Piano music score difficulty identification method and system

Abstract

The invention provides a piano music score difficulty identification method and system, and the method comprises the following steps: collecting music score data, and extracting the music score features of the music score data; fusing the music score features to obtain music score difficulty features; constructing a music score difficulty identification model based on the music score difficulty features; inputting the to-be-recognized piano music score into the music score difficulty recognition model to obtain a difficulty recognition result, and completing the recognition of the music score difficulty. According to the technical scheme of the invention, by automatically extracting and analyzing the music score features, objective evaluation of the difficulty of the music score is realized, and the efficiency and quality of piano playing teaching and learning are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of music score difficulty analysis, and in particular relates to a piano music score difficulty recognition method and system. Background Art

[0002] In the field of music score difficulty recognition, a variety of technologies and methods have been developed. Among them, machine learning and deep learning technologies have attracted widespread attention due to their powerful feature learning and pattern recognition capabilities. By training a large amount of music score data, these models can learn the complex relationship between music score difficulty and features, and achieve rapid and accurate evaluation of new music scores. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a method and system for identifying the difficulty of piano music scores.

[0004] To achieve the above object, the present invention provides the following solutions: A method for identifying difficulty of piano music score comprises the following steps: Collecting music score data and extracting music score features of the music score data; Combining the music score features to obtain music score difficulty features; Based on the difficulty features of the music score, a music score difficulty recognition model is constructed; The piano music score to be identified is input into the music score difficulty identification model to obtain a difficulty identification result, thereby completing the identification of the music score difficulty.

[0005] Preferably, the music score data includes notes, chords, rhythms, dynamic and emoticon symbols and technical difficulty; the music scores are of different composers, different musical periods and styles.

[0006] Preferably, the music score features include note features, chord features, rhythm features, dynamic and emoticon features, and technical difficulty features; and the method for extracting the music score features includes: Obtaining the note feature based on note density, note duration and note span; Obtaining the chord feature based on the chord type and the chord progression; Obtaining the rhythm feature based on beat analysis and rhythm pattern; Obtaining the dynamic and emoji features based on the volume change and the emoji; The technical difficulty characteristics are obtained based on finger span, fast running and two-hand coordination.

[0007] Preferably, the method for obtaining the note feature is: The note density is calculated according to measures, beats or other time units; Extracting the duration of each note to obtain the duration of the note; Calculating the span of adjacent notes or notes within a specific range to obtain the note span; The note density, note duration and note span features are vectorized and concatenated to obtain the note feature.

[0008] Preferably, based on the score difficulty feature, a method for constructing a score difficulty recognition model is: Linearly combining a preset feature transformation matrix with the music score difficulty feature to obtain a feature projection; Based on the feature projection, obtaining a distance measure of the projection space; Based on the distance measure, obtaining a feature projection matrix; Based on the feature projection matrix, the piano score is projected into a new feature space to obtain a new score difficulty feature; The new music score difficulty feature is classified based on a neural network to obtain the piano score difficulty, thereby completing the construction of the music score difficulty recognition model.

[0009] Preferably, the method for obtaining the feature projection matrix is: Obtaining a new distance measure based on the square of the distance measure; Construct an objective function based on the new distance measure; The objective function is solved based on stochastic gradient descent to obtain a feature projection matrix.

[0010] Preferably, the neural network structure includes an input layer, a hidden layer and an output layer; The number of nodes in the input layer matches the vector dimension of the new music score difficulty feature; The hidden layer uses a Sigmoid activation function, and each layer is set with a different number of nodes; The number of nodes in the output layer matches the number of difficulty categories, and the Softmax activation function is used to obtain the probability of each difficulty category.

[0011] The present invention also provides a piano score difficulty recognition system, which is used to implement the piano score difficulty recognition method, and comprises: A music score feature acquisition module, used to collect music score data and extract music score features of the music score data; A music score difficulty feature acquisition module, used to fuse the music score features to obtain the music score difficulty feature; A model building module, used to build a music score difficulty recognition model based on the music score difficulty feature; The difficulty recognition module is used to input the piano music score to be recognized into the music score difficulty recognition model, obtain the difficulty recognition result, and complete the recognition of the music score difficulty.

[0012] Preferably, the music score features include note features, chord features, rhythm features, dynamic and emoticon features, and technical difficulty features; the music score feature acquisition module includes: A note feature acquisition unit, used to obtain the note feature based on note density, note duration and note span; A chord feature acquisition unit, configured to acquire the chord feature based on the chord type and the chord progression; A rhythm feature acquisition unit, used for acquiring the rhythm feature based on beat analysis and rhythm pattern; A dynamic feature acquisition unit, used to obtain the dynamic and emoticon features based on volume changes and emoticons; The technical difficulty feature acquisition unit is used to obtain the technical difficulty feature based on finger span, fast running and two-hand coordination.

[0013] Preferably, the model building module includes: A feature projection unit, used for linearly combining a preset feature transformation matrix with the music score difficulty feature to obtain a feature projection; A distance measurement acquisition unit, configured to obtain a distance measurement of a projection space based on the feature projection; A feature projection matrix acquisition unit, used to obtain a feature projection matrix based on the distance measure; A new feature acquisition unit is used to project the piano music score to a new feature space based on the feature projection matrix to obtain a new music score difficulty feature; The difficulty classification unit is used to classify the difficulty features of the new music score based on the neural network, obtain the difficulty of the piano score, and complete the construction of the music score difficulty recognition model.

[0014] Compared with the prior art, the present invention has the following beneficial effects: Improved evaluation accuracy: This method collects and analyzes a large amount of music score data, extracts features that are closely related to the difficulty of the music score, and builds a recognition model based on these features. Compared with traditional manual evaluation methods, this method can more objectively and accurately reflect the actual difficulty of the music score, avoiding errors caused by subjective judgment.

[0015] Improve efficiency: As the number of piano scores continues to increase, manual evaluation of the difficulty of scores has become increasingly time-consuming and labor-intensive. This method can quickly evaluate the difficulty of a large number of scores through automated processing, greatly improving the evaluation efficiency.

[0016] Personalized learning support: For piano learners, choosing music scores that suit their level of difficulty is the key to improving learning outcomes. This method can provide learners with personalized music score difficulty assessments, helping them choose appropriate practice pieces based on their own level, thereby improving their performance skills more effectively. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0018] Figure 1 The present invention is a flowchart of a piano score difficulty recognition method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. Example

[0021] like Figure 1 As shown, a piano score difficulty recognition method comprises the following steps: S1: Collect music score data and extract music score features of the music score data; In this embodiment, first, piano music score data needs to be collected from various sources (such as music books, online databases, music score software, etc.). These data cover a wide range of difficulty to ensure the generalization ability and accuracy of the recognition model. The music score data is in digital format (such as MIDI files, PDF files, or special formats for music score editing software) or image format (such as scanned music score images).

[0022] A further implementation method is that the music score data includes notes, chords, rhythms, dynamic and expression symbols and technical difficulty; and the music scores are of different composers, different musical periods and styles.

[0023] A further implementation method is that the music score features include note features, chord features, rhythm features, dynamic and emoticon features, and technical difficulty features; and the method for extracting the music score features includes: The note features are obtained based on the note density, the note duration and the note span; a further implementation method is that the method for obtaining the note features is: Note density is calculated by measure, beat or other time unit; Extract the duration of each note to obtain the note duration; Calculate the span of adjacent notes or notes within a specific range to obtain the note span; The note density, note duration and note span features are quantized and concatenated to obtain the note features.

[0024] Obtain chord features based on chord type and chord progression; Obtain rhythmic features based on beat analysis and rhythmic patterns; Get dynamic and emoji features based on volume changes and emojis; Technical difficulty characteristics are obtained based on finger span, fast running and two-hand coordination.

[0025] S2: Integrate the music score features to obtain the music score difficulty features; In this embodiment, since different music score features may have different dimensions and value ranges, directly fusing these features may cause some features to dominate the final feature vector. Therefore, each feature needs to be standardized or normalized first to ensure that they are comparable in value.

[0026] Use the autoencoder in deep learning to learn a low-dimensional score feature representation that can capture the main information of the original features while removing redundancy and noise. Based on the fusion strategy of the autoencoder, transform the low-dimensional score feature representation and calculate the fused score difficulty feature The fused features are verified to check whether they can effectively reflect the difficulty of the score. In this embodiment, the fused features are verified by analyzing the correlation between the score difficulty feature and the target difficulty label and using the fused features (score difficulty feature) to train a preliminary model and evaluate its performance. If the fused features do not meet the preset requirements, it is necessary to return to the previous steps for adjustment, such as changing the feature selection method, adjusting the fusion strategy, or re-standardizing the features.

[0027] Once the fused features are verified to be effective, they can be used as input for subsequent music score difficulty recognition models. These fused features can fully and accurately reflect the difficulty level of the music score.

[0028] S3: Based on the difficulty features of the music score, a music score difficulty recognition model is constructed; a further implementation method is that based on the difficulty features of the music score, a method for constructing the music score difficulty recognition model is: The preset feature transformation matrix is ​​linearly combined with the score difficulty feature to obtain the feature projection; specifically, this process is actually a transformation of the original feature space, the purpose of which is to find a new feature space so that the score data of different difficulty levels can be more easily distinguished in this space.

[0029] Based on feature projection, the distance measure of the projection space is obtained; Based on the distance measure, obtain the feature projection matrix; Based on the feature projection matrix, the piano score is projected into a new feature space to obtain a new score difficulty feature; specifically, in this new feature space, a distance measure needs to be defined to measure the distance between different score samples. In this embodiment, Manhattan distance is selected as the distance measure, which is used to quantify the similarity or difference between samples.

[0030] Based on the neural network, the difficulty features of the new music scores are classified to obtain the difficulty of the piano scores and complete the construction of the music score difficulty recognition model.

[0031] A further implementation method is that the method for obtaining the feature projection matrix is: Based on the square of the distance measure, a new distance measure is obtained; the square of the original distance measure is taken to emphasize the difference in distance, so that during the optimization process, samples with farther distances have a greater impact on the objective function. This helps to pay more attention to those sample pairs that are originally difficult to distinguish during the optimization process.

[0032] The objective function is constructed based on the new distance measure; the objective function constructed using the new distance measure can reflect the performance of the model in distinguishing music scores of different difficulty levels. In this embodiment, the process of constructing the objective function is to maximize the inter-class distance while minimizing the intra-class distance, that is, to make music score samples of the same difficulty as close as possible in the projection space, while music score samples of different difficulty levels are as far away as possible.

[0033] The objective function is solved based on stochastic gradient descent to obtain the feature projection matrix.

[0034] A further embodiment is that the neural network structure includes an input layer, a hidden layer and an output layer; The number of nodes in the input layer matches the vector dimension of the new music score difficulty feature; The hidden layer uses the Sigmoid activation function, and each layer has a different number of nodes; The number of nodes in the output layer matches the number of difficulty categories, and the Softmax activation function is used to obtain the probability of each difficulty category. In this embodiment, the category with the highest probability is selected as the difficulty level of the music score.

[0035] In this embodiment, the cross entropy loss function is selected to evaluate the prediction performance of the model; gradient descent is used to optimize the network weights to minimize the loss function. The difficulty feature of the score is used as input, and the difficulty label is used as the target output, and the network is iteratively trained. In each iteration, the forward propagation calculates the prediction result, and the back propagation calculates the gradient and updates the weight.

[0036] S4: Input the piano music score to be identified into the music score difficulty recognition model, obtain the difficulty recognition result, and complete the recognition of the music score difficulty.

[0037] In this embodiment, specific difficulty levels include elementary, intermediate, and advanced, and a difficulty score is performed based on the difficulty level. The difficulty score is a value between 0 and 100. Example

[0038] The present invention also provides a piano score difficulty recognition system, which is used to implement a piano score difficulty recognition method, comprising: A music score feature acquisition module, used to collect music score data and extract music score features of the music score data; A music score difficulty feature acquisition module is used to fuse music score features to obtain music score difficulty features; A model building module, used to build a music score difficulty recognition model based on the music score difficulty characteristics; The difficulty recognition module is used to input the piano music score to be recognized into the music score difficulty recognition model, obtain the difficulty recognition result, and complete the recognition of the music score difficulty.

[0039] A further implementation method is that the music score features include note features, chord features, rhythm features, dynamic and emoticon features and technical difficulty features; the music score feature acquisition module includes: A note feature acquisition unit, used for acquiring note features based on note density, note duration and note span; A chord feature acquisition unit, used for acquiring chord features based on chord types and chord progressions; A rhythm feature acquisition unit, used for acquiring rhythm features based on beat analysis and rhythm patterns; A dynamic feature acquisition unit, used for acquiring dynamic and emoticon features based on volume changes and emoticons; The technical difficulty feature acquisition unit is used to obtain technical difficulty features based on finger span, fast running and two-hand coordination.

[0040] In a further embodiment, the model building module comprises: A feature projection unit, used for linearly combining a preset feature transformation matrix with a music score difficulty feature to obtain a feature projection; A distance measurement acquisition unit, used for obtaining a distance measurement of a projection space based on feature projection; A feature projection matrix acquisition unit, used for obtaining a feature projection matrix based on a distance measurement; A new feature acquisition unit, used to project the piano score into a new feature space based on a feature projection matrix to obtain a new score difficulty feature; The difficulty classification unit is used to classify the difficulty features of new music scores based on the neural network, obtain the difficulty of piano music scores, and complete the construction of the music score difficulty recognition model.

[0041] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for identifying the difficulty of piano music scores, characterized in that: The following steps are involved: Collecting music score data and extracting music score features of the music score data; Combining the music score features to obtain music score difficulty features; Based on the difficulty features of the music score, a music score difficulty recognition model is constructed; The piano music score to be identified is input into the music score difficulty identification model to obtain a difficulty identification result, thereby completing the identification of the music score difficulty.

2. A piano score difficulty identification method according to claim 1, characterized in that: The music score data includes notes, chords, rhythms, dynamic and expression symbols and technical difficulty; the music scores are of different composers, different music periods and styles.

3. A piano score difficulty identification method according to claim 2, characterized in that: The music score features include note features, chord features, rhythm features, dynamic and expression symbol features, and technical difficulty features; The method for extracting the music score features includes: Obtaining the note feature based on note density, note duration and note span; Obtaining the chord feature based on the chord type and the chord progression; Obtaining the rhythm feature based on beat analysis and rhythm pattern; Obtaining the dynamic and emoji features based on the volume change and the emoji; The technical difficulty characteristics are obtained based on finger span, fast running and two-hand coordination.

4. A piano score difficulty identification method according to claim 3, characterized in that: The method for obtaining the note feature is: The note density is calculated according to measures, beats or other time units; Extracting the duration of each note to obtain the duration of the note; Calculating the span of adjacent notes or notes within a specific range to obtain the note span; The note density, note duration and note span features are vectorized and concatenated to obtain the note feature.

5. A piano score difficulty identification method according to claim 3, characterized in that: Based on the score difficulty feature, the method for constructing a score difficulty recognition model is as follows: Linearly combining a preset feature transformation matrix with the music score difficulty feature to obtain a feature projection; Based on the feature projection, obtaining a distance measure of the projection space; Based on the distance measure, obtaining a feature projection matrix; Based on the feature projection matrix, the piano score is projected into a new feature space to obtain a new score difficulty feature; The new music score difficulty feature is classified based on a neural network to obtain the piano score difficulty, thereby completing the construction of the music score difficulty recognition model.

6. A piano score difficulty identification method according to claim 5, characterized in that: The method for obtaining the feature projection matrix is: Obtaining a new distance measure based on the square of the distance measure; Construct an objective function based on the new distance measure; The objective function is solved based on stochastic gradient descent to obtain a feature projection matrix.

7. A piano score difficulty identification method according to claim 5, characterized in that: The neural network structure includes an input layer, a hidden layer and an output layer; The number of nodes in the input layer matches the vector dimension of the new music score difficulty feature; The hidden layer uses a Sigmoid activation function, and each layer is set with a different number of nodes; The number of nodes in the output layer matches the number of difficulty categories, and the Softmax activation function is used to obtain the probability of each difficulty category.

8. A piano score difficulty recognition system, used to implement the piano score difficulty recognition method according to any one of claims 1 to 7, characterized in that: include: A music score feature acquisition module, used to collect music score data and extract music score features of the music score data; A music score difficulty feature acquisition module, used to fuse the music score features to obtain the music score difficulty feature; A model building module, used to build a music score difficulty recognition model based on the music score difficulty feature; The difficulty recognition module is used to input the piano music score to be recognized into the music score difficulty recognition model, obtain the difficulty recognition result, and complete the recognition of the music score difficulty.

9. The piano score difficulty identification system according to claim 8, characterized in that: The music score features include note features, chord features, rhythm features, dynamic and emoticon features, and technical difficulty features; the music score feature acquisition module includes: A note feature acquisition unit, used to obtain the note feature based on note density, note duration and note span; A chord feature acquisition unit, configured to acquire the chord feature based on the chord type and the chord progression; A rhythm feature acquisition unit, used for acquiring the rhythm feature based on beat analysis and rhythm pattern; A dynamic feature acquisition unit, used to obtain the dynamic and emoticon features based on volume changes and emoticons; The technical difficulty feature acquisition unit is used to obtain the technical difficulty feature based on finger span, fast running and two-hand coordination.

10. The piano score difficulty identification system according to claim 8, characterized in that: The model building module includes: A feature projection unit, used for linearly combining a preset feature transformation matrix with the music score difficulty feature to obtain a feature projection; A distance measurement acquisition unit, configured to obtain a distance measurement of a projection space based on the feature projection; A feature projection matrix acquisition unit, used to obtain a feature projection matrix based on the distance measure; A new feature acquisition unit is used to project the piano score to a new feature space based on the feature projection matrix to obtain a new score difficulty feature; The difficulty classification unit is used to classify the difficulty features of the new music score based on a neural network, obtain the difficulty of the piano score, and complete the construction of the music score difficulty recognition model.

Images