A method and related device for identifying songs with repeated content

Abstract

The present application discloses a method for identifying songs with duplicate content, including: performing song matching based on multiple audio clips of a song to be identified to obtain a target song; determining the offset time position of each audio clip in the target song among the multiple audio clips; wherein the offset time position is used to characterize the time point at which the audio clip matches the target song; calculating the difference between the distribution curve to be identified formed by all offset time positions and the standard time distribution curve; if the difference is greater than a threshold, the song to be identified is determined to be a song with duplicate content. The difference between multiple distribution curves to be identified and the standard time distribution curve is calculated. If the difference is greater, the song to be identified is determined to be a song with duplicate content, rather than using a model to identify songs with duplicate content, thereby improving the efficiency of identifying songs with duplicate content. The present application also discloses a server and a computer-readable storage medium, which have the above beneficial effects.

Description

Technical Field

[0001] The present application relates to the field of multimedia data processing technology, and in particular to a method for identifying songs with duplicate content, a server, and a computer-readable storage medium. Background Art

[0002] In multimedia data, songs composed of fragments of regular songs and then spliced together are called duplicate songs. Duplicate songs disrupt the normal viewing and listening experience and need to be identified and removed.

[0003] Related technologies use neural networks to annotate data on normal songs and songs with duplicate content, then train a model to identify duplicate songs. However, this training process relies on a large amount of annotated data, and the annotation costs and errors increase with the amount of data, making it difficult to maintain detection efficiency and control costs and model accuracy.

[0004] Therefore, how to improve the detection efficiency of songs with duplicate content is a key issue that technicians in this field are concerned about. Summary of the Invention

[0005] The purpose of this application is to provide a method, server and computer-readable storage medium for identifying duplicate songs to reduce the cost of identification while improving identification efficiency and accuracy.

[0006] To solve the above technical problems, the present application provides a method for identifying songs with duplicate content, comprising:

[0007] Perform song matching based on multiple audio clips of the song to be identified to obtain the target song;

[0008] Determining an offset time position of each audio segment in the target song; wherein the offset time position is used to represent a time point at which the audio segment matches the target song;

[0009] Calculating the difference between the to-be-identified distribution curve formed by all the offset time positions and the standard time distribution curve;

[0010] If the difference is greater than a threshold, the song to be identified is determined to be a song with duplicate content.

[0011] Optionally, determining the offset time position of each audio segment in the target song includes:

[0012] determining a plurality of time points at which each of the audio clips hits the target song;

[0013] Calculating a difference between each of the multiple time points and a corresponding multiple time points in the audio segment to obtain multiple time differences corresponding to each time point;

[0014] Adding the multiple time differences corresponding to each time point to obtain a cumulative time difference corresponding to each time point;

[0015] The time point with the largest cumulative time difference is used as the offset time position corresponding to the audio segment.

[0016] Optionally, taking the time point with the largest cumulative time difference as the offset time position corresponding to the audio segment includes:

[0017] constructing a histogram based on a plurality of the accumulated time differences;

[0018] A peak point in the histogram whose cumulative time difference is greater than a threshold is used as the offset time position corresponding to the audio segment.

[0019] Optionally, determining the offset time position of each audio segment in the target song includes:

[0020] determining a plurality of time points at which each of the audio clips hits the target song;

[0021] Determining whether there is a linear distribution portion between the multiple time points and corresponding time points in the audio segment;

[0022] If so, the starting time point of the linear distribution portion is used as the offset time position corresponding to the audio segment.

[0023] Optionally, calculating the difference between the to-be-identified distribution curve formed by all the offset time positions and the standard time distribution curve includes:

[0024] Determining a time point corresponding to each of the plurality of audio clips in the song to be identified;

[0025] Using the time point corresponding to each audio segment as the horizontal axis coordinate and the corresponding offset time position as the vertical axis coordinate, marking and connecting them in a rectangular coordinate system to obtain the distribution curve to be identified;

[0026] The difference between the to-be-identified distribution curve and the standard time distribution curve is calculated.

[0027] Optionally, calculating the difference between the to-be-identified distribution curve and the standard time distribution curve includes:

[0028] Calculating the mean Euclidean distance between the distribution curve to be identified and the standard time distribution curve;

[0029] The mean of the Euclidean distances is used as the degree of difference.

[0030] Optionally, calculating the difference between the to-be-identified distribution curve and the standard time distribution curve includes:

[0031] Calculating a mean similarity between the distribution curve to be identified and the standard time distribution curve;

[0032] The reciprocal of the mean similarity is used as the difference; wherein, the smaller the mean similarity is, the larger the reciprocal of the mean similarity is, and the greater the difference is.

[0033] Optionally, song matching is performed based on multiple audio clips of the song to be identified to obtain a target song, including:

[0034] Segmenting the audio of the song to be identified to obtain the multiple audio segments;

[0035] Extracting an audio fingerprint from each of the audio clips to obtain multiple audio fingerprints;

[0036] Matching the multiple audio fingerprints with the music library index to obtain multiple matching songs and corresponding matching degrees;

[0037] The matching song with the highest matching degree is used as the target song.

[0038] The present application also provides a server, comprising:

[0039] memory for storing computer programs;

[0040] A processor is used to implement the steps of the method for identifying duplicate songs when executing the computer program.

[0041] The present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the content duplicate song identification method as described above are implemented.

[0042] The present application provides a method for identifying songs with duplicate content, comprising: performing song matching based on multiple audio clips of a song to be identified to obtain a target song; determining the offset time position of each audio clip in the target song; wherein the offset time position is used to characterize the time point at which the audio clip matches the target song; calculating the difference between a distribution curve to be identified formed by all the offset time positions and a standard time distribution curve; if the difference is greater than a threshold, determining the song to be identified as a song with duplicate content.

[0043] The target song is obtained through fingerprint feature matching, and then the offset time position of each audio clip in the target song in the multiple audio clips in the song to be identified is determined. Finally, the difference between the distribution curve to be identified and the standard time distribution curve composed of all the offset time positions is calculated. If the difference is greater, it means that the content of the song to be identified does not completely match the target song, and it is determined to be a song with duplicate content, rather than using a model to identify songs with duplicate content, which improves the recognition efficiency of songs with duplicate content and reduces the cost of recognition.

[0044] The present application also provides a server and a computer-readable storage medium, which have the above beneficial effects and are not described in detail here. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0046] Figure 1 A flowchart of a first method for identifying duplicate songs provided in an embodiment of the present application;

[0047] Figure 2 A flowchart of a second method for identifying duplicate songs provided in an embodiment of the present application;

[0048] Figure 3 A schematic diagram of a first audio hit of the method for identifying duplicate songs provided in an embodiment of the present application;

[0049] Figure 4 A schematic diagram of a second audio hit of the method for identifying duplicate songs provided in an embodiment of the present application;

[0050] Figure 5 A flowchart of a third method for identifying duplicate songs provided in an embodiment of the present application;

[0051] Figure 6 A flowchart of a fourth method for identifying duplicate songs provided in an embodiment of the present application;

[0052] Figure 7 A schematic diagram of the first time position of the method for identifying duplicate songs provided in an embodiment of the present application;

[0053] Figure 8 A second time position diagram of the method for identifying songs with duplicate content provided by an embodiment of the present application;

[0054] Figure 9 A flowchart of a fifth method for identifying duplicate songs according to an embodiment of the present application;

[0055] Figure 10 A flowchart of a sixth method for identifying duplicate songs provided in an embodiment of the present application;

[0056] Figure 11 This application is a flowchart of the sixth method for identifying duplicate songs provided in an embodiment of the present application;

[0057] Figure 12 A schematic diagram of the structure of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] The core of this application is to provide a method, server and computer-readable storage medium for identifying songs with duplicate content, so as to reduce the cost of identification while improving the efficiency and accuracy of identification.

[0059] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0060] Related technologies use neural networks to annotate data on normal songs and songs with duplicate content, then train a model to identify duplicate songs. However, this training process relies on a large amount of annotated data, and the annotation costs and errors increase with the amount of data, making it difficult to maintain detection efficiency and control costs and model accuracy.

[0061] Therefore, the present application provides a method for identifying songs with duplicate content, which obtains a target song through fingerprint feature matching, then determines the offset time position of each audio clip in the target song in the song to be identified, and finally calculates the difference between the distribution curve to be identified and the standard time distribution curve composed of all the offset time positions. If the difference is greater, it means that the content in the song to be identified does not completely match the target song, and it is determined to be a song with duplicate content, rather than using a model to identify songs with duplicate content, thereby improving the recognition efficiency of songs with duplicate content and reducing the cost of recognition.

[0062] In order to avoid the recognition method of model detection, avoid the cost generated by training data, and improve the efficiency of identifying songs with duplicate content, a method for identifying songs with duplicate content can be provided to avoid detection through models.

[0063] The following describes a method for identifying songs with duplicate content provided by this application through an embodiment.

[0064] Please refer to Figure 1 , Figure 1 This is a flowchart of the first method for identifying duplicate songs provided in an embodiment of the present application.

[0065] In this embodiment, the method may include:

[0066] S101, performing song matching based on multiple audio clips of the song to be identified to obtain a target song;

[0067] This step aims to perform song matching based on multiple audio clips of the song to be identified to obtain a target song. This can be done by extracting fingerprint features from a database to determine the target song that is closest to the song to be identified.

[0068] In order to improve the efficiency of fingerprint feature extraction for songs to be identified and reduce the amount of data required for calculation, multiple audio clips are extracted from the songs to be identified in this step, and then fingerprint feature matching is performed based on the multiple audio clips.

[0069] The method for extracting multiple audio clips from the song to be identified can be based on the same time interval, randomly distributed time intervals, or a predetermined time interval array, where the elements in the array are used as the time points for extracting the audio clips. It can be seen that there is no single method for extracting audio clips in this step, and a suitable extraction method can be selected based on the actual application situation, and is not specifically limited here.

[0070] On the basis of extracting the audio clip, the fingerprint feature matching method can adopt any fingerprint feature matching method provided by the existing technology, which is not specifically limited here.

[0071] In addition, multiple similar songs can be matched in this step, and the song with the highest matching degree can be used as the target song.

[0072] S102, determining an offset time position of each of the multiple audio clips in the target song; wherein the offset time position is used to represent a time point at which the audio clip matches the target song;

[0073] Based on S101, this step aims to determine the offset time position of each of the multiple audio clips within the target song. In other words, it determines the time point at which each audio clip matches the target song. Since a song with duplicate content can be composed of, or at least includes, an audio segment within a normal song, the actual audio content within the song with duplicate content can be determined at a specific time point within the target song, i.e., the corresponding offset time position.

[0074] For example, if an audio clip between 3 and 5 seconds in the song to be identified is matched to between 15 and 17 seconds in the target song, or to a portion of that audio clip, then the audio content starting at 15 seconds can be matched. Therefore, the offset time position of the audio clip is determined to be 15 seconds.

[0075] Furthermore, due to the particularity of audio files, multiple offset time positions can be determined for a given audio clip within the target song. However, some offset time positions correspond to audio clips that only partially match the audio clip in the song to be identified, making them unqualified offset time positions. Therefore, it is necessary to determine the most accurate offset time position from among the multiple offset time positions.

[0076] The determination method can be based on the distribution of multiple time points matched in the audio clip. When the distribution is random and irregular, it is determined to be an inaccurate offset time position. Multiple consecutive time points cannot be connected to form a straight line segment, that is, a random and scattered distribution is presented. When it is in a regular linear distribution, it can be determined as an accurate offset time position. The regular distribution is that multiple time points can be connected to form a straight line segment. It can also be determined based on the statistical relationship between the multiple time points matched in the audio clip and the time point of the audio clip in the song to be identified.

[0077] S103, calculating the difference between the to-be-identified distribution curve formed by all offset time positions and the standard time distribution curve;

[0078] Based on S102, this step aims to calculate the difference between the distribution curve to be identified, which is composed of all offset time positions, and the standard time distribution curve. Among them, the distribution curve to be identified, which is composed of all offset time positions, represents the distribution of all time points that these audio clips match in the target song. If the song to be identified is a song with repeated content, the distribution of these time points cannot be matched with the standard time distribution curve of the song. If the song to be identified is not a song with repeated content, the distribution of the time points of the song to be identified is similar to that of the standard time distribution curve of the target song.

[0079] Furthermore, the distribution curve to be identified, formed by all offset time positions in this step, can refer to a distribution curve formed in a rectangular coordinate system using each audio clip's own time point and the corresponding offset time position as coordinate points. Correspondingly, the standard time distribution curve is the distribution curve formed when the song to be identified and the target song are perfectly matched, that is, the distribution curve formed in the coordinate system where x = y. Therefore, the standard time distribution curve is a straight line with the same horizontal and vertical coordinates.

[0080] The method for calculating the difference degree may be to calculate the distance between the two, or to calculate the similarity between the two, or to adopt any difference degree calculation method provided by the prior art, which is not specifically limited here.

[0081] S104: If the difference is greater than the threshold, the song to be identified is determined to be a song with duplicate content.

[0082] Based on S103, this step aims to determine that the song to be identified is a duplicate song if the difference is greater than a threshold. Thus, in this step, whether a song is a duplicate song is determined based on the size of the difference. The threshold can be set based on the technician's experience, statistical data on duplicate songs, or calculated based on historical differences.

[0083] In summary, this embodiment obtains the target song through fingerprint feature matching, then determines the offset time position of each audio clip in the target song in the multiple audio clips in the song to be identified, and finally calculates the difference between the distribution curve to be identified composed of all offset time positions and the standard time distribution curve. If the difference is greater, it means that the content of the song to be identified does not completely match the target song, and it is determined to be a song with duplicate content, rather than using a model to identify songs with duplicate content, which improves the recognition efficiency of songs with duplicate content and reduces the cost of recognition.

[0084] Furthermore, in order to improve the accuracy of matching target songs and determining the offset time position, the following further illustrates a method for identifying duplicate songs provided by the present application through another specific embodiment.

[0085] Please refer to Figure 2 , Figure 2 This is a flowchart of the second method for identifying duplicate songs provided in an embodiment of the present application.

[0086] In this embodiment, the method may include:

[0087] S201, performing song matching based on multiple audio clips of the song to be identified to obtain a target song;

[0088] S202, determining multiple time points at which each audio clip hits the target song;

[0089] S203, calculating the difference between each of the multiple time points and the corresponding multiple time points in the audio clip to obtain multiple time differences corresponding to each time point;

[0090] S204, adding up the multiple time differences corresponding to each time point to obtain the cumulative time difference corresponding to each time point;

[0091] S205, taking the time point with the largest accumulated time difference as the offset time position corresponding to the audio segment;

[0092] S206, calculating the difference between the distribution curve to be identified formed by all offset time positions and the standard time distribution curve;

[0093] S207: If the difference is greater than the threshold, the song to be identified is determined to be a song with duplicate content.

[0094] As can be seen, this optional solution primarily illustrates how to determine a more accurate offset time position through steps S202 to S205. This optional solution primarily calculates the cumulative time difference between each time point in the audio clip and the corresponding time point in the target song, and uses the time point with the largest cumulative time difference as the offset time position corresponding to the audio clip.

[0095] The threshold value can be set based on the technician's experience, or can be determined based on statistical data of songs with duplicate content, or can be determined based on calculated historical differences. When the threshold value is set based on the technician's experience, a difference analysis can be performed based on a large amount of sample data of songs with duplicate content, and the most accurate difference can be determined based on the analysis results as the threshold value.

[0096] If an audio clip cannot completely match the content of the corresponding target song, the time difference between each time point in the audio clip and the matching time point in the target song is irregularly distributed, and the difference between each group of time points has no regularity and is scattered.

[0097] If an audio clip completely matches the target song, that is, if there is an audio clip that completely matches a portion of the target song, the offset time of the matching portion is the exact offset time position. The time differences between each time point in the audio clip and the matching time point in the target song are distributed in a certain pattern, with the differences concentrated, resulting in the calculated cumulative time difference being much larger than the cumulative time differences at other time points.

[0098] Furthermore, in this embodiment, the time point with the largest cumulative time difference is used as the offset time position corresponding to the audio segment, including:

[0099] Step 1, constructing a histogram based on multiple accumulated time differences;

[0100] Step 2: The peak point in the histogram whose cumulative time difference is greater than the threshold is used as the offset time position corresponding to the audio segment.

[0101] As can be seen, this alternative solution primarily illustrates how to determine the offset time position based on the accumulated time difference. In this alternative solution, a histogram is constructed based on multiple accumulated time differences, and the peak points in the histogram where the accumulated time difference exceeds a threshold are used as the corresponding offset time position of the audio segment. In other words, the corresponding time point is determined from the constructed histogram.

[0102] Please refer to Figure 3 and Figure 4 , Figure 3 This is a schematic diagram of the first audio hit of the method for identifying duplicate songs provided in an embodiment of the present application. Figure 4 This is a schematic diagram of the second audio hit of the content duplicate song identification method provided in an embodiment of the present application.

[0103] Figure 3 The top middle figure shows a schematic diagram of audio fingerprint segment hits. The horizontal axis of the diagram represents the database audio fingerprint time point, which is the time point in the target song, and the vertical axis represents the request audio fingerprint time point, which is the time point in the audio segment of the song to be identified. The bottom figure shows a histogram of offset time position statistics. The horizontal axis of the histogram represents the time point in the target song in the database, and the vertical axis represents the cumulative time difference.

[0104] Figure 4 The top middle figure shows a schematic diagram of audio fingerprint segment hits. The horizontal axis of the diagram represents the database audio fingerprint time point, which is the time point in the target song, and the vertical axis represents the request audio fingerprint time point, which is the time point in the audio segment of the song to be identified. The bottom figure shows a histogram of offset time position statistics. The horizontal axis of the histogram represents the time point in the target song in the database, and the vertical axis represents the cumulative time difference.

[0105] Due to the similarity of audio fingerprints of segments, a continuous audio segment may hit different time points of the same song. In this case, the error of the returned offset time position is large.

[0106] by Figure 3 and Figure 4 For example, this embodiment draws a histogram by counting the cumulative time difference between the hit song time point and the requested segment time point, and selects the peak point in the histogram distribution that is greater than the threshold as the offset time position of the hit song segment. Figure 3 As shown, Figure 3Although a large number of audio fingerprint segments are hit, they are relatively scattered as shown in the histogram, and no significant peaks are generated, so it is impossible to return an accurate offset time position estimate. Figure 4 As shown, Figure 4 A peak significantly larger than the threshold appears at 30s, which means that the requested audio fingerprint is more likely to hit the song at the offset time position of 30s. This ensures the continuity of the audio fingerprint hitting the song segment and increases the accuracy of the offset time position of the target song returned by the audio fingerprint segment.

[0107] It can be seen that this embodiment calculates the cumulative time difference between time points, and then determines the most accurate offset time position based on the statistical characteristics of the cumulative time difference, and selects the most accurate time point from multiple time points as the offset time position, thereby improving the accuracy of the offset time position and further improving the accuracy of subsequent repeated content judgment.

[0108] Furthermore, in order to improve the accuracy of matching target songs and determining the offset time position, the following further illustrates a method for identifying duplicate songs provided by the present application through another specific embodiment.

[0109] Please refer to Figure 5 , Figure 5 This is a flowchart of the third method for identifying duplicate songs provided in an embodiment of the present application.

[0110] In this embodiment, the method may include:

[0111] S301, performing song matching based on multiple audio clips of the song to be identified to obtain a target song;

[0112] S302, determining multiple time points at which each audio clip hits the target song;

[0113] S303, determining whether there is a linear distribution portion between the multiple time points and the corresponding time points in the audio clip; if so, executing S304; if not, terminating the execution;

[0114] S304, taking the starting time point of the linear distribution portion as the offset time position corresponding to the audio segment;

[0115] S305, calculating the difference between the distribution curve to be identified formed by all offset time positions and the standard time distribution curve;

[0116] S306: If the difference is greater than the threshold, the song to be identified is determined to be a song with duplicate content.

[0117] As can be seen, this optional solution primarily illustrates how to determine a more accurate offset time position through steps S302 to S304. This optional solution primarily involves determining multiple time points at which each audio clip occurs in the target song and determining whether there is a linear distribution between these multiple time points and the corresponding time points in the audio clip. If a corresponding linear distribution exists, the starting time point of the linear distribution is used as the corresponding offset time position for the audio clip.

[0118] That is, determine whether there is a linear distribution part between multiple time points and the corresponding time points in the audio clip. If there is a linear distribution part, it means that the time position of the audio clip corresponding to the part in the target song is continuous, and the starting position of the part can be used as the offset time position of the audio clip.

[0119] It can be seen that this embodiment determines whether there is a linear distribution part between multiple time points and the corresponding time points in the audio clip. When a linear distribution part appears, it indicates that there are continuous time points. The time of the linear distribution part can be used as the offset time position of the audio clip, thereby improving the accuracy of determining the time position.

[0120] Furthermore, in order to accurately evaluate the difference between the distribution curve to be identified composed of all offset time positions and the standard time distribution curve, the following further illustrates a method for identifying songs with repeated content provided by the present application through another specific embodiment.

[0121] Please refer to Figure 6 , Figure 6 This is a flowchart of the fourth method for identifying duplicate songs provided in an embodiment of the present application.

[0122] In this embodiment, the method may include:

[0123] S401, performing song matching based on multiple audio clips of the song to be identified to obtain a target song;

[0124] S402, determining an offset time position of each audio segment in the target song among the multiple audio segments;

[0125] S403, determining a time point corresponding to each of the multiple audio clips in the song to be identified;

[0126] S404: Using the time point corresponding to each audio clip as the horizontal axis coordinate and the corresponding offset time position as the vertical axis coordinate, the audio clips are marked and connected in a rectangular coordinate system to obtain a distribution curve to be identified;

[0127] S405, calculating the mean Euclidean distance between the distribution curve to be identified and the standard time distribution curve;

[0128] S406: Taking the mean of the Euclidean distance as the degree of difference.

[0129] S407: If the difference is greater than the threshold, the song to be identified is determined to be a song with duplicate content.

[0130] As can be seen, this optional solution mainly uses S403 to S405 to accurately evaluate the difference between the time distributions of the two. In this optional solution, the time point corresponding to each audio clip in the song to be identified is determined; the time point corresponding to each audio clip is used as the horizontal axis coordinate, and the corresponding offset time position is used as the vertical axis coordinate. These are identified and connected in a rectangular coordinate system to obtain a distribution curve to be identified; the mean Euclidean distance between the distribution curve to be identified and the standard time distribution curve is calculated, and the mean Euclidean distance is used as the difference.

[0131] Please refer to Figure 7 , Figure 7 This is a schematic diagram of the first time position of the method for identifying songs with repeated content provided in an embodiment of the present application. Figure 7 Specifically, it is a schematic diagram of the normal song offset time position, wherein the horizontal axis is the test audio time offset position, that is, the time point of the song to be identified, and the vertical axis is the hit song ID time offset position, that is, the time point of the target song.

[0132] Please refer to Figure 8 , Figure 8 This is a second time position diagram of the method for identifying songs with repeated content provided in an embodiment of the present application. Figure 8 Specifically, it is a schematic diagram of the time position of the content-duplicate song offset, wherein the horizontal axis is the test audio time offset position, that is, the time point of the song to be identified, and the vertical axis is the time offset position of the hit song ID, that is, the time point of the target song.

[0133] That is, first determine the time distribution between the offset time position and the time point corresponding to the audio segment (i.e., the distribution curve to be identified), that is, determine the distribution of the offset time position and the time point in the coordinate system. Among them, the offset time position of each audio fingerprint segment of the normal song is similar to the offset time position of the target song, and the overall distribution shows y=x, as shown in Figure 7 On the contrary, due to the presence of a large number of repeated audio segments in songs with duplicate content, the offset time positions of different audio fingerprint segments cannot be consistent with the offset time positions in the target song, and the overall distribution deviates from y = x, as shown in Figure 8 The method of the present invention calculates the mean of the Euclidean distance between the distribution curve to be identified at the offset time position and the straight line y=x, and compares it with a threshold value to characterize the difference between the distribution curve to be identified at the offset time position and the straight line y=x. If the difference is greater than the threshold value, the song is determined to have duplicate content.

[0134] It can be seen that this embodiment evaluates the difference by using the Euclidean distance, thereby improving the accuracy of the difference evaluation and maintaining high accuracy in identifying the songs to be identified.

[0135] Furthermore, in order to accurately evaluate the difference between the distribution curve to be identified composed of all offset time positions and the standard time distribution curve, the following further illustrates a method for identifying songs with repeated content provided by the present application through another specific embodiment.

[0136] Please refer to Figure 9 , Figure 9 This is a flowchart of the fifth method for identifying duplicate songs provided in an embodiment of the present application.

[0137] In this embodiment, the method may include:

[0138] S501, performing song matching based on multiple audio clips of the song to be identified to obtain a target song;

[0139] S502, determining an offset time position of each audio segment in the target song among the multiple audio segments;

[0140] S503, determining a time point corresponding to each of the multiple audio clips in the song to be identified;

[0141] S504: Using the time point corresponding to each audio clip as the horizontal axis coordinate and the corresponding offset time position as the vertical axis coordinate, the audio clips are marked and connected in a rectangular coordinate system to obtain a distribution curve to be identified;

[0142] S505, calculating the mean similarity between the distribution curve to be identified and the standard time distribution curve;

[0143] S506: The reciprocal of the mean similarity value is used as the difference.

[0144] S507: If the difference is greater than the threshold, the song to be identified is determined to be a song with duplicate content.

[0145] As can be seen, this optional solution primarily involves steps S503 to S505 to accurately assess the difference between the time distributions of the two. In this optional solution, the corresponding time point of each of the multiple audio clips in the song to be identified is determined. The corresponding time point of each audio clip is used as the horizontal axis coordinate, and the corresponding offset time position is used as the vertical axis coordinate. These are then labeled and connected in a rectangular coordinate system to obtain the distribution curve to be identified. The mean similarity between the distribution curve to be identified and the standard time distribution curve is calculated. The smaller the mean similarity, the greater the reciprocal of the mean similarity, and the greater the difference. In other words, the similarity between the time distribution and the standard time distribution curve is calculated, and the reciprocal of the similarity is used as the difference. The more similar the distribution curve to be identified is to the standard time distribution curve, and the greater the similarity, the lower the probability that the song to be identified is a duplicate song, meaning the smaller the difference. The less similar the standard time distribution curve is to the standard time distribution curve, and the smaller the similarity, the greater the probability that the song to be identified is a duplicate song, meaning the greater the difference.

[0146] The method for calculating the similarity may adopt any similarity calculation method provided by the prior art, and is not specifically limited here.

[0147] It can be seen that this embodiment improves the accuracy of the evaluation and further improves the accuracy of recognition by evaluating the difference between the song to be identified and the target song through similarity and then making a judgment.

[0148] Furthermore, in order to improve the accuracy of matching target songs, the method for identifying songs with duplicate content provided by this application is further described below through another specific embodiment.

[0149] Please refer to Figure 10 , Figure 10 This is a flowchart of the sixth method for identifying duplicate songs provided in an embodiment of the present application.

[0150] In this embodiment, the method may include:

[0151] S601, segmenting the audio of the song to be recognized to obtain multiple audio clips;

[0152] S602, extracting audio fingerprints from each audio clip to obtain multiple audio fingerprints;

[0153] S603, matching multiple audio fingerprints with the music library index to obtain multiple matching songs and corresponding matching degrees; and selecting the matching song with the highest matching degree as the target song;

[0154] S604, determining an offset time position of each audio segment in the target song among the multiple audio segments;

[0155] S605, calculating the difference between the to-be-identified distribution curve formed by all offset time positions and the standard time distribution curve;

[0156] S606: If the difference is greater than the threshold, the song to be identified is determined to be a song with duplicate content.

[0157] As can be seen, this optional solution primarily improves the accuracy and precision of matching the target song through steps S601 to S603. In this optional solution, the song to be identified is segmented to obtain multiple audio segments. Audio fingerprints are extracted from each audio segment to obtain multiple audio fingerprints. Song matching is performed based on these multiple audio fingerprints to obtain the target song. As can be seen, by using the audio fingerprints corresponding to multiple audio segments for song matching in this embodiment, the accuracy and precision of identifying the target song can be improved.

[0158] The matching degree may be obtained by any matching degree obtaining method provided by the prior art, and is not specifically limited here.

[0159] Please refer to Figure 11 , Figure 11 This application is a flowchart of the sixth method for identifying duplicate songs provided in an embodiment of the present application.

[0160] For example, in this embodiment, song matching can be performed based on Landmark's song recognition system.

[0161] The audio fingerprint used in this embodiment is the Landmark audio fingerprint, such as Figure 11 As shown. First, the time domain audio clip is transformed into the frequency domain through Fourier transform, as shown Figure 11 As shown in the 1A spectrogram (Spectrogram), the horizontal axis represents the time index and the vertical axis represents the frequency index; then the local peak of the time-frequency point signal is extracted, as shown in Figure 11 As shown in the 1B-Constellation Map (constellation map); then find the remaining peak points around the local peak point to form a combination (t1, f1, t2, f2), that is, the time-frequency point (t1, f1) and the time-frequency point (t2, f2) both have peak points, such as Figure 11 As shown in 1C-CombinatorialHash Generation (combinatorial hash generation diagram); finally, further simplify the information and retain only (f1, f2, t2-t1), as shown in Figure 11 As shown in the 1D-Hash details (hash details diagram).

[0162] The extracted audio fingerprint is then used to construct a music library index through a hash table. Finally, after the audio fingerprint is extracted from the audio clip, it is matched with the music library index and the song ID with the highest matching score and the offset time position of the hit are returned, thus obtaining the target song and the corresponding offset time position.

[0163] It can be seen that this embodiment obtains the target song through fingerprint feature matching, and then determines the offset time position of each audio clip in the target song in the multiple audio clips in the song to be identified, and finally calculates the difference between the distribution curve to be identified composed of all offset time positions and the standard time distribution curve. If the difference is greater, it means that the content of the song to be identified does not completely match the target song, and it is determined to be a song with duplicate content, rather than using a model to identify songs with duplicate content, which improves the recognition efficiency of songs with duplicate content and reduces the cost of recognition.

[0164] Please refer to Figure 12 , Figure 12 A schematic diagram of the structure of a server provided in an embodiment of the present application.

[0165] The embodiment of the present application further provides a server, characterized by comprising:

[0166] Memory 100, for storing computer programs;

[0167] The processor 200 is configured to implement the steps of the method for identifying duplicate songs as described in the above embodiment when executing the computer program.

[0168] An embodiment of the present application also provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the content duplicate song identification method as described in the above embodiment are implemented.

[0169] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0170] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0171] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0172] The above describes in detail a method for identifying duplicate songs, a server, and a computer-readable storage medium provided by the present application. This document uses specific examples to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only intended to help understand the method and core concept of the present application. It should be noted that, for those skilled in the art, without departing from the principles of the present application, several improvements and modifications may be made to the present application, and such improvements and modifications also fall within the scope of protection of the claims of the present application.

Claims

1. A method for identifying songs with duplicate content, characterized in that: include: Perform song matching based on multiple audio clips of the song to be identified to obtain the target song; Determining an offset time position of each audio segment in the target song; wherein the offset time position is used to represent a time point at which the audio segment matches the target song; Calculating the difference between the distribution curve to be identified formed by all the offset time positions and the standard time distribution curve; wherein the distribution curve to be identified is formed in a rectangular coordinate system by using the time point of each audio segment itself and the corresponding offset time position as coordinate points; If the difference is greater than a threshold, the song to be identified is determined to be a song with duplicate content.

2. The method for identifying duplicate songs according to claim 1, wherein: Determining an offset time position of each audio segment in the target song includes: determining a plurality of time points at which each of the audio clips hits the target song; Calculating a difference between each of the multiple time points and a corresponding multiple time points in the audio segment to obtain multiple time differences corresponding to each time point; Adding the multiple time differences corresponding to each time point to obtain a cumulative time difference corresponding to each time point; The time point with the largest cumulative time difference is used as the offset time position corresponding to the audio segment.

3. The method for identifying duplicate songs according to claim 2, wherein: The time point with the largest cumulative time difference is used as the offset time position corresponding to the audio segment, including: constructing a histogram based on a plurality of the accumulated time differences; A peak point in the histogram whose cumulative time difference is greater than a threshold is used as the offset time position corresponding to the audio segment.

4. The method for identifying duplicate songs according to claim 1, wherein: Determining an offset time position of each audio segment in the target song includes: determining a plurality of time points at which each of the audio clips hits the target song; Determining whether there is a linear distribution portion between the multiple time points and corresponding time points in the audio segment; If so, the starting time point of the linear distribution portion is used as the offset time position corresponding to the audio segment.

5. The method for identifying duplicate songs according to any one of claims 1 to 4, characterized in that: Calculating the difference between the distribution curve to be identified formed by all the offset time positions and the standard time distribution curve, including: Determining a time point corresponding to each of the plurality of audio clips in the song to be identified; Using the time point corresponding to each audio segment as the horizontal axis coordinate and the corresponding offset time position as the vertical axis coordinate, marking and connecting them in a rectangular coordinate system to obtain the distribution curve to be identified; The difference between the to-be-identified distribution curve and the standard time distribution curve is calculated.

6. The method for identifying duplicate songs according to claim 5, wherein: Calculating the difference between the to-be-identified distribution curve and the standard time distribution curve includes: Calculating the mean Euclidean distance between the distribution curve to be identified and the standard time distribution curve; The mean of the Euclidean distances is used as the degree of difference.

7. The method for identifying duplicate songs according to claim 5, wherein: Calculating the difference between the to-be-identified distribution curve and the standard time distribution curve includes: Calculating a mean similarity between the distribution curve to be identified and the standard time distribution curve; The reciprocal of the mean similarity is used as the difference; wherein, the smaller the mean similarity is, the larger the reciprocal of the mean similarity is, and the greater the difference is.

8. The method for identifying duplicate songs according to claim 5, wherein: Song matching is performed based on multiple audio clips of the song to be identified to obtain the target song, including: Segmenting the audio of the song to be identified to obtain the multiple audio segments; Extracting an audio fingerprint from each of the audio clips to obtain multiple audio fingerprints; Matching the multiple audio fingerprints with the music library index to obtain multiple matching songs and corresponding matching degrees; The matching song with the highest matching degree is used as the target song.

9. A server, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the method for identifying duplicate songs according to any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for identifying duplicate songs according to any one of claims 1 to 8.

Images