Music selection method and apparatus, computer device, and storage medium

Abstract

The application relates to a music selection method and device, computer equipment and a storage medium. The method comprises the following steps: acquiring multiple pieces of music, wherein the multiple pieces of music comprise multiple pieces of therapeutic music, multiple pieces of unlabeled music and multiple pieces of user-preferred music; extracting audio features of each piece of music to obtain a therapeutic music audio feature set, an unlabeled music audio feature set and a preferred music audio feature set; determining a therapeutic score and a preference score of each piece of unlabeled music according to the therapeutic music audio feature set, the unlabeled music audio feature set and the preferred music audio feature set; and screening target music meeting a preset condition from the multiple pieces of unlabeled music according to the therapeutic score and the preference score of each piece of unlabeled music. The embodiment of the application can automatically select music works with therapeutic property and user preference from unlabeled music.

Description

Technical Field

[0001] This application relates to the field of music therapy, and in particular to a music selection method, apparatus, computer device, and storage medium. Background Technology

[0002] The following statements are provided only as background information in relation to this application and do not necessarily constitute prior art.

[0003] Music therapy is an evidence-based adjunctive therapy that can produce particularly effective results in the treatment of psychosomatic illnesses. The musical works used in music therapy (referred to as therapeutic music) are typically selected by relevant professionals (such as music therapists) based on their professional empirical experience and judgment. However, this selection method relies heavily on the individual professional empirical experience of the professionals, and because this experience is difficult to quantify, it is difficult to automate the selection process.

[0004] An existing approach proposes learning therapeutic-related audio features from existing therapeutic music, and then using these learned audio features to assign therapeutic value scores to unknown music (i.e., music works whose suitability for music therapy is uncertain). Compared to the aforementioned method where professionals select music based on experience, this approach can more efficiently determine whether a piece of music is therapeutic music and / or the strength of its therapeutic value.

[0005] However, the inventors discovered that experienced music therapists usually adjust the selection of music based on the patient's musical preferences. Using music that is both therapeutic and preference-based often yields better results. In contrast, the above approach only scores unknown music from the perspective of the music itself, ignoring the patient's musical preferences. This results in the final selection of therapeutic music that lacks personalization and has poor therapeutic effects. Summary of the Invention

[0006] To address the aforementioned shortcomings or disadvantages, this application provides a music selection method, apparatus, computer device, and storage medium. The embodiments of this application can automatically select therapeutic music works from unlabeled music that also take into account user preferences.

[0007] This application provides a music selection method according to a first aspect, in one embodiment of which the method includes:

[0008] Acquire multiple music tracks, including multiple therapeutic music tracks, multiple unlabeled music tracks, and multiple user-preferred music tracks;

[0009] The audio features of each piece of music are extracted to obtain the audio feature set of therapeutic music, the audio feature set of unlabeled music, and the audio feature set of preferred music.

[0010] The therapeutic score and preference score for each unlabeled music piece were determined based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preference music audio feature set.

[0011] Based on the therapeutic and preference ratings of each unlabeled song, target songs that meet the preset criteria are selected from the above-mentioned unlabeled songs.

[0012] In one embodiment, the therapeutic score and preference score for each unlabeled music track are determined based on a therapeutic music audio feature set, an unlabeled music audio feature set, and a preference music audio feature set, including:

[0013] The therapeutic score for each unlabeled music piece was determined based on the audio feature set of therapeutic music and the audio feature set of unlabeled music.

[0014] The preference score for each unlabeled music is determined based on the unlabeled music audio feature set and the preferred music audio feature set.

[0015] In one embodiment, determining the therapeutic score for each unlabeled music piece based on a therapeutic music audio feature set and an unlabeled music audio feature set includes:

[0016] Construct a therapeutic evaluation model;

[0017] The target weighting vector of the therapeutic evaluation model is determined based on the feature sets of therapeutic music audio and unlabeled music audio.

[0018] The therapeutic score for each unlabeled music track is determined based on the target weighted vector of the therapeutic assessment model.

[0019] In one embodiment, the therapeutic evaluation model includes:

[0020]

[0021]

[0022] Where, σ 2 It is the fractional variance of therapeutic music; s t and S' t These are the sequence number and total number of the therapeutic music tracks; s u and S' u These are the sequence number and total number of unlabeled music tracks, respectively. It is the sth t The audio characteristics of the first therapeutic music, It is the sth u The audio characteristics of unlabeled music; w t It is a weighted vector, b t It is the bias.

[0023] In one embodiment, determining a preference score for each unlabeled piece of music based on an unlabeled music audio feature set and a preferred music audio feature set includes:

[0024] Construct a preference evaluation model;

[0025] The target weighting vector of the preference evaluation model is determined based on the feature set of unlabeled music audio and the feature set of preferred music audio.

[0026] The preference score for each unlabeled music track is determined based on the target weighted vector of the preference evaluation model.

[0027] In one embodiment, the preference evaluation model includes:

[0028]

[0029]

[0030] Where, σ 2 It is the fractional variance of music preference; s p and S' p These are the preferred music number and the total number, respectively; s u and S' u These are the sequence number and total number of unlabeled music tracks, respectively. It is the sth p The primary preference is for the audio characteristics of music. It is the sth u The audio characteristics of unlabeled music; w p It is a weighted vector, b p It is the bias.

[0031] In one embodiment, target music that meets preset conditions is selected from the plurality of unlabeled music tracks based on the therapeutic and preference scores of each track, including:

[0032] The overall score for each unlabeled song is determined based on its therapeutic and preference scores.

[0033] Based on the overall score of each unlabeled song, target songs that meet the preset conditions are selected from the above multiple unlabeled songs.

[0034] This application provides a music selection device according to a second aspect, wherein in one embodiment the device includes:

[0035] The music acquisition module is used to acquire multiple music tracks, including multiple therapeutic music tracks, multiple unlabeled music tracks, and multiple user-preferred music tracks.

[0036] The extraction module is used to extract the audio features of each piece of music, resulting in a set of audio features for therapeutic music, a set of audio features for unlabeled music, and a set of audio features for preferred music.

[0037] The evaluation module is used to determine the therapeutic score and preference score for each unlabeled music based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preference music audio feature set.

[0038] The selection module is used to filter out target music that meets preset conditions from the above multiple unlabeled music tracks based on the therapeutic rating and preference rating of each unlabeled music track.

[0039] This application provides a computer device according to a third aspect, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the above-described methods.

[0040] According to a fourth aspect, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described methods.

[0041] The embodiments described above in this application can automatically learn the audio features related to music therapy in therapeutic music and the audio features related to user preferences in user-preferred music when selecting music. This allows for the evaluation of the therapeutic and preference-related aspects of each unlabeled piece of music. Based on the therapeutic and preference scores of each unlabeled piece, therapeutic music that also considers user preferences can be automatically selected from the unlabeled music. Compared to the commonly used method of having professionals select music with both therapeutic and preference-related aspects based on experience, the embodiments of this application are more efficient, enabling the selection of music with both therapeutic and preference-related aspects from a large batch of unlabeled music. Furthermore, compared to selecting therapeutic music solely based on its audio features, the embodiments of this application can select music with both therapeutic and preference-related aspects, resulting in more personalized music that better matches user preferences and can produce better therapeutic effects. Attached Figure Description

[0042] Figure 1 A flowchart illustrating a music selection method provided for one or more embodiments of this application;

[0043] Figure 2 This is a schematic diagram illustrating the principle of comprehensive evaluation of unlabeled music, provided as an example in this application.

[0044] Figure 3 A structural block diagram of a music selection device provided for one or more embodiments of this application;

[0045] Figure 4 This is an internal structural diagram of a computer device provided for one or more embodiments of this application. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0047] This application provides a music selection method that can be used to select music with both therapeutic and preferred qualities from unlabeled music. Compared to the method of selecting music with both therapeutic and preferred qualities by professionals based on experience, this method is more efficient and can filter out music with both therapeutic and preferred qualities from a large number of unlabeled music tracks. Compared to the method of selecting therapeutic music based solely on the audio characteristics of the music, this method can select music with both therapeutic and preferred qualities, and the selected music is more personalized and more in line with the user's preferences, thus producing better therapeutic effects.

[0048] The execution subject of this music selection method can be a computing device that assists in music therapy, such as a user terminal device like a laptop, smartphone, tablet, or desktop computer, or a remote computing device like a server. The server can be a standalone server or a server cluster consisting of multiple servers. Alternatively, the execution subject can also be a smart electronic musical instrument.

[0049] This music selection method can be applied to various music therapy scenarios. For example, in one scenario, a music therapist can use this method to select a wider range of therapeutic and preferred music for the patient. Experienced music therapists often select music that is both therapeutic and preferred to enhance the therapeutic effect. However, manual music selection is inefficient, resulting in a small number of selected tracks that are insufficient to address patient tolerance to specific types of music. This method, however, allows music therapists to efficiently select a larger range of music suitable for music therapy, thus preventing patients from developing tolerance to the therapist's chosen music.

[0050] In one embodiment, the music selection method may include, for example: Figure 1 The steps shown below will be explained using the method applied to a server as an example.

[0051] S110: Obtain multiple music tracks, including multiple therapeutic music tracks, multiple unlabeled music tracks, and multiple user-preferred music tracks.

[0052] The therapeutic music mentioned above refers to musical works that have been confirmed by professionals (such as music therapists) as suitable for music therapy. This embodiment does not limit the selection of therapeutic music, and users can select according to the information and data they have. Unlabeled music refers to other music besides therapeutic music, which has not yet been confirmed as suitable for music therapy. User-preferred music refers to musical works that users (such as patients receiving music therapy) prefer.

[0053] Here, the server can retrieve various types of music from a pre-defined music library. There can be multiple music libraries, each storing different types of music. For example, library A might store therapeutic music, library B might store unlabeled music, and library C might store music based on user preferences. Furthermore, music based on different user preferences can be stored in different music libraries.

[0054] S120: Extract the audio features of each piece of music to obtain the audio feature set of therapeutic music, the audio feature set of unlabeled music, and the audio feature set of preferred music.

[0055] The aforementioned therapeutic music audio feature set includes audio features extracted from each therapeutic music track; the aforementioned unlabeled music audio feature set includes audio features extracted from each unlabeled music track; and the aforementioned preferred music audio feature set includes audio features extracted from each user-preferred music track.

[0056] It should be noted that this embodiment does not specifically limit the content of the audio features extracted by the server from the music. The server can extract one or more types of audio features from the music. For example, these could be Mel Frequency Cepstrum Coefficient (MFCC), rhythm intensity, and other audio features. Understandably, the more types of audio features the server extracts from the music, the more accurately it can select therapeutic music that also caters to user preferences from unlabeled music. However, at the same time, the processing speed will be slower when analyzing the audio features of each piece of music. Therefore, the number and / or types of audio features that the server needs to extract can be adjusted according to the specific requirements of accuracy and processing speed in the actual application scenario.

[0057] S130: Determine the therapeutic score and preference score for each unlabeled music piece based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preference music audio feature set.

[0058] The therapeutic rating of unlabeled music is a quantitative description of its therapeutic ability in music therapy; a higher therapeutic rating indicates a stronger therapeutic ability. The preference rating of unlabeled music is a quantitative description of a user's degree of preference for unlabeled music; a higher preference rating indicates a stronger user preference.

[0059] When determining the therapeutic score and preference score of each unlabeled music piece based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preference music audio feature set, the server can first determine the therapeutic score of each unlabeled music piece based on the therapeutic music audio feature set and the unlabeled music audio feature set, and then determine the preference score of each unlabeled music piece based on the unlabeled music audio feature set and the preference music audio feature set.

[0060] When the server determines the therapeutic score for each unlabeled piece of music based on the therapeutic music audio feature set and the unlabeled music audio feature set, it may include the following steps:

[0061] 1. Construct a therapeutic evaluation model; wherein, the therapeutic evaluation model includes:

[0062]

[0063]

[0064] Where, σ 2 It is the fractional variance of therapeutic music; s t and S' t These are the sequence number and total number of the therapeutic music tracks; s u and S' u These are the sequence number and total number of unlabeled music tracks, respectively. It is the sth t The audio characteristics of the first therapeutic music, It is the sth u The audio characteristics of unlabeled music; w t It is a weighted vector, b t It is the bias value.

[0065] 2. Determine the target weighting vector of the therapeutic evaluation model based on the therapeutic music audio feature set and the unlabeled music audio feature set; wherein, the target weighting vector of the therapeutic evaluation model refers to the aforementioned w. t ;

[0066] 3. Determine the therapeutic score for each unlabeled music piece based on the target weighted vector of the therapeutic assessment model.

[0067] The steps described above are explained in detail below.

[0068] In this embodiment, the audio features of the therapeutic music are used as its feature vector, and the s-th... t The eigenvector of the first therapeutic music is denoted as Among them, s t The sequence number for the therapeutic music, s t =1,2,...,S' t S't It is the total number of therapeutic music tracks; using the audio features of unlabeled music as its feature vector, the s-th... u The feature vector of the first unlabeled music is denoted as Among them, s u The sequence number for unlabeled music, s u =1,2,...,S' u S' u It represents the total number of unlabeled music tracks; the scoring function used to measure the therapeutic effect of music x can be denoted as f. t (x), taking linear representation as an example, the scoring function can be described as:

[0069]

[0070] Among them, w t b is a weighted vector t This is the bias value.

[0071] To learn the common patterns inherent in therapeutic music, it can be modeled as a minimum variance problem, namely:

[0072]

[0073] Where, σ 2 The fractional variance of therapeutic music.

[0074] Furthermore, since therapeutic music is the only reliable learning sample available and the best known learning target, this superiority can be modeled as follows:

[0075]

[0076] By adopting the above constraints, the excellence of therapeutic music can be guaranteed, while tolerating that some unlabeled music scores higher than therapeutic music (in fact, this possibility does exist and cannot be ignored).

[0077] To standardize the representation, the therapeutic score can be further normalized:

[0078]

[0079]

[0080] Combining the above formulas, we can obtain the following therapeutic evaluation model:

[0081]

[0082]

[0083] Furthermore, the objective function in the equation can be rewritten as:

[0084]

[0085] Among them, R t The sum of the following covariance matrices:

[0086]

[0087] Next, in order to solve the problem, we transform it into the following equivalent convex optimization problem:

[0088]

[0089]

[0090] Where δ (δ≥0) is the fractional margin between therapeutic music and unlabeled music; C (C≥0) is a penalty factor that adjusts the fractional variance and the margin. The value of the penalty factor can be flexibly set according to the actual problem. This embodiment does not impose specific limitations. For example, different penalty factors can be set for different diseases. The formula can be solved using the Lagrange multiplier method. In addition, in actual calculations, expanding x by one dimension and assigning 1 to the elements of the corresponding dimension will allow b to be calculated. t It is solved together with the extended dimension of w.

[0091] After solving for a set of optimal weighted vectors After that, you can give each unlabeled piece of music a therapeutic score, which will give you a result. The value of .

[0092] Furthermore, when the server determines the preference rating for each unlabeled piece of music based on the unlabeled music audio feature set and the preferred music audio feature set, it may include the following steps:

[0093] a. Construct a preference evaluation model; wherein, the preference evaluation model includes:

[0094]

[0095]

[0096] Where, σ 2 It is the fractional variance of music preference; s p and S' p These are the preferred music number and the total number, respectively; s u and S' u These are the sequence number and total number of unlabeled music tracks, respectively. It is the sth p The primary preference is for the audio characteristics of music. It is the sth u The audio characteristics of unlabeled music; wp It is a weighted vector, b p It is the bias value.

[0097] b. Determine the target weighting vector of the preference evaluation model based on the unlabeled music audio feature set and the preferred music audio feature set; wherein, the target weighting vector of the preference evaluation model refers to the w mentioned above. p ;

[0098] c. Determine the preference score for each unlabeled music track based on the target weighted vector of the preference evaluation model.

[0099] The steps described above are explained in detail below.

[0100] In this embodiment, based on the above formula and the same method, a scoring function for measuring preference for music x can be learned.

[0101] Specifically, in this embodiment, the audio features of preferred music are used as its feature vector, and the s-th... p The eigenvector of the primary music preference is denoted as Among them, s p The sequence number is for music preferences, s p =1,2,...,S' p S p It represents the total number of preferred music tracks; using the audio features of unlabeled music as its feature vector, the s-th... u The feature vector of the first unlabeled music is denoted as Among them, s u The sequence number for unlabeled music, s u =1,2,...,S' u S' u This represents the total number of unlabeled music tracks; for example, the scoring function can be represented linearly. Described as:

[0102]

[0103] Among them, w p b is a weighted vector p This is the bias value.

[0104] To learn the inherent common patterns in music preferences, this can be modeled as a minimum variance problem, i.e.:

[0105]

[0106] Where, σ 2 The fractional variance represents the preference for music.

[0107] Furthermore, since music preference is the known optimal learning objective, this superiority can be modeled as a constraint:

[0108]

[0109] By adopting the above constraints, we can ensure the superiority of the preferred music while tolerating that some unlabeled music scores higher than the preferred music (in fact, this possibility does exist and cannot be ignored).

[0110] To standardize the representation, the preference scores can be further normalized:

[0111]

[0112]

[0113] Combining equations (10), (11), and (12) above, the following preference evaluation model can be obtained:

[0114]

[0115]

[0116] Furthermore, the objective function in the equation can be rewritten as:

[0117]

[0118] Among them, R p The sum of the following covariance matrices:

[0119]

[0120] Next, in order to solve the problem, we transform it into the following equivalent convex optimization problem:

[0121]

[0122]

[0123] Where δ (δ≥0) is the fractional margin between preferred music and unlabeled music, and C (C≥0) is the penalty factor that modifies the fractional variance and the margin. Equation (16) can be solved using the Lagrange multiplier method. Furthermore, in practical calculations, extending x by one dimension and assigning 1 to the elements of that dimension allows b to be... p It is solved together as an extended dimension of w.

[0124] After solving for a set of optimal weighted vectors Then, you can assign a preference score to each unlabeled piece of music, thus obtaining... The value of .

[0125] Furthermore, in one possible implementation, when selecting music, the optimal therapeutic weighting vector and optimal preference weighting vector obtained from previous therapeutic and preference optimizations can be directly used to score each unlabeled piece of music, thus obtaining... and When it is necessary to select target music from a large number of unlabeled music tracks, this implementation method can quickly perform a comprehensive evaluation of each unlabeled music track, thereby speeding up the selection of target music.

[0126] Furthermore, as music therapy progresses, more and more empirical and training data can be obtained. Therefore, the optimal therapeutic weighting vector and the optimal preference weighting vector can be recalculated using the newly acquired data, thereby enabling a more accurate evaluation of the therapeutic and preference properties of unlabeled music.

[0127] S140: Based on the therapeutic and preference scores of each unlabeled music track, select target music tracks from the aforementioned unlabeled music tracks that meet preset criteria. Target music tracks refer to unlabeled music tracks that meet the preset criteria.

[0128] The server can determine the overall score of each unlabeled music piece based on its therapeutic and preference scores, and then select target music pieces that meet preset conditions from the multiple unlabeled music pieces based on their overall scores.

[0129] There are various ways to determine the overall score based on the therapeutic and preference scores of unlabeled music.

[0130] In one example, a comprehensive scoring function can be defined. Make it the product of the scores from both aspects:

[0131]

[0132] That is, the overall score of an unlabeled piece of music is equal to the product of its therapeutic score and its preference score.

[0133] In another example, a simple weighted summation could be used to obtain the final composite score, i.e.:

[0134]

[0135] Among them, w 1 w2 and w2 are the weights set for therapeutic and preference ratings of unlabeled music.

[0136] After calculating the overall score for each unlabeled music track, the overall scores of each unlabeled music track can be sorted, and a specified number of music tracks that balance therapeutic and personal preferences can be selected based on the sorting.

[0137] If a user prioritizes both therapeutic and personal preferences, the server can filter target music based on the user's specific needs by sorting and then filtering.

[0138] Taking a focus on therapeutic effects as an example: The server can first sort all the unlabeled music according to the therapeutic effect rating from high to low, then set a preference rating threshold θ and the final number of music N required. Then, in the sorted unlabeled music, starting from the unlabeled music with a high therapeutic effect rating, the server scans each music one by one. If the preference rating of the scanned unlabeled music is higher than or equal to θ, it is included in the final selection set, until the number of music in the final selection set reaches N.

[0139] To better understand the above embodiments, this application also provides an example of comprehensively evaluating unlabeled music. In this example, three therapeutic music tracks, two preferred music tracks, and four unlabeled music tracks are used for illustration. For details, please refer to [link to relevant documentation]. Figure 2 Given the three types of music sets mentioned above, the optimal therapeutic weighting vector and the optimal preference weighting vector can be learned through two processes: therapeutic optimization and preference optimization. Finally, the comprehensive evaluation of each unlabeled music is based on these two vectors. The comprehensive evaluation is calculated by simple multiplication, i.e., Equation (17).

[0140] It should be noted that, regarding the various steps included in the music selection method provided in any of the above embodiments, unless explicitly stated herein, there is no strict order restriction on the execution of these steps; they can be executed in other orders. Furthermore, at least some of these steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is also not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0141] Based on the same inventive concept, this application also provides a music selection device. In this embodiment, as... Figure 3 As shown, the music selection device includes the following modules:

[0142] The music acquisition module 110 is used to acquire multiple pieces of music, including multiple therapeutic music, multiple unlabeled music, and multiple user-preferred music.

[0143] The extraction module 120 is used to extract the audio features of each piece of music, resulting in a therapeutic music audio feature set, an unlabeled music audio feature set, and a preference music audio feature set.

[0144] Evaluation module 130 is used to determine the therapeutic score and preference score for each unlabeled music based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preference music audio feature set.

[0145] Selection module 140 is used to filter target music that meets preset conditions from the above multiple unlabeled music tracks based on the therapeutic score and preference score of each unlabeled music track.

[0146] In one embodiment, the evaluation module 130 includes:

[0147] The therapeutic assessment submodule is used to determine the therapeutic score for each unlabeled music piece based on the audio feature set of therapeutic music and the audio feature set of unlabeled music.

[0148] The preference rating submodule is used to determine the preference score for each unlabeled music based on the unlabeled music audio feature set and the preferred music audio feature set.

[0149] In one embodiment, the therapeutic evaluation submodule includes:

[0150] The first building block is used to construct the therapeutic evaluation model;

[0151] The first weighted vector determination unit is used to determine the target weighted vector of the therapeutic evaluation model based on the therapeutic music audio feature set and the unlabeled music audio feature set.

[0152] Therapeutic scoring unit is used to determine the therapeutic score for each unlabeled music piece based on the target weighted vector of the therapeutic assessment model.

[0153] In one embodiment, the therapeutic evaluation model includes:

[0154]

[0155]

[0156] Where, σ 2 It is the fractional variance of therapeutic music; s t and S' t These are the sequence number and total number of the therapeutic music tracks; s u and S' u These are the sequence number and total number of unlabeled music tracks, respectively. It is the sth t The audio characteristics of the first therapeutic music, It is the sth uThe audio characteristics of unlabeled music; w t It is a weighted vector, b t It is the bias.

[0157] In one embodiment, the preference evaluation submodule includes:

[0158] The second building block is used to construct the preference evaluation model;

[0159] The second weighted vector determination unit is used to determine the target weighted vector of the preference evaluation model based on the unlabeled music audio feature set and the preferred music audio feature set.

[0160] The preference rating unit is used to determine the preference rating for each unlabeled music piece based on the target weighted vector of the preference rating model.

[0161] In one embodiment, the preference evaluation model includes:

[0162]

[0163]

[0164] Where, σ 2 It is the fractional variance of music preference; s p and S' p These are the preferred music number and the total number, respectively; s u and S' u These are the sequence number and total number of unlabeled music tracks, respectively. It is the sth p The primary preference is for the audio characteristics of music. It is the sth u The audio characteristics of unlabeled music; w p It is a weighted vector, b p It is the bias value.

[0165] In one embodiment, the selection module 140 includes:

[0166] The comprehensive scoring submodule is used to determine the comprehensive score for each unlabeled music based on its therapeutic and preference scores.

[0167] The filtering submodule is used to filter out target music that meets preset conditions from the above multiple unlabeled music tracks based on the comprehensive score of each unlabeled music track.

[0168] For specific limitations regarding the music selection device, please refer to the limitations on the music selection method above, which will not be repeated here. Each module in the aforementioned music selection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0169] In one embodiment, a computer device is provided, the internal structure of which can be shown as follows: Figure 4 As shown.

[0170] The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and the database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The database stores data such as audio features and weighted vectors; the specific stored data may also be defined in the above method embodiments. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a music selection method.

[0171] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0172] This embodiment also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the following steps:

[0173] Multiple music tracks are acquired, including multiple therapeutic music tracks, multiple unlabeled music tracks, and multiple user-preferred music tracks. Audio features of each track are extracted to obtain audio feature sets for therapeutic music, unlabeled music, and preferred music tracks. A therapeutic score and a preference score are determined for each unlabeled music track based on these sets. Based on the therapeutic and preference scores, target music tracks that meet preset conditions are selected from the multiple unlabeled music tracks.

[0174] In one implementation, when the processor executes a computer program to determine the therapeutic score and preference score of each unlabeled piece of music based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preferred music audio feature set, it also performs the following steps:

[0175] The therapeutic score for each unlabeled music piece is determined based on the therapeutic music audio feature set and the unlabeled music audio feature set; the preference score for each unlabeled music piece is determined based on the unlabeled music audio feature set and the preference music audio feature set.

[0176] In one implementation, when the processor executes a computer program to determine the therapeutic score of each unlabeled piece of music based on the therapeutic music audio feature set and the unlabeled music audio feature set, it also performs the following steps:

[0177] Construct a therapeutic evaluation model; determine the target weighting vector of the therapeutic evaluation model based on the audio feature sets of therapeutic music and unlabeled music; determine the therapeutic score for each unlabeled music based on the target weighting vector of the therapeutic evaluation model.

[0178] In one implementation, when the processor executes a computer program to determine a preference score for each unlabeled piece of music based on the unlabeled music audio feature set and the preferred music audio feature set, it also performs the following steps:

[0179] Construct a preference evaluation model; determine the target weighting vector of the preference evaluation model based on the audio feature set of unlabeled music and the audio feature set of preferred music; determine the preference score for each unlabeled music based on the target weighting vector of the preference evaluation model.

[0180] In one implementation, when the processor executes a computer program to select target music that meets preset conditions from the plurality of unlabeled music tracks based on the therapeutic and preference scores of each unlabeled music track, it also performs the following steps:

[0181] A comprehensive score for each unlabeled music piece is determined based on its therapeutic and preference scores; target music pieces that meet preset criteria are then selected from the aforementioned unlabeled music pieces based on their comprehensive scores.

[0182] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0183] Multiple music tracks are acquired, including multiple therapeutic music tracks, multiple unlabeled music tracks, and multiple user-preferred music tracks. Audio features of each track are extracted to obtain audio feature sets for therapeutic music, unlabeled music, and preferred music tracks. A therapeutic score and a preference score are determined for each unlabeled music track based on these sets. Based on the therapeutic and preference scores, target music tracks that meet preset conditions are selected from the multiple unlabeled music tracks.

[0184] In one implementation, the computer program, executed by a processor, determines the therapeutic score and preference score for each unlabeled piece of music based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preferred music audio feature set, and further performs the following steps:

[0185] The therapeutic score for each unlabeled music piece is determined based on the therapeutic music audio feature set and the unlabeled music audio feature set; the preference score for each unlabeled music piece is determined based on the unlabeled music audio feature set and the preference music audio feature set.

[0186] In one implementation, the computer program, executed by a processor, determines the therapeutic score of each unlabeled piece of music based on the therapeutic music audio feature set and the unlabeled music audio feature set, and further performs the following steps:

[0187] Construct a therapeutic evaluation model; determine the target weighting vector of the therapeutic evaluation model based on the audio feature sets of therapeutic music and unlabeled music; determine the therapeutic score for each unlabeled music based on the target weighting vector of the therapeutic evaluation model.

[0188] In one implementation, the computer program, executed by a processor, determines a preference score for each unlabeled piece of music based on the unlabeled music audio feature set and the preferred music audio feature set, and further performs the following steps:

[0189] Construct a preference evaluation model; determine the target weighting vector of the preference evaluation model based on the audio feature set of unlabeled music and the audio feature set of preferred music; determine the preference score for each unlabeled music based on the target weighting vector of the preference evaluation model.

[0190] In one implementation, the computer program, executed by a processor, selects target music that meets preset conditions from the plurality of unlabeled music tracks based on the therapeutic and preference scores of each track, and further performs the following steps:

[0191] A comprehensive score for each unlabeled music piece is determined based on its therapeutic and preference scores; target music pieces that meet preset criteria are then selected from the aforementioned unlabeled music pieces based on their comprehensive scores.

[0192] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0193] Those skilled in the art will understand that implementing all or part of the processes in the above method embodiments can be accomplished by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), memory bus (RAMbus), direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0194] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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 specification.

[0195] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A music selection method, characterized in that, The method includes: Acquire multiple music tracks, including multiple therapeutic music tracks, multiple unlabeled music tracks, and multiple user-preferred music tracks; The audio features of each piece of music are extracted to obtain the audio feature set of therapeutic music, the audio feature set of unlabeled music, and the audio feature set of preferred music. The therapeutic score and preference score for each unlabeled music piece are determined based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preference music audio feature set. Based on the therapeutic and preference scores of each unlabeled music piece, target music that meets preset conditions is selected from the multiple unlabeled music pieces. The therapeutic score and preference score for each unlabeled music track are determined based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preference music audio feature set, including: The therapeutic score for each unlabeled music piece is determined based on the therapeutic music audio feature set and the unlabeled music audio feature set; A preference score for each unlabeled music track is determined based on the unlabeled music audio feature set and the preferred music audio feature set.

2. The method as described in claim 1, characterized in that, The therapeutic score for each unlabeled music piece is determined based on the therapeutic music audio feature set and the unlabeled music audio feature set, including: Construct a therapeutic evaluation model; The target weighting vector of the therapeutic evaluation model is determined based on the therapeutic music audio feature set and the unlabeled music audio feature set; The therapeutic score for each unlabeled music track is determined based on the target weighted vector of the therapeutic evaluation model.

3. The method as described in claim 2, characterized in that, The therapeutic evaluation model includes: ； in, It is the fractional variance of therapeutic music; and These are the sequence number and total number of the therapeutic music tracks; and These are the sequence number and total number of unlabeled music tracks, respectively. It is the first The audio characteristics of primary therapeutic music, It is the first The audio characteristics of unlabeled music; It is a weighted vector. It is the bias.

4. The method as described in claim 1, characterized in that, The preference score for each unlabeled music track is determined based on the unlabeled music audio feature set and the preferred music audio feature set, including: Construct a preference evaluation model; The target weighting vector of the preference evaluation model is determined based on the unlabeled music audio feature set and the preferred music audio feature set. The preference score for each unlabeled music track is determined based on the target weighted vector of the preference evaluation model.

5. The method as described in claim 4, characterized in that, The preference evaluation model includes: ； in, It is the fractional variance of music preference; and These are the preferred music number and the total number, respectively; and These are the sequence number and total number of unlabeled music tracks, respectively. It is the first The primary preference is for the audio characteristics of music. It is the first The audio characteristics of unlabeled music; It is a weighted vector. It is the bias.

6. The method as described in claim 1, characterized in that, Based on the therapeutic and preference ratings of each unlabeled music track, target music that meets preset criteria is selected from the plurality of unlabeled music tracks, including: The overall score for each unlabeled song is determined based on its therapeutic and preference scores. Based on the overall score of each unlabeled song, target songs that meet preset conditions are selected from the multiple unlabeled songs.

7. A music selection device, characterized in that, The device includes: The music acquisition module is used to acquire multiple pieces of music, including multiple therapeutic music tracks, multiple unlabeled music tracks, and multiple user-preferred music tracks. The extraction module is used to extract the audio features of each piece of music, resulting in a set of audio features for therapeutic music, a set of audio features for unlabeled music, and a set of audio features for preferred music. The evaluation module is used to determine the therapeutic score and preference score for each unlabeled music based on the therapeutic music audio feature set, the unlabeled music audio feature set, and the preference music audio feature set. The selection module is used to filter target music that meets preset conditions from the multiple unlabeled music tracks based on the therapeutic rating and preference rating of each unlabeled music track. The evaluation module includes: The therapeutic evaluation submodule is used to determine the therapeutic score of each unlabeled music based on the therapeutic music audio feature set and the unlabeled music audio feature set; The preference evaluation submodule is used to determine the preference score for each unlabeled music based on the unlabeled music audio feature set and the preferred music audio feature set.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images