side
Face facilitates analyst's program writing, on the other hand mathematically ensure that code efficiency again, when realizing manual time and machine
Between Dual high-efficiency.
Important object in Keras: activation object, initialization object and regularization object.
Activation object: when defining network layer, the critically important selection of what activation primitive formula is used.Keras is provided greatly
The activation primitive predefined is measured, a variety of different network structures of customization are facilitated.Using activation object, there are two types of sides in Keras
Method: first is that individually defining an active coating;Second is that defining required activation primitive by activation option inside prefilter layer.
Initialize object: the initial value for weighted value or bias term in random setting network layer activation primitive.It is good
Weights initialisation value can help to accelerate model convergence rate.
Regularization object: when modeling, regularization is to prevent a very common means of overfitting.In nerve
The means of regularization are also provided in network, are respectively applied to weight parameter, bias term and activation primitive.
The preset audio feature extraction model in scheme that the embodiment of the present invention 1 provides can accomplish unsupervised identification simultaneously
Feature extraction is carried out, which can simulate the behavior of the mankind, extract respective musical features for music data collection.
In addition, the scheme that the embodiment of the present invention 1 provides, due to constructing convolutional neural networks model, solution using Keras framework
The excessively complicated disadvantage of existing building model library of having determined, not only simplifies the building process of model library, and mould created
Type library is complete, and iteration is fast, and accelerates training using GPU (Graphics Processing Unit, graphics processor).Wherein,
GPU also known as shows core, vision processor, display chip, is a kind of specially in PC, work station, game machine and some
The microprocessor that image operation works in mobile device (such as tablet computer, smart phone).
As shown in Fig. 2, being volume used in a kind of determination method for recommendation music track that the embodiment of the present invention 1 provides
The circuit theory schematic diagram of product neural network model.
Description below is done for the English in Fig. 2:
Input: input;
Feature maps: Feature Mapping;
Convolutions: convolution;
Subsampling: double sampling;
Full connection: full connection;
Gaussian connections: Gauss connection.
According to structural schematic diagram as shown in Figure 2 it is found that the embodiment of the present invention 1 provide scheme in convolutional neural networks
Each convolutional layer in model includes three parts: convolution, pond and nonlinear activation function.
It should be noted that convolution, for common convolution operation, Keras provides corresponding convolutional layer API, including
One-dimensional, two and three dimensions convolution operations, cutting operation, zero padding operation etc..One-dimensional convolution is commonly known as convolution, because
Main application carries out convolution behaviour on the sequence data with Time alignment, using adjacent signal of the convolution kernel to one-dimensional data
Make to generate a tensor.Two-dimensional convolution is commonly known as airspace convolution, is typically employed in input data relevant to image,
It is also that convolution operation is carried out to input data using convolution kernel.The operation that Three dimensional convolution performs equally.
Pond is a kind of processing in convolutional neural networks to characteristics of image, is carried out usually after convolution operation.Pond
The purpose of change is to calculate sufficient statistic of the feature in part and prevent overfitting to reduce overall feature quantity
With reduction calculation amount.Equally, the extraction of the audio frequency characteristics in the scheme of the offer of the embodiment of the present invention 1, process class are provided
Seemingly, details are not described herein.
The pond layer of Keras is divided into maximum statistic pond and average statistic pond according to the statistic of calculating;According to
Dimension is divided into one-dimensional, two and three dimensions eating layer;It is divided into local pondization and global pool according to normalized set region.The present invention
The convolutional neural networks model in scheme that embodiment 1 provides extracts space characteristics using convolution, and using average pond.
In an optional example, in the scheme for the determination method that the embodiment of the present invention 1 provides, the method is also wrapped
It includes: according to the preference information of user, music track will be recommended to push on the mobile terminal of corresponding user.
In practical applications, by analyzing the preference information of user, can know: the music class that active user likes
Type, for example, active user likes rock music, alternatively, active user likes lyrical music, if passing through the preference of analysis user
Information, analyzed accordingly as a result, and based on analysis result known to: the music type that active user likes is rock music,
And the recommendation music for being recommended that the 1 preset audio feature extraction Model Matching provided goes out through the embodiment of the present invention
Song is also rock music, then the recommendation music track is pushed to the movement for currently equally liking the active user of rock music
In terminal, in this way, the music track is appreciated whenever and wherever possible convenient for active user, the recommendation side relative to existing music track
For method, the scheme that the embodiment of the present invention 1 provides can be provided accurately accurately to target user and meet user preference
The music track of degree charges to the music track of recommendation in this way, also can be applied to business model, for example, being known as meeting
Member carries out the charge for recommending music track monthly, quarterly or per year, improves user experience, avoid user great
It is aimlessly scanned in vast existing musical database, the music liked to active user can be rapidly searched for
Mesh, has saved a large amount of quality time of active user, and pushes on the mobile terminal of active user, improves user experience
Degree.
It should be noted that the music type in the music track for analyzing user's recommendation is consistent with the preference of user
When, it is conventional push technology by the push technology for recommending music track to push on the mobile terminal of corresponding user, herein
It repeats no more.
In conclusion a kind of determination method for recommendation music track that the embodiment of the present invention 1 provides, has below beneficial to effect
Fruit: the precision for recommending music track is improved, user experience is improved.
Embodiment 2
Embodiment according to the present invention 2 additionally provides a kind of determining device for recommending music track, as shown in figure 3, for this
A kind of structural schematic diagram of the determining device for recommendation music track that inventive embodiments 2 provide.The one of the offer of the embodiment of the present invention 2
It includes acquiring unit 301, extraction unit 302, matching unit 303 and determination unit 304 that kind, which recommends the determining device of music track,.
Specifically, acquiring unit 301, obtains audio file to be identified;
Extraction unit 302, according to preset audio feature extraction model, that extracts audio file to be identified includes at least sound
The audio feature information of frequency generic;
Matching unit 303 searches for the audio feature information extracted with extraction unit 302 in audio classification index database
The audio categories information matched;And
According to audio categories information, the music track with audio categories information matches is searched in music assorting index database;
The music track that matching unit 303 matches is determined as the recommendation music for being used to recommend by determination unit 304
Song;In this way, the determining device that the embodiment of the present invention 2 provides, can accomplish: improving the precision for recommending music track, improve
User experience.
In an optional example, described device further includes push unit (being not shown in Fig. 3), push unit, root
According to the preference information of user, the recommendation music track that determination unit 304 is determined is pushed to the mobile end of corresponding user
On end;In such manner, it is possible to the recommendation music track for being used to recommend accurately is pushed to user, in order to which user can be at any time
Music track is appreciated everywhere, improves the Experience Degree of user.
In an optional example, described device further includes model construction unit (being not shown in Fig. 3), model construction
Unit constructs preset audio feature extraction model by convolutional neural networks model;The detailed content of this part is tired of
The same or similar description in the method scheme of the embodiment of the present invention 1 is referred to, details are not described herein.
In an optional example, convolutional neural networks model is constructed using Keras framework;This part it is detailed in
Hold, please be referring to the same or similar description in the method scheme of the embodiment of the present invention 1, details are not described herein.
Part in the partial content in scheme that the embodiment of the present invention 2 provides and the scheme of the offer of the embodiment of the present invention 1
The same or similar part of content, please be referring to the description of the corresponding portion for the embodiment of the present invention 1, and details are not described herein.
In conclusion a kind of determining device for recommendation music track that the embodiment of the present invention 2 provides, has below beneficial to effect
Fruit: the precision for recommending music track is improved, user experience is improved.
Embodiment 3
Embodiment according to the present invention 3 additionally provides one kind for determining and recommends music destination device, and feature exists
In including that perhaps more than one program one of them or more than one program is stored in storage by memory and one
In device, and be configured to be executed by one or more than one processor the one or more programs include for into
The following instruction operated of row: audio file to be identified is obtained;According to preset audio feature extraction model, audio to be identified is extracted
The audio feature information including at least audio generic of file;Search and audio feature information in audio classification index database
Matched audio categories information;According to audio categories information, search and audio categories information matches in music assorting index database
Music track, and the music track matched is determined as to the recommendation music track for being used to recommend.
What the partial content in scheme that the embodiment of the present invention 3 provides was provided with the embodiment of the present invention 1 or embodiment 2
The same or similar part of partial content in scheme, please be referring to for the corresponding of the embodiment of the present invention 1 or embodiment 2
Partial description, details are not described herein.
In conclusion one kind that the embodiment of the present invention 3 provides recommends music destination device for determining, have with following
Beneficial effect: the precision for recommending music track is improved, user experience is improved.
Embodiment 4
Embodiment according to the present invention 4 also provides a kind of computer readable storage medium, is stored thereon with computer journey
Sequence, the step of any of the above-described the method is realized when described program is executed by processor.
Part in the partial content in scheme that the embodiment of the present invention 4 provides and the scheme of the offer of the embodiment of the present invention 1
The same or similar part of content, please be referring to the description of the corresponding portion for the embodiment of the present invention 1, and details are not described herein.
In conclusion a kind of computer readable storage medium that the embodiment of the present invention 4 provides, has the advantages that
The precision for recommending music track is improved, user experience is improved.
Although above having used general explanation and specific embodiment, the present invention is described in detail, at this
On the basis of invention, it can be made some modifications or improvements, this will be apparent to those skilled in the art.Therefore,
These modifications or improvements without departing from theon the basis of the spirit of the present invention are fallen within the scope of the claimed invention.
Claims (10)
1. a kind of determination method for recommending music track characterized by comprising
Obtain audio file to be identified;
According to preset audio feature extraction model, that extracts the audio file to be identified includes at least audio generic
Audio feature information;
Search and the matched audio categories information of the audio feature information in audio classification index database;
According to the audio categories information, the music with the audio categories information matches is searched in music assorting index database
Mesh, and the music track matched is determined as to the recommendation music track for being used to recommend.
2. the method according to claim 1, wherein the method also includes:
According to the preference information of user, the recommendation music track is pushed on the mobile terminal of corresponding user.
3. the method according to claim 1, wherein the method also includes:
The preset audio feature extraction model is constructed by convolutional neural networks model.
4. according to the method described in claim 3, it is characterized in that,
The convolutional neural networks model is constructed using Keras framework.
5. a kind of determining device for recommending music track characterized by comprising
Acquiring unit obtains audio file to be identified;
Extraction unit, according to preset audio feature extraction model, that extracts the audio file to be identified includes at least audio
The audio feature information of generic;
Matching unit is searched in audio classification index database and is matched with the audio feature information that the extraction unit extracts
Audio categories information;And
According to the audio categories information, the music with the audio categories information matches is searched in music assorting index database
Mesh;
Determination unit, the music track that the matching unit is matched are determined as the recommendation music for being used to recommend
Mesh.
6. device according to claim 5, which is characterized in that described device further includes push unit,
The push unit, according to the preference information of user, the recommendation music track that the determination unit is determined
It pushes on the mobile terminal of corresponding user.
7. device according to claim 5, which is characterized in that described device further includes model construction unit,
The model construction unit constructs the preset audio feature extraction model by convolutional neural networks model.
8. device according to claim 7, which is characterized in that
The convolutional neural networks model is constructed using Keras framework.
9. one kind recommends music destination device for determining, which is characterized in that include memory and one or one
Above program, one of them perhaps more than one program be stored in memory and be configured to by one or one with
It includes the instruction for performing the following operation that upper processor, which executes the one or more programs:
Obtain audio file to be identified;
According to preset audio feature extraction model, that extracts the audio file to be identified includes at least audio generic
Audio feature information;
Search and the matched audio categories information of the audio feature information in audio classification index database;
According to the audio categories information, the music with the audio categories information matches is searched in music assorting index database
Mesh, and the music track matched is determined as to the recommendation music track for being used to recommend.
10. a kind of computer readable storage medium, which is characterized in that be stored thereon with computer program, described program is processed
The step of claim 1-4 any the method is realized when device executes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811362181.7A CN109726310A (en) | 2018-11-15 | 2018-11-15 | A kind of determination method, apparatus and storage medium for recommending music track |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811362181.7A CN109726310A (en) | 2018-11-15 | 2018-11-15 | A kind of determination method, apparatus and storage medium for recommending music track |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109726310A true CN109726310A (en) | 2019-05-07 |
Family
ID=66295703
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811362181.7A Pending CN109726310A (en) | 2018-11-15 | 2018-11-15 | A kind of determination method, apparatus and storage medium for recommending music track |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109726310A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111488485A (en) * | 2020-04-16 | 2020-08-04 | 北京雷石天地电子技术有限公司 | Music recommendation method based on convolutional neural network, storage medium and electronic device |
CN115312074A (en) * | 2022-10-10 | 2022-11-08 | 江苏米笛声学科技有限公司 | Cloud server based on audio processing |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106991163A (en) * | 2017-03-31 | 2017-07-28 | 福州大学 | A kind of song recommendations method based on singer's sound speciality |
US20170220582A1 (en) * | 2012-08-30 | 2017-08-03 | deviantArt, Inc. | Tag Selection, Clustering, and recommendation for Content Hosting Services |
US20180053233A1 (en) * | 2016-08-16 | 2018-02-22 | Ebay Inc. | Expandable service architecture with configurable orchestrator |
CN107918653A (en) * | 2017-11-16 | 2018-04-17 | 百度在线网络技术(北京)有限公司 | A kind of intelligent playing method and device based on hobby feedback |
CN108334617A (en) * | 2018-02-07 | 2018-07-27 | 大连大学 | The method of semantic-based music retrieval |
-
2018
- 2018-11-15 CN CN201811362181.7A patent/CN109726310A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20170220582A1 (en) * | 2012-08-30 | 2017-08-03 | deviantArt, Inc. | Tag Selection, Clustering, and recommendation for Content Hosting Services |
US20180053233A1 (en) * | 2016-08-16 | 2018-02-22 | Ebay Inc. | Expandable service architecture with configurable orchestrator |
CN106991163A (en) * | 2017-03-31 | 2017-07-28 | 福州大学 | A kind of song recommendations method based on singer's sound speciality |
CN107918653A (en) * | 2017-11-16 | 2018-04-17 | 百度在线网络技术(北京)有限公司 | A kind of intelligent playing method and device based on hobby feedback |
CN108334617A (en) * | 2018-02-07 | 2018-07-27 | 大连大学 | The method of semantic-based music retrieval |
Non-Patent Citations (2)
Title |
---|
裔隽等: "《Python机器学习实践》", 28 February 2018 * |
高志强等: "《深度学习从入门到实践》", 30 June 2018 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111488485A (en) * | 2020-04-16 | 2020-08-04 | 北京雷石天地电子技术有限公司 | Music recommendation method based on convolutional neural network, storage medium and electronic device |
CN111488485B (en) * | 2020-04-16 | 2023-11-17 | 北京雷石天地电子技术有限公司 | Music recommendation method based on convolutional neural network, storage medium and electronic device |
CN115312074A (en) * | 2022-10-10 | 2022-11-08 | 江苏米笛声学科技有限公司 | Cloud server based on audio processing |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110929161B (en) | Large-scale user-oriented personalized teaching resource recommendation method | |
CN109902708A (en) | A kind of recommended models training method and relevant apparatus | |
CN111177569A (en) | Recommendation processing method, device and equipment based on artificial intelligence | |
CN104573130B (en) | The entity resolution method and device calculated based on colony | |
CN110209807A (en) | A kind of method of event recognition, the method for model training, equipment and storage medium | |
CN111444394A (en) | Method, system and equipment for obtaining relation expression between entities and advertisement recalling system | |
CN107391582B (en) | The information recommendation method of user preference similarity is calculated based on context ontology tree | |
CN112380453B (en) | Article recommendation method and device, storage medium and equipment | |
CN111444395A (en) | Method, system and equipment for obtaining relation expression between entities and advertisement recalling system | |
CN109933656A (en) | Public sentiment polarity prediction technique, device, computer equipment and storage medium | |
CN107786943A (en) | A kind of tenant group method and computing device | |
CN110532351A (en) | Recommend word methods of exhibiting, device, equipment and computer readable storage medium | |
Zubiaga et al. | Content-based clustering for tag cloud visualization | |
CN109902823A (en) | A kind of model training method and equipment based on generation confrontation network | |
CN110737805A (en) | Method and device for processing graph model data and terminal equipment | |
CN112131261A (en) | Community query method and device based on community network and computer equipment | |
CN109726310A (en) | A kind of determination method, apparatus and storage medium for recommending music track | |
CN111506755A (en) | Picture set classification method and device | |
CN115687760A (en) | User learning interest label prediction method based on graph neural network | |
KR102376652B1 (en) | Method and system for analazing real-time of product data and updating product information using ai | |
CN111192170A (en) | Topic pushing method, device, equipment and computer readable storage medium | |
CN107291894A (en) | A kind of probability matrix decomposition model for merging similitude and common scoring item quantity | |
KR102119518B1 (en) | Method and system for recommending product based style space created using artificial intelligence | |
CN112784899A (en) | Method, device and equipment for mining frequent pattern of power transformation operation and maintenance knowledge and storage medium | |
CN109977979A (en) | Position method, apparatus, electronic equipment and the storage medium of seed user |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190507 |