Audio track identification method, device and readable storage medium

Abstract

The present application discloses an audio track recognition method, device and readable storage medium, the method comprising: obtaining the target audio to be identified, and after segmenting the target audio, extracting the audio fingerprint features of each segment; obtaining the song IDs that match each audio fingerprint feature; clustering the song IDs to obtain clustering results; using the clustering results, determining the confidence of each song ID; using the confidence, filtering out the target tracks of the target audio from the tracks corresponding to each song ID. The present application can identify the target tracks of the target audio without marking the tracks included in the target audio. In order to further display the tracks to the user, the audio is labeled and track information is provided.

Description

Technical Field

[0001] The present application relates to the field of audio technology, and in particular to an audio track identification method, device and readable storage medium. Background Art

[0002] There are many types of songs in the music library, such as ordinary songs and medley songs. In the field of long audio, many long audio programs belonging to the music category will recommend multiple songs, or other types of long audio will also be interspersed with many songs. In other words, in some audio, there may be a single song or multiple songs.

[0003] Usually, when a user is listening to an audio, there may be songs that the user is interested in. However, when the audio does not have a corresponding track list, it is impossible to determine the track that the audio corresponds to, and it is difficult to display the track that the audio corresponds to to the user.

[0004] In summary, how to effectively solve the problem of identifying the corresponding audio tracks is a technical problem that technical personnel in this field urgently need to solve. Summary of the invention

[0005] The purpose of this application is to provide an audio track identification method, device and readable storage medium, which can effectively identify the track corresponding to the audio.

[0006] In order to solve the above technical problems, this application provides the following technical solutions:

[0007] In a first aspect, the present application provides an audio track identification method, comprising:

[0008] Acquire target audio to be identified, and after segmenting the target audio, extract audio fingerprint features of each segment;

[0009] Obtain song IDs that match each of the audio fingerprint features;

[0010] Clustering the song IDs to obtain a clustering result;

[0011] Using the clustering results, determining the confidence of each song ID;

[0012] The confidence level is used to select the target tracks of the target audio from the tracks corresponding to the song IDs.

[0013] In a feasible implementation manner, the song IDs are clustered to obtain a clustering result, including:

[0014] Take each of the song IDs as a point;

[0015] Connect the corresponding points of the song IDs of the same song to obtain an undirected graph;

[0016] Each of the undirected graphs is respectively determined as a cluster of the same song to obtain the clustering result.

[0017] In a feasible implementation, the confidence of each song ID is determined by using the clustering result, including:

[0018] Using the clustering results, determine the total number of hits for the same song and the number of hits corresponding to each song ID;

[0019] Compare each of the hit counts with the total hit count for the same song to obtain the confidence level.

[0020] In a feasible implementation manner, the confidence level is used to select a target track of the target audio from tracks corresponding to each song ID, including:

[0021] If the clustering result corresponds to at least two cluster centers, then for the same song, the song ID corresponding to the song with the highest confidence is determined as the medley component track; the medley component track corresponding to each song is determined as the target track; wherein one cluster center corresponds to the same song;

[0022] If the clustering result corresponds to a cluster center, the track corresponding to the song ID with the highest confidence is determined as the target track.

[0023] In a feasible implementation, it also includes:

[0024] For the same song, if there are at least two song IDs with the same confidence, the playback volume corresponding to the song ID is obtained, and the track corresponding to the song ID with the highest playback volume is used as the medley component track or the target track.

[0025] In a feasible implementation manner, the obtaining of the target audio to be identified includes:

[0026] Receive a track identification request sent by a long audio client;

[0027] The long audio corresponding to the track identification request is determined as the target audio.

[0028] In a feasible implementation manner, after using the confidence to select the target track of the target audio from the tracks corresponding to each of the song IDs, the method further includes:

[0029] If the clustering result corresponds to at least two cluster centers, the target audio is determined to be a medley song, and a medley song label is added to the target audio.

[0030] In a feasible implementation, it also includes:

[0031] The target track is output, and a play button corresponding to the target track is popped up.

[0032] In a second aspect, the present application provides an electronic device, including:

[0033] Memory for storing computer programs;

[0034] A processor is used to implement the steps of the above-mentioned audio track identification method when executing the computer program.

[0035] In a third aspect, the present application provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned audio track identification method are implemented.

[0036] Apply the method provided in the embodiment of the present application to obtain the target audio to be identified, and after segmenting the target audio, extract the audio fingerprint features of each segment; obtain the song ID that matches each audio fingerprint feature; cluster the song IDs with the same song to obtain clustering results; use the clustering results to determine the confidence of each song ID; use the confidence to filter out the target track of the target audio from the tracks corresponding to each song ID.

[0037] After obtaining the target audio to be identified, in order to avoid the situation where the tracks cannot be identified due to the inclusion of multiple songs in the same audio. In this application, it is proposed to segment the target audio and extract the audio fingerprint features of each segment, and then obtain the song ID corresponding to each audio fingerprint feature. In this way, even if the same audio corresponds to multiple songs, each track can be accurately identified. Specifically, in order to effectively identify the tracks, the song IDs can be clustered with the same song, and the confidence of each song ID can be determined based on the clustering results. The confidence level indicates whether the track of the corresponding song ID corresponds to the track involved in the target audio. Therefore, finally, based on the confidence level, the target track of the target audio can be screened out from the tracks corresponding to each song ID. In other words, the present application can identify the target track of the target audio without marking the tracks contained in the target audio. In order to further display the tracks to the user, the audio is labeled and track information is provided.

[0038] Correspondingly, the embodiments of the present application also provide a device and a readable storage medium corresponding to the above-mentioned audio track identification method, which have the above-mentioned technical effects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 This is a flowchart of an implementation method of an audio track identification method in an embodiment of the present application;

[0041] Figure 2 This is a specific slicing schematic diagram in an embodiment of the present application;

[0042] Figure 3a This is a schematic diagram of an undirected graph drawing stage 1 in an embodiment of the present application;

[0043] Figure 3b This is a schematic diagram of an undirected graph drawing stage 2 in an embodiment of the present application;

[0044] Figure 3c This is a schematic diagram of an undirected graph drawing stage 3 in an embodiment of the present application;

[0045] Figure 4 This is a schematic diagram of the clustering results of the same song based on an undirected graph in an embodiment of the present application;

[0046] Figure 5 This is a schematic diagram of the structure of an audio track identification device in an embodiment of the present application;

[0047] Figure 6 This is a schematic diagram of the structure of an electronic device in an embodiment of the present application;

[0048] Figure 7 This is a schematic diagram of the specific structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to enable those skilled in the art to better understand the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0050] Please refer to Figure 1 , Figure 1 This is a flowchart of an audio track identification method in an embodiment of the present application, the method comprising the following steps:

[0051] S101, obtaining target audio to be identified, and after segmenting the target audio, extracting audio fingerprint features of each segment.

[0052] The target audio may be any audio in the long audio field, or a song whose tracks are unknown in a music library. In this embodiment, the number and type of tracks specifically corresponding to the target audio are not limited.

[0053] In this embodiment, the target audio can be determined by searching whether there is a track in the music library, and then the audio without a track is determined as the target audio; or the target audio can be obtained by receiving audio sent by an upper-layer application / external device. In this embodiment, the specific method of obtaining the target audio is not limited.

[0054] Preferably, obtaining the target audio to be recognized in step S101 may specifically include:

[0055] Step 1: Receive a track identification request sent by a long audio client;

[0056] Step 2: Determine the long audio corresponding to the track identification request as the target audio.

[0057] That is, after the long audio client sends a track identification request, the long audio corresponding to the track identification request can be determined as the target audio. In this way, in the long audio service, the track information in the long audio that the customer is interested in can be identified.

[0058] After obtaining the target audio, in order to effectively identify the track, the target audio needs to be segmented and the audio fingerprint features of each segment need to be extracted. Then, the track of the target audio can be determined based on these audio fingerprint features.

[0059] Specifically, when the target audio is segmented, the target audio can be segmented in sequence according to the fixed length. In this embodiment, the specific length of the fixed length can be set and adjusted according to actual needs, and can generally be set between 6 and 15 seconds, and can certainly be set to other values.

[0060] The methods of slicing include but are not limited to the following:

[0061] Method 1: Evenly segment the target audio according to a fixed length. For example, if the target audio is 180 seconds long, you can divide the 180 seconds of target audio into 30 segments of a fixed length of 6 seconds according to the preset segment length of 6 seconds. Figure 2 As shown, in Figure 2 In the figure, each fingerprint can be represented as fp1 (corresponding to Figures 1 to 6s), fp2 (corresponding to Figures 7-12s), ...

[0062] Method 2: In order to further improve the accuracy of song recognition, the slicing density can also be improved. For example, if the target audio is 18 seconds long, the 18-second target audio can be divided into 7 segments of 1-6 seconds, 3-8 seconds, 4-10 seconds, 6-12 seconds, 8-14 seconds, 10-16 seconds, and 12-18 seconds according to the preset segment length of 6 seconds and the segment interval of 2 seconds.

[0063] Method 3: In practical applications, if the target audio is long, in order to avoid wasting time and a large amount of computer resources, the target audio can also be sliced ​​by uniform sampling, such as cutting out a 6-second slice every 30 seconds.

[0064] After the target audio is segmented into several segments, the operation of step S102 can be performed.

[0065] S102: Obtain song IDs that match each audio fingerprint feature.

[0066] In this embodiment, songs of known tracks can be segmented in advance, and then the audio fingerprint features of each segment can be extracted, and the corresponding song ID can be marked for each audio fingerprint feature. In this way, after the target audio is segmented and the audio fingerprint features corresponding to each segment are extracted, the song ID corresponding to these audio fingerprint features corresponding to the target audio can be determined by feature comparison.

[0067] It should be noted that in this embodiment, one shard corresponds to one audio fingerprint feature, but one audio fingerprint feature can correspond to 0, 1 or more song IDs. In other words, for the same shard, there may be no corresponding song ID, it may correspond to 1 song ID, or it may correspond to multiple song IDs. That is, the audio fingerprint feature of each shard may match N song IDs (N is indefinite, it may be 0, it may be 1. For ease of processing, N can be set not to be greater than M. The value of M can be determined according to actual application requirements, such as it can be set to the average number of versions of the same song, that is, at most M results can be returned). In this application, the song ID corresponds to the audio fingerprint feature, and the same song can have multiple different song IDs and corresponding tracks due to its different versions, etc. (generally, the track corresponds to the song name plus a brief description of the version, such as the song name and singer name).

[0068] For example, for the target audio, an id of 000 is corresponding, and the corresponding song IDs of the first three fragments are as follows:

[0069] The first hit: 123, 124,

[0070] 2nd piece hits: 124,

[0071] The third piece hits: 123, 125;

[0072] Among them, 123, 124 and 125 refer to different song IDs respectively.

[0073] S103, clustering the songs with the same ID to obtain a clustering result.

[0074] Considering that there is usually only one version of a song in the same audio, after obtaining the song IDs corresponding to each shard, these song IDs can be clustered to determine the number of songs and other information, so as to subsequently determine which song ID corresponds to the track of the same song.

[0075] Specifically, in practical applications, the correspondence between different song IDs and songs can be saved in advance. When the song IDs are clustered, they can be classified into one category based on the correspondence between the song ID and a certain song, and finally the clustering result can be obtained.

[0076] Specifically, the clustering result indicates which song IDs are clustered into one category, which song IDs are discrete, the number of song IDs specifically included in a cluster center, etc. Generally, a cluster center corresponds to one song.

[0077] S104: Determine the confidence level of each song ID using the clustering results.

[0078] From the clustering results, the confidence level corresponding to each song ID can be determined, which indicates the possibility that the track corresponding to the song ID belongs to the track included in the target audio.

[0079] Specifically, for the same song, the higher the confidence of the song ID, the more likely the track of the song ID is to be the actual corresponding song in the target audio.

[0080] S105. Filter out target tracks of target audio from tracks corresponding to each song ID using the confidence level.

[0081] After obtaining the confidence level, the target track of the target audio can be filtered out from the tracks corresponding to each song ID according to the confidence level.

[0082] Specifically, a confidence threshold can be set in advance, and then the tracks corresponding to the song IDs greater than the predicted confidence level are determined as the target tracks of the target audio; of course, the confidence levels can also be sorted, and then the tracks corresponding to the top N song IDs are taken as the target tracks of the target audio.

[0083] In a specific implementation of the present application, the target track can also be output, and a play button corresponding to the target track can be popped up. In this way, the target track information corresponding to the audio clip / full song without the track list information can be identified, and the corresponding play button can be popped up, so that the user can directly click to play the song / full song corresponding to the target track when interested.

[0084] Apply the method provided in the embodiment of the present application to obtain the target audio to be identified, and after segmenting the target audio, extract the audio fingerprint features of each segment; obtain the song ID that matches each audio fingerprint feature; cluster the song IDs with the same song to obtain clustering results; use the clustering results to determine the confidence of each song ID; use the confidence to filter out the target track of the target audio from the tracks corresponding to each song ID.

[0085] After obtaining the target audio to be identified, in order to avoid the situation where the tracks cannot be identified due to the inclusion of multiple songs in the same audio. In this application, it is proposed to segment the target audio and extract the audio fingerprint features of each segment, and then obtain the song ID corresponding to each audio fingerprint feature. In this way, even if the same audio corresponds to multiple songs, each track can be accurately identified. Specifically, in order to effectively identify the tracks, the song IDs can be clustered with the same song, and the confidence of each song ID can be determined based on the clustering results. The confidence level indicates whether the track of the corresponding song ID corresponds to the track involved in the target audio. Therefore, finally, based on the confidence level, the target track of the target audio can be screened out from the tracks corresponding to each song ID. In other words, the present application can identify the target track of the target audio without marking the tracks contained in the target audio. In order to further display the tracks to the user, the audio is labeled and track information is provided.

[0086] It should be noted that, based on the above embodiments, the embodiments of the present application also provide corresponding improved solutions. In the preferred / improved embodiments, the same steps or corresponding steps as those in the above embodiments can be referenced to each other, and the corresponding beneficial effects can also be referenced to each other, which will not be described one by one in the preferred / improved embodiments of this article.

[0087] In a specific embodiment of the present application, step 103 may specifically include:

[0088] Step 1: Use each song ID as a point;

[0089] Step 2: Connect the corresponding points of the song ID of the same song to obtain an undirected graph;

[0090] Step 3: Determine each undirected graph as a cluster of the same song and obtain the clustering result.

[0091] For ease of description, the following Figure 3a-3c Take as an example, the above three steps are combined for explanation.

[0092] like Figure 3a As shown, each song ID hit is regarded as a point in the graph; according to the hit result of the first slice, song IDs 123 and 124 belong to the same song, so a line is drawn between the two, as shown in Figure 3b ; Since the hit result of the third piece shows that 123 and 125 are the same song, a line is connected between 123 and 125, such as Figure 3c As shown; if the same steps are followed, the clustering result of the entire song will be obtained, as shown Figure 4 As shown, there are three independent undirected graphs, that is, they are clustered into three categories, indicating that three tracks may be mixed in the target audio, and song IDs 123, 124 and 125 correspond to one song, song IDs 128 and 129 correspond to one song, and song IDs 126 and 127 correspond to one song.

[0093] Since there is no need to calculate the similarity distance between each point in an undirected graph, the undirected graph method can be used to achieve fast clustering of the same songs.

[0094] In a specific implementation of the present application, the above step S104 uses the clustering results to determine the confidence of each song ID, which specifically includes:

[0095] Step 1: Using the clustering results, determine the total number of hits for the same song and the number of hits corresponding to each song ID;

[0096] Step 2: Compare each hit count with the total hit count of the same song to get the confidence level.

[0097] For ease of description, the above two steps are combined for explanation below.

[0098] Based on the clustering results, the total number of hits for the same song and the number of hits for each song ID in the same song can be determined. In this way, it can be determined which song ID has the highest confidence for the same song. Specifically, the ratio of each hit number to the total number of hits for the same song can be used to determine the corresponding confidence. In other words, if confidence is expressed as confidence, confidence = the number of hit time slices for a certain song ID / the total number of hits for the same song with the song ID. For example, the hit results of the first three segments 1, 2, and 3 of a target audio indicate that the same song (song ID: 123, 124, 125) hits a time period of 1 to 18 seconds (each segment is 6 seconds long), i.e., three segments are hit, of which song ID=123 hits two segments, so the confidence of song ID=123 is 2*6 / 3*6=2 / 3, song ID=124 hits two segments, corresponding to confidence=2*6 / 3*6=2 / 3, song ID=125 hits one segment, so confidence=1*6 / 3*6=1 / 3.

[0099] In other words, the higher the ratio of the number of hits for a song ID to the total number of hits for the same song, the higher the confidence. In this way, the confidence can represent the proportion of the song ID hits in the same song, and can effectively represent the possibility that the track corresponding to the song ID is the track corresponding to the target audio, which is further conducive to screening out more accurate target tracks.

[0100] In a specific implementation of the present application, the above step S105, using the confidence level to select the target tracks of the target audio from the tracks corresponding to each song ID, may specifically include the following situations:

[0101] Case 1: If the clustering result corresponds to at least two cluster centers, then for the same song, the track corresponding to the song ID with the highest confidence is determined as the medley component track; the medley component track corresponding to each song is determined as the target track.

[0102] Among them, one cluster center corresponds to the same song.

[0103] When the clustering result corresponds to at least two cluster centers, it can be determined that the target audio includes at least 2 songs. Therefore, the target audio corresponds to at least 2 target tracks. For the convenience of distinguishing descriptions, in this embodiment, when the target track corresponds to at least 2 tracks, each corresponding track is referred to as a medley track. Specifically, for each medley track, that is, it is determined from a cluster center, for example, the track corresponding to a song ID with the highest confidence can be selected from the multiple song IDs corresponding to the cluster center as the medley track. Preferably, for the convenience of output and statistics, when the target track corresponds to multiple medley tracks, these medley tracks can also be listed to obtain a track list of the target audio.

[0104] Case 2: If the clustering result corresponds to a cluster center, the track corresponding to the song ID with the highest confidence is determined as the target track.

[0105] If the clustering result shows that there is only one cluster center, it means that the target audio includes only one song, so the track corresponding to the song ID with the highest confidence can be directly determined as the target track.

[0106] By screening the tracks according to different situations, it is possible to avoid the situation where the target audio corresponds to only one song, while the determined target tracks correspond to two different versions of the song. That is, by screening the tracks according to different situations, the final target tracks can be more in line with the actual situation.

[0107] Preferably, in actual applications, the same confidence may occur within the same song. In order to make the final target track as authentic as possible, for the same song, if there are at least two song IDs with the same confidence, the playback volume of the corresponding song ID is obtained, and the track corresponding to the song ID with the highest playback volume is used as the medley component track or target track. That is, the higher the reference playback volume, the more likely it is to be authentic. Therefore, in the case of the same confidence, the track corresponding to the song ID with the highest playback volume can be determined as the target track or medley component track.

[0108] Preferably, after selecting the target tracks of the target audio from the tracks corresponding to each song ID by using the confidence level, if the clustering result corresponds to at least two cluster centers, the target audio is determined to be a medley song, and a medley song label is added to the target audio.

[0109] That is to say, when it is determined that the clustering result corresponds to at least two cluster centers, it can be clear that the target audio includes at least two songs, so the target audio can be determined as a medley song, and a medley song label is added to the target audio. In this way, when recommending songs to users, more accurate songs can be recommended to users in combination with the transmitted song labels, thereby improving user experience.

[0110] In a specific implementation of the present application, after executing step S105, the target tracks of the target audio are screened out from the tracks corresponding to each song ID using the confidence level, and then the constituent tracks can be effectively screened. Specific screening implementation process:

[0111] Step 1: Perform same song statistics for each song ID to obtain the actual hit play time corresponding to each song and the same song hit play time after same song processing;

[0112] Step 2: Compare the corresponding actual hit play time with the hit play time of the same song to obtain the play ratio of each song;

[0113] Step 3: Eliminate, from the target tracks, the tracks that make up the medley whose play percentage is less than the play percentage threshold.

[0114] Among them, the specific implementation process of same song statistics for song ID can be referred to the above description, which will not be repeated here.

[0115] In this embodiment, after the same song statistics are performed, the actual playback time corresponding to each song can be obtained. Same song processing is to merge the time period of the same song hit and the smaller interval segments, and regard the merged time period as the same song hit playback time. For example, the hit time period of group A songs (i.e. a certain song) is 1 to 20s and 30 to 40s, and there is no hit in 20 to 30s, but because the duration of the miss is short, when performing the same song processing, the hit time of group A songs can be adjusted to 1 to 40s, that is, the actual hit playback time is 30s, and the same song hit playback time is 40s.

[0116] After obtaining the actual hit play time and the same song hit play time, the play ratio of each song can be calculated. For example, if the actual hit play time of the songs in group A is 30s and the same song hit play time is 40s, the play ratio of the songs in group A is 30 / 40, or 75%.

[0117] It can be seen that if scale is used to represent the play ratio, then scale = the actual play time of a group of the same song / the play time of the group of the same song after processing. The larger the scale, the more time the song is played in this time period, which can be used to determine whether the song is inserted when the host is speaking or is a complete song. Specifically, the smaller the scale, the more likely the host's confession is interspersed in the middle of the song, and the larger the scale, the greater the probability of playing the complete song during this period.

[0118] In this way, the identified target tracks can be filtered based on the play ratio, and the tracks with a play ratio less than the play ratio threshold can be removed. The play ratio threshold can be set according to actual application requirements, such as 80%.

[0119] Based on the playback ratio, the tracks that make up the medley can be effectively screened, making the final target tracks more accurate.

[0120] Corresponding to the above method embodiment, the embodiment of the present application further provides an audio track identification device. The audio track identification device described below and the audio track identification method described above can refer to each other.

[0121] See also Figure 5 As shown, the device includes the following modules:

[0122] The feature extraction module 101 is used to obtain the target audio to be identified, and after segmenting the target audio, extract the audio fingerprint features of each segment;

[0123] The song ID determination module 102 is used to obtain the song IDs that match the audio fingerprint features respectively;

[0124] The same song clustering module 103 is used to cluster the song IDs to obtain clustering results;

[0125] The confidence calculation module 104 is used to determine the confidence of each song ID using the clustering results;

[0126] The track determination module 105 is used to filter out the target track of the target audio from the tracks corresponding to each song ID by using the confidence level.

[0127] By using the device provided in the embodiment of the present application, the target audio to be identified is obtained, and after the target audio is segmented, the audio fingerprint features of each segment are extracted; the song IDs that match each audio fingerprint feature are obtained; the song IDs are clustered together to obtain clustering results; the confidence of each song ID is determined using the clustering results; and the target tracks of the target audio are screened out from the tracks corresponding to each song ID using the confidence.

[0128] After obtaining the target audio to be identified, in order to avoid the situation where the tracks cannot be identified due to the inclusion of multiple songs in the same audio. In this application, it is proposed to segment the target audio and extract the audio fingerprint features of each segment, and then obtain the song ID corresponding to each audio fingerprint feature. In this way, even if the same audio corresponds to multiple songs, each track can be accurately identified. Specifically, in order to effectively identify the tracks, the song IDs can be clustered with the same song, and the confidence of each song ID can be determined based on the clustering results. The confidence level indicates whether the track of the corresponding song ID corresponds to the track involved in the target audio. Therefore, finally, based on the confidence level, the target track of the target audio can be screened out from the tracks corresponding to each song ID. In other words, the present application can identify the target track of the target audio without marking the tracks contained in the target audio. In order to further display the tracks to the user, the audio is labeled and track information is provided.

[0129] In a specific implementation of the present application, the same song clustering module 103 is specifically used to take each song ID as a point; connect the corresponding points of the song ID of the same song to obtain an undirected graph; determine each undirected graph as a same song cluster to obtain a clustering result.

[0130] In a specific implementation of the present application, the confidence calculation module 104 is specifically used to use the clustering results to determine the total number of hits for the same song and the number of hits corresponding to each song ID; and compare each hit number with the total number of hits corresponding to the same song to obtain the confidence.

[0131] In a specific implementation of the present application, the track determination module 105 is specifically used to, if the clustering result corresponds to at least two clustering centers, then for the same song, the track corresponding to the song ID with the highest confidence is determined as the medley component track; the medley component track corresponding to each song is determined as the target track; wherein, one clustering center corresponds to the same song; if the clustering result corresponds to one clustering center, the track corresponding to the song ID with the highest confidence is determined as the target track.

[0132] In a specific embodiment of the present application, the track determination module 105 is also used to obtain the playback volume of the corresponding song ID if there are at least two song IDs with the same confidence level for the same song, and use the track corresponding to the song ID with the highest playback volume as the medley component track or target track.

[0133] In a specific implementation of the present application, the feature extraction module 101 is specifically used to receive a track identification request sent by a long audio client; and determine the long audio corresponding to the track identification request as the target audio.

[0134] In a specific implementation of the present application, it also includes:

[0135] The label marking module is used to use confidence to filter out the target tracks of the target audio from the tracks corresponding to each song ID. If the clustering result corresponds to at least two clustering centers, it is determined that the target audio is a medley song and a medley song label is added to the target audio.

[0136] In a specific implementation of the present application, it also includes:

[0137] The output module is used to output the target track and pop up the play button corresponding to the target track.

[0138] Corresponding to the above method embodiment, the embodiment of the present application further provides an electronic device. The electronic device described below and the audio track identification method described above can refer to each other.

[0139] See also Figure 6 As shown, the electronic device includes:

[0140] A memory 332, for storing computer programs;

[0141] The processor 322 is used to implement the steps of the audio track identification method of the above method embodiment when executing the computer program.

[0142] For details, please refer to Figure 7 , Figure 7 A specific structural diagram of an electronic device provided for this embodiment, the electronic device may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 322 (for example, one or more processors) and a memory 332, and the memory 332 stores one or more computer applications 342 or data 344. Among them, the memory 332 can be a temporary storage or a permanent storage. The program stored in the memory 332 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the data processing device. Furthermore, the central processing unit 322 can be configured to communicate with the memory 332 to execute a series of instruction operations in the memory 332 on the electronic device 301.

[0143] The electronic device 301 may further include one or more power supplies 326 , one or more wired or wireless network interfaces 350 , one or more input and output interfaces 358 , and / or one or more operating systems 341 .

[0144] The steps in the audio track identification method described above can be implemented by the structure of an electronic device.

[0145] Corresponding to the above method embodiment, the embodiment of the present application also provides a readable storage medium. The readable storage medium described below and the audio track identification method described above can refer to each other.

[0146] A readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the audio track identification method of the above method embodiment.

[0147] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0148] Those skilled in the art may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

Claims

1. A method for identifying an audio track, characterized in that: include: Acquire target audio to be identified, and after segmenting the target audio, extract audio fingerprint features of each segment; Obtain song IDs that match each of the audio fingerprint features; Clustering the song IDs to obtain a clustering result; Using the clustering results, determining the confidence of each song ID; Using the confidence level, select a target track of the target audio from tracks corresponding to each of the song IDs; Wherein, using the confidence level to select the target track of the target audio from the tracks corresponding to each of the song IDs includes: If the clustering result corresponds to at least two cluster centers, then for the same song, the song ID corresponding to the song with the highest confidence is determined as the medley component track; the medley component track corresponding to each song is determined as the target track; wherein one cluster center corresponds to the same song; If the clustering result corresponds to one of the cluster centers, the track corresponding to the song ID with the highest confidence is determined as the target track; For the same song, if there are at least two song IDs with the same confidence, the playback volume corresponding to the song ID is obtained, and the track corresponding to the song ID with the highest playback volume is used as the medley component track or the target track.

2. The audio track identification method according to claim 1, characterized in that: The song IDs are clustered to obtain clustering results, including: Take each of the song IDs as a point; Connect the corresponding points of the song IDs of the same song to obtain an undirected graph; Each of the undirected graphs is respectively determined as a cluster of the same song to obtain the clustering result.

3. The audio track identification method according to claim 1, characterized in that: Using the clustering results, the confidence of each song ID is determined, including: Using the clustering results, determine the total number of hits for the same song and the number of hits corresponding to each song ID; Compare each of the hit counts with the total hit count for the same song to obtain the confidence level.

4. The audio track identification method according to claim 1, characterized in that: The step of obtaining the target audio to be identified includes: Receive a track identification request sent by a long audio client; The long audio corresponding to the track identification request is determined as the target audio.

5. The audio track identification method according to claim 1, characterized in that: Also includes: The target track is output, and a play button corresponding to the target track is popped up.

6. The audio track identification method according to any one of claims 1 to 5, characterized in that: After using the confidence to select the target track of the target audio from the tracks corresponding to the song IDs, the method further includes: If the clustering result corresponds to at least two cluster centers, the target audio is determined to be a medley song, and a medley song label is added to the target audio.

7. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the audio track identification method as claimed in any one of claims 1 to 6 when executing the computer program.

8. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the audio track identification method according to any one of claims 1 to 6.

Images