Chorus audio segmentation method, computer equipment, storage medium and program product

Abstract

The invention relates to a chorus audio segmentation method, computer equipment, a storage medium and a program product. The method comprises the following steps: separating a pure human voice audio from a target chorus audio, and dividing the pure human voice audio into a plurality of audio clips; determining timbre features of each audio clip through a pre-trained timbre feature extraction model; determining an audio segmentation result of the target chorus audio according to the similarity between the timbre features of the audio clips; the audio segmentation result represents audio clips corresponding to sentences sung by different singers in the target chorus audio. By adopting the method, the timbre characteristics can be used as the basis of audio segmentation, the similarity degree of each segment in the chorus audio in the timbre dimension is analyzed, the segments sung by different singers in the chorus audio are accurately distinguished, and the accuracy of the chorus audio segmentation is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a chorus audio segmentation method, apparatus, computer equipment, computer-readable storage medium and computer program product. Background Art

[0002] Users can sing songs online through the Karaoke application, record and share audio or video of their own singing. Users can choose songs from a rich music library to sing, and improve their singing skills and singing effects based on the real-time scores generated by the application for their singing. At the same time, the Karaoke application also provides a song chorus function, where users can sing the same song with other users, and jointly create and share chorus works.

[0003] Traditional technology mainly segments audio by analyzing the audio structure of music, dividing the audio into structures such as prelude, verse, chorus, etc., so that users can sing in segments. However, when users sing in a chorus, using traditional technology to segment the audio cannot accurately distinguish the singing segments of different users, and there is a problem of confusing the singing segments of different users, which is not conducive to improving the accuracy of chorus audio segmentation. Summary of the invention

[0004] Based on this, it is necessary to provide a chorus audio segmentation method, device, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of chorus audio segmentation in response to the above technical problems.

[0005] In a first aspect, the present application provides a chorus audio segmentation method, comprising:

[0006] Separating pure vocal audio from the target chorus audio, and dividing the pure vocal audio into a plurality of audio segments;

[0007] Determining the timbre feature of each of the audio clips by using a pre-trained timbre feature extraction model;

[0008] The audio segmentation result of the target chorus audio is determined according to the similarity between the timbre features of each of the audio segments; the audio segmentation result represents the audio segments corresponding to the sentences sung by different singers in the target chorus audio.

[0009] In one embodiment, determining the audio segmentation result of the target chorus audio according to the similarity between the timbre features of the audio segments includes:

[0010] According to the similarity between the timbre features of the audio clips, the audio clips are divided into at least one audio clip set; the similarity between the timbre features of the audio clips in the same audio clip set is greater than or equal to a preset similarity threshold;

[0011] An audio segmentation result of the target chorus audio is determined according to each of the audio segment sets.

[0012] In one of the embodiments, dividing the audio segments into at least one audio segment set according to the similarity between the timbre features of the audio segments includes:

[0013] For any audio segment among the audio segments, an audio segment whose similarity with the timbre feature of the any audio segment is greater than or equal to a preset similarity threshold is selected from the audio segments as an audio segment matching the any audio segment;

[0014] An audio segment set is constructed according to the any audio segment and the audio segment matching the any audio segment.

[0015] In one of the embodiments, dividing the audio segments into at least one audio segment set according to the similarity between the timbre features of the audio segments includes:

[0016] For any audio segment among the audio segments, if an audio segment whose degree of similarity with the timbre feature of the any audio segment is greater than or equal to a preset similarity threshold cannot be screened out from the audio segments, an audio segment set is constructed based on the any audio segment.

[0017] In one of the embodiments, determining the audio segmentation result of the target chorus audio according to each of the audio segment sets includes:

[0018] For each of the audio segment sets, taking the audio segments in the audio segment set as the singing segments of the singer matching the audio segment set;

[0019] The singing segments of each of the singers are used as the audio segmentation results of the target chorus audio.

[0020] In one embodiment, the method further comprises:

[0021] Acquire length information of the timbre feature of each of the audio segments, and determine the similarity between the timbre features of each of the audio segments according to the dot product between the timbre features of each of the audio segments and the length information; or

[0022] The similarity between the timbre features of the audio segments is determined according to the distances between the timbre features of the audio segments.

[0023] In one embodiment, dividing the pure human voice audio into a plurality of audio segments includes:

[0024] Obtaining the start time of each lyric sentence in the target chorus audio;

[0025] The pure human voice audio is divided according to the start time to obtain the plurality of audio segments.

[0026] In a second aspect, the present application further provides a chorus audio segmentation device, comprising:

[0027] A separation module, used to separate pure vocal audio from the target chorus audio, and divide the pure vocal audio into a plurality of audio segments;

[0028] A determination module, used to determine the timbre feature of each of the audio clips through a pre-trained timbre feature extraction model;

[0029] The segmentation module is used to determine the audio segmentation result of the target chorus audio according to the similarity between the timbre features of each of the audio segments; the audio segmentation result represents the audio segments corresponding to the sentences sung by different singers in the target chorus audio.

[0030] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the above method are implemented.

[0031] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0032] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0033] The above-mentioned chorus audio segmentation method, device, computer equipment, computer-readable storage medium and computer program product separate pure vocal audio from the target chorus audio to avoid interference of the accompaniment on the audio segmentation, divide the pure vocal audio into several audio segments, and determine the timbre characteristics of each audio segment. By analyzing the similarity of each audio segment in the timbre characteristic dimension, the audio segments corresponding to the sentences sung by different singers in the target chorus audio are determined. The timbre characteristics can be used as the basis for audio segmentation, and the similarity of each segment in the chorus audio in the timbre dimension can be analyzed to accurately distinguish the segments sung by different singers in the chorus audio, thereby improving the accuracy of chorus audio segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0035] Figure 1 A diagram of an application environment of a chorus audio segmentation method in one embodiment;

[0036] Figure 2 A schematic flow chart of a chorus audio segmentation method in one embodiment;

[0037] Figure 3 A schematic diagram of a process for determining a timbre feature in one embodiment;

[0038] Figure 4 A schematic diagram of determining the similarity between timbre features in one embodiment;

[0039] Figure 5 is a structural block diagram of a chorus audio segmentation device in one embodiment;

[0040] Figure 6 The figure is a diagram of the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0042] The chorus audio segmentation method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal 102 separates the pure human voice audio from the target chorus audio, and divides the pure human voice audio into several audio segments; the terminal 102 determines the timbre features of each audio segment through a pre-trained timbre feature extraction model; the terminal 102 determines the audio segmentation result of the target chorus audio according to the similarity between the timbre features of each audio segment; the audio segmentation result represents the audio segments corresponding to the sentences sung by different singers in the target chorus audio. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices, and the Internet of Things devices can be smart speakers, smart TVs, smart car-mounted devices, etc. Portable wearable devices can be smart watches, head-mounted devices, etc. The head mounted device may be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0043] In an exemplary embodiment, Figure 2 As shown, a chorus audio segmentation method is provided, and the method is applied to a terminal as an example for explanation. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server, including the following steps S202 to S206. Among them:

[0044] Step S202, separate the pure vocal audio from the target chorus audio, and divide the pure vocal audio into a plurality of audio segments.

[0045] Among them, the target chorus audio may refer to the audio obtained by at least two singers singing a specific song together. In actual applications, each segment in the target chorus audio may be sung by a single singer alone or by multiple singers together.

[0046] Among them, pure vocal audio may refer to pure vocals without music. In practical applications, the target chorus audio may be separated into pure vocal audio and pure accompaniment audio by a pre-trained vocal / accompaniment separation deep learning model.

[0047] The audio clips may refer to audios that can together constitute the target chorus audio. In practical applications, the number of audio clips may be the same as the number of lyrics of the target chorus audio.

[0048] As an example, when the traditional technology uses music audio structured segmentation to segment the audio, it analyzes the structure of the song and separates the prelude, main song, chorus, etc. The accuracy is insufficient and the segmentation results are prone to errors. This error has little effect on the song singing when the song is sung by a single singer alone / solo songs. However, when multiple singers sing a song together, the traditional technology is still used to segment the song, and obvious segmentation errors will occur, such as dividing the lyrics / fragments sung by different singers into lyrics / fragments sung by the same singer, which affects the segmentation accuracy of the chorus audio. When the chorus audio segmentation method is used to segment the chorus audio, the terminal can first obtain the audio obtained by singing a song by multiple singers as the target chorus audio, and then separate the pure vocal audio from the target chorus audio through a pre-trained vocal accompaniment separation model (such as a vocal / accompaniment separation deep learning model) to avoid the accompaniment interfering with the segmentation result. After that, the terminal can segment the pure vocal audio according to the lyrics information of the target chorus audio and divide the pure vocal audio into several audio segments. It is understandable that the terminal may first segment the target chorus audio according to the lyrics information of the target chorus audio, and separate the pure vocal audio through a pre-trained vocal accompaniment separation model (such as a vocal / accompaniment separation deep learning model) to obtain several audio clips.

[0049] Step S204: determining the timbre feature of each audio clip by using a pre-trained timbre feature extraction model.

[0050] Among them, the timbre feature extraction model may refer to a model used to determine the timbre features of an audio clip / audio. In practical applications, the timbre feature extraction model may include a deep learning model for extracting timbre features of singing voices.

[0051] The timbre feature may refer to information used to characterize or distinguish the timbre of different audio clips / different audios. In practical applications, the timbre feature may include but is not limited to a timbre feature vector (such as a timbre embedding).

[0052] As an example, timbre features can be used to characterize audio clips / audio. Therefore, in order to accurately segment the audio, the terminal can input each audio clip into a pre-trained timbre feature extraction model. The timbre feature extraction model can analyze the audio clip and determine the timbre features of the audio clip.

[0053] Step S206, determining the audio segmentation result of the target chorus audio according to the similarity between the timbre features of each audio segment; the audio segmentation result represents the audio segments corresponding to the sentences sung by different singers in the target chorus audio.

[0054] Among them, the audio segmentation result may refer to information representing audio segments corresponding to sentences sung by different singers in the target chorus audio.

[0055] As an example, the terminal can compare the timbre features of each audio clip in pairs to determine the similarity between the timbre features of each audio clip. If two audio clips are sung by the same singer, the similarity between the timbre features of the two audio clips will be greater (such as closer to 1). If the two audio clips are sung by different singers, the similarity between the timbre features of the two audio clips will be smaller (such as closer to 0). Therefore, the terminal can analyze whether the singers of each audio clip are the same based on the similarity between the timbre features of each audio clip, thereby determining the audio segmentation result of the target chorus audio.

[0056] In the above-mentioned chorus audio segmentation method, pure vocal audio is separated from the target chorus audio to avoid interference of the accompaniment on the audio segmentation, the pure vocal audio is divided into several audio segments, and the timbre characteristics of each audio segment are determined. By analyzing the similarity of each audio segment in the timbre feature dimension, the audio segments corresponding to the sentences sung by different singers in the target chorus audio are determined. The timbre characteristics can be used as the basis for audio segmentation, and the similarity of each segment in the chorus audio in the timbre dimension can be analyzed. The segments sung by different singers in the chorus audio can be accurately distinguished, thereby improving the accuracy of chorus audio segmentation.

[0057] In an exemplary embodiment, the audio segmentation result of the target chorus audio is determined according to the similarity between the timbre features of the audio clips, including: dividing the audio clips into at least one audio clip set according to the similarity between the timbre features of the audio clips; the similarity between the timbre features of the audio clips in the same audio clip set is greater than or equal to a preset similarity threshold; and determining the audio segmentation result of the target chorus audio according to the audio clip sets.

[0058] The audio clip set may refer to a set consisting of a certain number of audio clips. In practical applications, the number of the audio clip set may be greater than or equal to the number of singers of the target chorus audio.

[0059] Among them, the preset similarity threshold may refer to pre-set information used to determine whether the singers of two audio clips are the same. In actual applications, the preset similarity threshold may include but is not limited to 0.2, and the preset similarity threshold can be flexibly adjusted based on actual needs.

[0060] As an example, after comparing the timbre features of each audio clip in pairs and determining the degree of similarity between the timbre features of each audio clip, the terminal can compare the degree of similarity between the timbre features of any two audio clips and a preset similarity threshold. When the degree of similarity between the timbre features of two audio clips is greater than or equal to the preset similarity threshold, the terminal can determine that the two audio clips are audio sung by the same singer. The terminal can take the set of audio clips sung by the same singer as an audio clip set. After the terminal analyzes the degree of similarity between the timbre features of all audio clips and the preset similarity threshold, the terminal can obtain several audio clip sets. The terminal can take the audio clips sung by the singers corresponding to each audio clip set as the audio segmentation result of the target chorus audio.

[0061] In this embodiment, by analyzing the similarity between the timbre features of the audio clips, the audio clips are divided into at least one audio clip set, and the audio segmentation results of the target chorus audio are determined according to each audio clip set. Audio clips with a similarity greater than or equal to a preset similarity threshold can be divided into the same audio clip set, and accurate audio segmentation results can be determined in combination with each audio clip set to achieve accurate audio segmentation.

[0062] In some embodiments, each audio segment is divided into at least one audio segment set according to the similarity between the timbre features of each audio segment, including: for any audio segment in each audio segment, selecting from each audio segment an audio segment whose similarity with the timbre features of any audio segment is greater than or equal to a preset similarity threshold as an audio segment that matches the any audio segment; and constructing an audio segment set according to any audio segment and the audio segment that matches the any audio segment.

[0063] As an example, the terminal can take any audio segment among the audio segments as the audio segment to be analyzed, the terminal can add a classification identifier for the audio segment to be analyzed, and then compare the timbre characteristics of each audio segment with the timbre characteristics of the audio segment to be analyzed one by one. When the degree of similarity between the timbre characteristics of an audio segment and the timbre characteristics of the audio segment to be analyzed is greater than or equal to a preset similarity threshold, the terminal can determine that the audio segment and the audio segment to be analyzed are sung by the same singer, and the terminal can add a corresponding classification identifier for the audio segment, which is the same as the classification identifier of the audio segment to be analyzed. After the timbre characteristics of each audio segment are compared with the timbre characteristics of the audio segment to be analyzed one by one, each audio segment has a corresponding classification identifier, and the terminal can take a set of audio segments with the same classification identifier as an audio segment set, and the classification identifier can represent the singer of the audio segment in the audio segment set.

[0064] In this embodiment, by screening audio clips in each audio clip whose degree of similarity with the timbre features of a specific audio clip is greater than or equal to a preset similarity threshold and constructing an audio clip set, the similarity between the timbre features of the audio clips can be accurately analyzed, ensuring that the degree of similarity between the timbre features of the audio clips in the same audio clip set is greater than or equal to the preset similarity threshold, thereby improving the accuracy of the audio clip set, so that accurate audio segmentation can be performed in combination with the audio clip set later.

[0065] In some embodiments, each audio segment is divided into at least one audio segment set according to the similarity between the timbre features of each audio segment, including: for any audio segment among the audio segments, if an audio segment whose similarity between the timbre features of any audio segment is greater than or equal to a preset similarity threshold cannot be screened out from the audio segments, an audio segment set is constructed based on any audio segment.

[0066] As an example, the terminal can take any audio segment among the audio segments as the audio segment to be analyzed, the terminal can add a classification identifier for the audio segment to be analyzed, and then compare the timbre characteristics of each audio segment with the timbre characteristics of the audio segment to be analyzed one by one. When the degree of similarity between the timbre characteristics of the audio segments other than the audio segment to be analyzed itself in each audio segment and the timbre characteristics of the audio segment to be analyzed is less than a preset similarity threshold, the terminal can determine that the singers of the audio segments other than the audio segment to be analyzed itself in each audio segment are different from the singers of the audio segment to be analyzed, and the terminal can treat the audio segments to be analyzed separately as an audio segment set. After the timbre characteristics of each audio segment are compared with the timbre characteristics of the audio segment to be analyzed one by one, each audio segment has a corresponding classification identifier, and the terminal can treat the set of audio segments with the same classification identifier as an audio segment set, and the classification identifier can represent the singers of the audio segments in the audio segment set.

[0067] In this embodiment, when the degree of similarity between the timbre features of an audio clip and the timbre features of each audio clip other than the audio clip itself is less than a preset similarity threshold, an audio clip set is constructed using the audio clip to avoid incorrect classification of the audio clips and improve the accuracy of the audio clip set, so that the audio clip set can be subsequently combined for accurate audio segmentation.

[0068] In some embodiments, based on each audio clip set, an audio segmentation result of the target chorus audio is determined, including: for each audio clip set, taking the audio clips in the audio clip set as the singing clips of the singers matching the audio clip set; and taking the singing clips of each singer as the audio segmentation result of the target chorus audio.

[0069] Among them, the singer matching the audio clip set may refer to the singer of each audio clip in the audio clip set. In actual applications, since the similarity between the timbre characteristics of each audio clip in the same audio clip set is greater than or equal to the preset similarity threshold, it can be determined that each audio clip in the same audio clip set is sung by the same singer, that is, the singer matching the audio clip set may be the singer of each audio clip in the audio clip set, and the singer may include users of the K song platform.

[0070] As an example, if two audio clips are sung by the same singer, the similarity between the timbre features of the two audio clips will be greater (such as closer to 1), and the similarity between the timbre features of each audio clip in the same audio clip set is greater than or equal to a preset similarity threshold. Therefore, the terminal can determine that each audio clip in the same audio clip set is sung by the same singer. For each audio clip set, the terminal can regard the singer of each audio clip in the audio clip set as the singer matching the audio clip set. Further, each audio clip in the audio clip set can be used as a singing clip of the singer matching the audio clip set, and the terminal can regard the singing clip of each singer as the audio segmentation result of the target chorus audio.

[0071] In this embodiment, by taking the audio clips in the audio clip set as the singing clips of the singers matching the audio clip set, and taking the singing clips of each singer as the audio segmentation results of the target chorus audio, it is possible to accurately distinguish the clips sung by each singer in the target chorus audio based on the audio clip set, obtain accurate audio segmentation results, and improve the accuracy of audio segmentation.

[0072] In some embodiments, the above method also includes: obtaining length information of the timbre features of each audio clip, and determining the similarity between the timbre features of each audio clip based on the dot product and length information between the timbre features of each audio clip; or determining the similarity between the timbre features of each audio clip based on the distance between the timbre features of each audio clip.

[0073] The length information may refer to information representing the length of the timbre feature. In practical applications, the length information may include the modulus of the timbre feature vector.

[0074] The distance between timbre features may include but is not limited to Euclidean distance, Manhattan distance, Chebyshev distance, etc.

[0075] As an example, the similarity between the timbre features of each audio clip may include cosine similarity information and Euclidean distance, etc.; when the similarity between the timbre features of each audio clip is cosine similarity information, the terminal may first obtain the length information of the timbre features of each audio clip, and then calculate the dot product between the timbre features of each audio clip, and then the terminal may calculate the ratio between the dot product between the timbre features of any two audio clips and the product of the length information of the timbre features of the two audio clips, determine the cosine similarity information between the timbre features of the two audio clips, and thus determine the similarity between the timbre features of the two audio clips. The terminal may also calculate the Euclidean distance / Manhattan distance / Chebyshev distance, etc. between the timbre features of each audio clip, and use the calculated Euclidean distance / Manhattan distance / Chebyshev distance, etc. as the similarity between the timbre features of each audio clip.

[0076] In this embodiment, by analyzing the cosine similarity or distance between the timbre features, the similarity between the timbre features is determined, the similarity between the timbre features can be accurately calculated, the accuracy of the similarity between the timbre features is improved, and a data basis is provided for the subsequent construction of an audio clip set and audio segmentation.

[0077] In some embodiments, the pure vocal audio is divided into a number of audio segments, including: obtaining the start time of each lyric sentence in the target chorus audio; and dividing the pure vocal audio according to the start time to obtain a number of audio segments.

[0078] The lyrics sentence may refer to information representing each sentence / each lyric in the lyrics information of the target chorus audio. In practical applications, the lyrics information of the target chorus audio may include several sentences / several lyrics.

[0079] Among them, the start time can refer to the information of the starting position of the lyrics sentence in the target chorus audio. For example, when a line of lyrics appears between 1 minute 20 seconds and 1 minute 25 seconds of the target chorus audio, the start time of the line of lyrics can be 1 minute 20 seconds of the target chorus audio.

[0080] As an example, when singers sing a song together to generate the target chorus audio, different singers can sing different lyrics, so the terminal can divide the pure vocal audio of the target chorus audio into several audio segments according to the lyrics information of the target chorus audio. Specifically, the terminal can first obtain the lyrics information of the target chorus audio, and then the terminal can use the starting time of each sentence of lyrics as a segmentation point, and divide / segment the pure vocal audio according to the segmentation point to obtain several audio segments. It is understandable that the terminal can also obtain the starting time of each lyric sentence in the target chorus audio, segment the target chorus audio according to the starting time, and extract the pure vocal audio to obtain several audio segments.

[0081] In this embodiment, by segmenting the audio based on the start time of the lyrics, accurate audio segmentation can be performed according to the start time of the lyrics, so as to subsequently analyze the similarity between the timbre features of each audio segment obtained by segmentation.

[0082] In some embodiments, taking two singers (singer A and singer B) singing a song together as an example, singer A and singer B sing a song together to obtain a target chorus audio, the terminal can first divide the target chorus audio into n audio segments according to the starting time of each sentence (such as a lyric sentence), and then determine the timbre characteristics of each audio segment, such as Figure 3 As shown, a flow chart of determining timbre features is provided. The terminal inputs each audio clip into a pre-trained vocal / accompaniment separation deep learning model. The vocal / accompaniment separation deep learning model can separate pure vocal audio and pure accompaniment audio from the target chorus audio. Separating the pure vocal audio can avoid interference from the accompaniment during subsequent extraction of timbre feature vectors. The terminal then inputs the pure vocal audio of each audio clip into a pre-trained singing timbre feature vector extraction deep learning model to determine the timbre features (timbre feature vectors) corresponding to the pure vocal audio of each audio clip, and finally obtains n timbre feature vectors. The terminal can calculate the similarity between the n timbre feature vectors in pairs. If the similarity is higher than a preset threshold, the terminal can determine that the lyrics (or sentences) corresponding to the corresponding two audio clips are sung by the same singer. Otherwise, the terminal can determine that the lyrics (or sentences) corresponding to the corresponding two audio clips are sung by different singers. Figure 4 As shown, a schematic diagram for determining the similarity between timbre features is provided. Based on the similarity between n timbre feature vectors, the terminal can divide n audio clips into several audio clip sets, each audio clip set has a matching singer, and the singers matching different audio clip sets can be different. Therefore, the terminal can determine the singer of each audio clip in the n audio clips according to the audio clip set, and then the terminal can segment the target chorus audio according to whether the audio clips are sung by the same singer. In actual applications, the terminal can identify the audio clip sung by singer A in the n audio clips as segment A, and identify the audio clip sung by singer B in the n audio clips as segment B; the terminal can also identify the lyrics of the audio clip sung by singer A in the n audio clips as segment A, and identify the lyrics of the audio clip sung by singer B in the n audio clips as segment B. Singer A can sing the lyrics marked by segment A, and singer B can sing the lyrics marked by segment B.

[0083] In this embodiment, by analyzing the similarity between the timbre features of each audio segment in the chorus audio, the segments sung by different singers in the chorus audio are distinguished, so as to perform audio segmentation. The timbre features of the singers can be utilized to correct the problem of unreasonable segmentation, accurately segment the audio, and improve the user experience.

[0084] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0085] Based on the same inventive concept, the embodiment of the present application also provides a chorus audio segmentation device for implementing the chorus audio segmentation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more chorus audio segmentation device embodiments provided below can refer to the limitations of the chorus audio segmentation method above, and will not be repeated here.

[0086] In an exemplary embodiment, Figure 5 As shown, a chorus audio segmentation device is provided, including: a separation module 502, a determination module 504 and a segmentation module 506, wherein:

[0087] The separation module 502 is used to separate the pure vocal audio from the target chorus audio and divide the pure vocal audio into a plurality of audio segments.

[0088] The determination module 504 is used to determine the timbre feature of each of the audio clips by using a pre-trained timbre feature extraction model.

[0089] The segmentation module 506 is used to determine the audio segmentation result of the target chorus audio according to the similarity between the timbre features of each of the audio segments; the audio segmentation result represents the audio segments corresponding to the sentences sung by different singers in the target chorus audio.

[0090] In one of the exemplary embodiments, the segmentation module 506 is further specifically used to divide each of the audio segments into at least one audio segment set according to the degree of similarity between the timbre features of each of the audio segments; the degree of similarity between the timbre features of each of the audio segments in the same audio segment set is greater than or equal to a preset similarity threshold; and the audio segmentation result of the target chorus audio is determined based on each of the audio segment sets.

[0091] In one of the exemplary embodiments, the segmentation module 506 is further configured to select, for any audio segment in the audio segments, an audio segment whose similarity with the timbre characteristics of the any audio segment is greater than or equal to a preset similarity threshold as an audio segment matching the any audio segment; and construct an audio segment set based on the any audio segment and the audio segment matching the any audio segment.

[0092] In one of the exemplary embodiments, the segmentation module 506 is further specifically used to construct an audio segment set based on any audio segment among the audio segments, if an audio segment whose degree of similarity with the timbre characteristics of any audio segment is greater than or equal to a preset similarity threshold cannot be screened out from the audio segments.

[0093] In one of the exemplary embodiments, the segmentation module 506 is further specifically used to, for each of the audio clip sets, use the audio clips in the audio clip set as the singing clips of the singers matching the audio clip set; and use the singing clips of each of the singers as the audio segmentation results of the target chorus audio.

[0094] In one of the exemplary embodiments, the above-mentioned device also includes an analysis module, which is specifically used to obtain length information of the timbre features of each of the audio clips, and determine the similarity between the timbre features of each of the audio clips based on the dot product between the timbre features of each of the audio clips and the length information; or determine the similarity between the timbre features of each of the audio clips based on the distance between the timbre features of each of the audio clips.

[0095] In one of the exemplary embodiments, the separation module 502 is further used to obtain the start time of each lyric sentence in the target chorus audio; and divide the pure vocal audio according to the start time to obtain the plurality of audio segments.

[0096] Each module in the above-mentioned chorus audio segmentation device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.

[0097] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a chorus audio segmentation method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0098] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0099] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0100] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0101] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0102] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0103] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0104] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0105] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A chorus audio segmentation method, characterized in that: The method comprises: Separating pure vocal audio from the target chorus audio, and dividing the pure vocal audio into a plurality of audio segments; Determining the timbre feature of each of the audio clips by using a pre-trained timbre feature extraction model; The audio segmentation result of the target chorus audio is determined according to the similarity between the timbre features of each of the audio segments; the audio segmentation result represents the audio segments corresponding to the sentences sung by different singers in the target chorus audio.

2. The method according to claim 1, characterized in that The step of determining the audio segmentation result of the target chorus audio according to the similarity between the timbre features of the audio segments comprises: According to the similarity between the timbre features of the audio clips, the audio clips are divided into at least one audio clip set; the similarity between the timbre features of the audio clips in the same audio clip set is greater than or equal to a preset similarity threshold; An audio segmentation result of the target chorus audio is determined according to each of the audio segment sets.

3. The method according to claim 2, characterized in that The step of dividing the audio segments into at least one audio segment set according to the similarity between the timbre features of the audio segments comprises: For any audio segment among the audio segments, an audio segment whose similarity with the timbre feature of the any audio segment is greater than or equal to a preset similarity threshold is selected from the audio segments as an audio segment matching the any audio segment; An audio segment set is constructed according to the any audio segment and the audio segment matching the any audio segment.

4. The method according to claim 2, characterized in that: The step of dividing the audio segments into at least one audio segment set according to the similarity between the timbre features of the audio segments comprises: For any audio segment among the audio segments, if an audio segment whose degree of similarity with the timbre feature of the any audio segment is greater than or equal to a preset similarity threshold cannot be screened out from the audio segments, an audio segment set is constructed based on the any audio segment.

5. The method according to claim 2, characterized in that: The step of determining the audio segmentation result of the target chorus audio according to each of the audio segment sets includes: For each of the audio segment sets, taking the audio segments in the audio segment set as the singing segments of the singer matching the audio segment set; The singing segments of each of the singers are used as the audio segmentation results of the target chorus audio.

6. The method according to claim 1, characterized in that The method further comprises: Acquire length information of the timbre feature of each of the audio segments, and determine the similarity between the timbre features of each of the audio segments according to the dot product between the timbre features of each of the audio segments and the length information; or The similarity between the timbre features of the audio segments is determined according to the distances between the timbre features of the audio segments.

7. The method according to claim 1, characterized in that The step of dividing the pure human voice audio into a plurality of audio segments comprises: Obtaining the start time of each lyric sentence in the target chorus audio; The pure human voice audio is divided according to the start time to obtain the plurality of audio segments.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images