Classical music multi-mode emotion cognition method and system for full-scene teaching application

Abstract

The invention relates to the crossing field of artificial intelligence and art education, and provides a classical music multi-mode emotion cognition method and system for full-scene teaching application. The method comprises the following steps: S1, constructing a training sample and a test sample; s2, constructing an improved SVM classification model, and training the model by using the multi-modal emotion feature parameters to obtain optimal parameters of the model; s3, testing the test sample by using the improved SVM classification model; and S4, performing multi-modal artistic emotion feature recognition on the classical music data by using the improved SVM classification model. The system comprises a training sample and test sample construction unit, an improved SVM classification model construction and training unit and an emotion feature recognition unit. According to the method, the problems that the multi-modal artistic emotion feature samples of the teaching resources of the online classical music education major are not linearly separable, the samples are sparse, and the model parameter training is poor can be effectively solved.

Description

Technical Field

[0001] The present invention relates to the intersection of artificial intelligence and art education, and in particular to a multimodal emotion recognition method and system for classical music for full-scenario teaching applications. Background Art

[0002] Classical music education is a crucial component of quality-oriented education. It aims to cultivate students' aesthetic sense, appreciation, and creative abilities (expressiveness and creativity) through the acquisition of specialized knowledge and the observation and practice of classical music works. Online classical music education is a necessary path to overcome long-standing bottlenecks in the development of classical music education, such as inappropriate resource allocation, uneven development of the ecosystem, and the lack of shared teaching content and technology.

[0003] However, at present, most professional art colleges at home and abroad, as well as mainstream "online art education platforms", mostly follow the traditional teaching models such as "face-to-face" and "one-to-one" for art professional teaching - "online teaching" also means moving "recorded courses / live courses" from "offline" to "online". There is a lack of mature, standardized and systematic online art education professional course system and advanced teaching technology support. The exploration of art professional teaching resources based on the Internet (mobile Internet) and the application of advanced teaching technologies for full-scene teaching are limited. The teaching model (content / technology / scenario) is single and the teaching efficiency is low, which cannot adapt to the sustainable development of online classical music education.

[0004] Classical music possesses not only aesthetic qualities but also ideological attributes, embodying and materializing human emotions. Classical music can generally be categorized based on emotional perception and perception. Generally speaking, art encompasses four major categories: language arts, plastic arts, performing arts, and mixed arts. Artistic works encompass diverse forms such as literature, painting, music, dance, architecture, sculpture, drama, and film. Arts education is generally interdisciplinary, and specialized arts instruction differs from traditional cultural instruction. In addition to learning and understanding the specialized art knowledge taught by specialized arts teachers, learners must also learn and appreciate the emotions (artistic emotions) of artworks through observation or demonstration of specialized teaching, thereby enhancing their own artistic appreciation and artistic creation (artistic cognition / artistic expression). This requires face-to-face, or even one-on-one, inspiration and demonstration from specialized arts teachers, which is crucial for internet (mobile internet) learners. Therefore, the sustainable development of online art education requires not only the support of a mature, standardized and systematic "online art education professional course system (knowledge base system)", but also the support of art emotion cognition technology for online art education professional teaching resources to improve the emotional cognition and artistic creation ability of Internet learners.

[0005] Emotions and feelings are the human brain's perception of the relationship between objective things and subjective needs, expressed through facial expressions, body language, language, and voice. Emotions are categorized as basic and complex (complex emotions are often composed of combinations of basic emotions). Feelings are used to express feelings with profound and stable social significance, including moral, aesthetic, and rational feelings. Feelings are expressed through emotions, and cognitive processes play a key role in the generation of emotions. Affective computing, a field that intersects information science, cognitive science, neuroscience, and even social science, is a system built on neural networks and deep learning algorithms, applying knowledge engineering to simulate human thought processes. Accurately describing emotions is a key research topic in affective computing.

[0006] In recent years, research on emotion recognition technology at home and abroad has mainly focused on "intelligent question-answering systems", and related research includes question understanding, dialogue management, dialogue generation, dialogue evaluation, etc.; refined evaluation technology based on (learner) emotion cognition has become a hot topic in the research of smart education application technology in recent years, including process evaluation and final evaluation, but it needs to comprehensively consider factors such as learners, teachers, and teaching environment, capture and analyze learners' learning environment (teaching scenarios), establish feedback mechanisms and emotional communication mechanisms between learners and teachers, etc. to achieve refined evaluation, but the relevant research results are not very practical.

[0007] As an important branch of sentiment computing, multimodal art emotion recognition objects (for online art education professional teaching resources) are often not linearly separable, and traditional classifiers cannot be used for art emotion classification. Usually, online art education professional teaching resources (graphics / audio / video) formed after structured and knowledge-based processing contain rich artistic emotions and need to be processed by multimodal art emotion recognition (feature extraction / emotion classification) to determine the classification and sorting of professional knowledge and related knowledge, and continuously optimize and refine them to achieve multimodal resource reorganization and automatic updating to form an "online art education professional course system." In the process of extracting multimodal art emotion features for online art education professional teaching resources, due to the small number of samples and the fact that art emotion features are not linearly separable, the art emotion recognition rate decreases and the amount of calculation increases. Multimodal art emotion recognition technology can help solve this problem. Summary of the Invention

[0008] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a multimodal emotion recognition method and system for classical music for full-scene teaching applications, aiming to solve the above-mentioned and other potential problems of the existing technology in the multimodal art emotion recognition process of online art education.

[0009] The present invention adopts the following technical solutions:

[0010] On the one hand, the present invention provides a multimodal emotion recognition method for classical music for full-scenario teaching applications, comprising:

[0011] S1. Constructing training samples and testing samples of artistic emotion features of classical music;

[0012] S2. Constructing an improved SVM classification model, wherein the improved SVM classification model is used to extract multimodal emotion feature parameters of the artistic emotion feature training sample; training the model using the multimodal emotion feature parameters to obtain optimal parameters of the improved SVM classification model;

[0013] S3, using the improved SVM classification model obtained in step S2, testing the artistic emotion feature test sample in step S1;

[0014] S4. Use the improved SVM classification model tested in step S3 to perform multimodal artistic emotion feature recognition on classical music data.

[0015] According to any of the possible implementations described above, there is further provided an implementation, wherein the method further includes:

[0016] Structuring and knowledge-based processing of multi-source and heterogeneous classical music course data to form classical music professional teaching data;

[0017] Structural and knowledge-based processing of multi-source, heterogeneous classical music Internet big data to form classical music-related teaching data;

[0018] Using the tested improved SVM classification model, the artistic emotion feature recognition is performed on the classical music professional teaching data and the classical music related teaching data, and based on the artistic emotion feature recognition results, a classical music professional teaching course system that integrates the professional teaching data and the related teaching data is constructed and optimized;

[0019] According to the learner's professional foundation and emotional level, the classical music professional teaching curriculum system is matched to obtain the learner's classical music professional teaching curriculum.

[0020] Any of the possible implementations described above further provides an implementation, in which, in step S1, the classical music professional teaching data is preprocessed to eliminate redundant data.

[0021] For any of the possible implementations described above, a further implementation is provided, in which step S2 is specifically:

[0022] S21, constructing a multimodal art emotion feature set of the art emotion feature training samples of the classical music

[0023] {(x1,y1),(x2,y2),...(x n ,y n )},x i Represents the multimodal artistic emotion feature vector of the i-th training sample, y i Represents the multimodal artistic emotional feature category label of the i-th training sample;

[0024] S22, calculate the multimodal artistic emotional feature weights of the training samples according to the information gain method d = (d1, d2, ... d m ) T ;

[0025] S23, introducing the multimodal artistic emotion feature weights of step S22 into the Gaussian kernel function and the polynomial kernel function for correction, and using the two corrected kernel functions to form a hybrid kernel function;

[0026] S24. Calculate the optimal parameters of the improved SVM classification model.

[0027] In any of the possible implementations described above, a further implementation is provided, in step S24, using a momentum particle swarm optimization algorithm to find the optimal parameters of the improved SVM classification model.

[0028] For any of the possible implementations described above, a further implementation is provided, which uses a momentum particle swarm optimization algorithm to seek the optimal parameters of the improved SVM classification model, specifically:

[0029] X1. Set parameters: population size m, number of iterations, inertia weight w, acceleration constants C1, C2 and SVM parameter solution space;

[0030] X2. Initialization: Randomly initialize the speed and position of m particles in the search space, calculate the value of the initial fitness of the particles, and the P of each particle ibest Initialized to the current position of the particle, P gbest Initialized to the local optimal position of the particle f with the best fitness value among the moving particles, the algorithm uses the SVM classification accuracy under H-fold cross validation as the fitness function;

[0031] X3. Update the speed and position of particles: Update the local optimal position of each particle, calculate the fitness of each particle and compare it with P ibest Compare, if its value is better than P ibest , then update to the current position of the particle, otherwise remain unchanged;

[0032] X4, update the global optimal position, compare the updated P ibest With P gbest , if better than P gbestIf yes, update it, otherwise keep it unchanged;

[0033] X5. Determine whether the termination condition is met. If so, output the optimal parameter combination; otherwise, go to step X3.

[0034] As for any possible implementation described above, a further implementation is provided, in which in step X5, the termination condition is that the maximum number of iterations is reached or the obtained fitness value meets the specified accuracy.

[0035] As for any possible implementation described above, a further implementation is provided, in step S23, a Gaussian kernel function and a polynomial kernel function are linearly combined to obtain a modified kernel function.

[0036] On the other hand, the present invention also provides a classical music multimodal emotion recognition system for full-scenario teaching applications, the system is used to implement the above method, the system includes:

[0037] A training sample and test sample construction unit, used to construct artistic emotion feature training samples and artistic emotion feature test samples of classical music;

[0038] An improved SVM classification model construction and training unit is used to construct an improved SVM classification model, train the model using artistic emotion feature training samples, and obtain the optimal parameters of the improved SVM classification model;

[0039] The emotion feature recognition unit uses an improved SVM classification model to perform multimodal artistic emotion feature recognition on classical music data.

[0040] According to any of the possible implementations described above, there is further provided an implementation, wherein the system further includes:

[0041] The classical music professional teaching curriculum system optimization unit identifies the artistic emotion characteristics of classical music professional teaching data and classical music related teaching data, and optimizes the classical music professional teaching curriculum system that integrates professional teaching data and related teaching data based on the results of artistic emotion feature identification;

[0042] The learning matching unit matches the classical music professional teaching curriculum system according to the learner's professional foundation and emotional level, and obtains the learner's classical music professional teaching curriculum.

[0043] The beneficial effects of this invention are as follows: it can effectively solve the problems of linear inseparability, sample sparsity, and poor model parameter training in multimodal artistic emotion feature samples of online classical music education professional teaching resources. It helps to improve the multimodal artistic emotion recognition ability of online classical music education professional teaching resources and solve the problem caused by insufficient sample size of online classical music education professional teaching resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 The figure shows a schematic diagram of the multimodal artistic emotion feature extraction process in the embodiment.

[0045] Figure 2 Shown is a schematic diagram of the training / testing process of the improved SVM classification model in the embodiment. DETAILED DESCRIPTION

[0046] The following will describe in detail specific embodiments of the present invention with reference to the accompanying drawings. It should be noted that the technical features or combinations of technical features described in the following embodiments should not be considered isolated, and they can be combined with each other to achieve better technical effects.

[0047] Based on an improved support vector machine (ISVM) algorithm, the present invention realizes the extraction and recognition of multimodal art emotion features of online art education professional teaching resources, thereby improving the professional knowledge learning and art emotion cognition ability of internet learners. The present invention first pre-processes the results of the structured and knowledge-based processing of multi-source, heterogeneous, and cross-platform art professional course data (text / audio / video) from professional art colleges to realize multimodal art emotion information endpoint detection, thereby extracting the multimodal art emotion feature parameters of online art education professional teaching resources; then, dimensionality reduction processing is performed to eliminate redundant features; furthermore, training samples and test samples of the multimodal art emotion features of online art education professional teaching resources are constructed, and art emotion classification of online art education professional teaching resources is performed based on the multimodal art emotion features; finally, a kernel function is introduced to perform multimodal art emotion recognition on the multimodal art emotion feature test samples. The ISVM algorithm with the kernel function has good learning ability for small, high-dimensional, and nonlinear multimodal art emotion feature samples in online art education professional teaching resources. When the number of samples is small, the ISVM can still accurately identify the test samples and obtain the optimal solution under these conditions.

[0048] The "Improved Support Vector Machine (ISVM) Algorithm" disclosed in the present invention mainly plays the role of a classifier in the process of multimodal art emotion recognition (feature extraction and classification recognition) of online classical music education professional teaching resources. It adopts a "one-to-one" classification algorithm to construct multiple support vector machines based on improved kernel functions to achieve multimodal art emotion recognition of online classical music education professional teaching resources. For a single ISVM, first, the two art emotion categories to be distinguished are used to construct a training sample feature matrix and a test sample feature vector set, and then the SVM model is trained to perform category judgment on the test samples of online art education professional teaching resources. Figure 1 shown.

[0049] The present invention obtains a multimodal art emotion feature training sample set and a test sample set for online classical music education professional teaching resources, and then extracts the multimodal art emotion feature parameters of online art education professional teaching resources, and then uses Figure 1 The processes shown are classified.

[0050] The present invention provides a multimodal emotion recognition method for classical music for all-scenario teaching applications, specifically comprising:

[0051] S1. Constructing training samples and testing samples of artistic emotion features of classical music;

[0052] S2. Constructing an improved SVM classification model, wherein the improved SVM classification model is used to extract multimodal emotion feature parameters of the artistic emotion feature training sample; training the model using the multimodal emotion feature parameters to obtain optimal parameters of the improved SVM classification model;

[0053] S3, using the improved SVM classification model obtained in step S2, testing the artistic emotion feature test sample in step S1;

[0054] S4. Use the improved SVM classification model tested in step S3 to perform multimodal artistic emotion feature recognition on classical music data.

[0055] In a specific embodiment, the method further comprises:

[0056] Structuring and knowledge-based processing of multi-source and heterogeneous classical music course data to form classical music professional teaching data;

[0057] Structural and knowledge-based processing of multi-source, heterogeneous classical music Internet big data to form classical music-related teaching data;

[0058] Using the tested improved SVM classification model, the artistic emotion feature recognition is performed on the classical music professional teaching data and the classical music related teaching data, and based on the artistic emotion feature recognition results, a classical music professional teaching course system that integrates the professional teaching data and the related teaching data is constructed and optimized;

[0059] According to the learner's professional foundation and emotional level, the classical music professional teaching curriculum system is matched to obtain the learner's classical music professional teaching curriculum.

[0060] In a specific embodiment, in step S1, the classical music professional teaching data is preprocessed to eliminate redundant data.

[0061] In a specific embodiment, step S2 is specifically as follows:

[0062] S21, constructing a multimodal art emotion feature set of the art emotion feature training samples of the classical music

[0063] {(x1,y1),(x2,y2),...(x n ,y n )},x i Represents the multimodal artistic emotion feature vector of the i-th training sample, y i Represents the multimodal artistic emotional feature category label of the i-th training sample;

[0064] S22, calculate the multimodal artistic emotional feature weights of the training samples according to the information gain method d = (d1, d2, ... d m ) T ;

[0065] S23, introducing the multimodal artistic emotion feature weights of step S22 into the Gaussian kernel function and the polynomial kernel function for correction, and using the two corrected kernel functions to form a hybrid kernel function;

[0066] S24. Calculate the optimal parameters of the improved SVM classification model.

[0067] In a specific embodiment, in step S24, the momentum particle swarm optimization algorithm is used to find the optimal parameters of the improved SVM classification model.

[0068] In a specific embodiment, the momentum particle swarm optimization algorithm is used to find the optimal parameters of the improved SVM classification model, specifically:

[0069] X1. Set parameters: population size m, number of iterations, inertia weight w, acceleration constants C1, C2 and SVM parameter solution space;

[0070] X2. Initialization: Randomly initialize the speed and position of m particles in the search space, calculate the value of the initial fitness of the particles, and the P of each particle ibest Initialized to the current position of the particle, P gbest Initialized to the local optimal position of the particle f with the best fitness value among the moving particles, the algorithm uses the SVM classification accuracy under H-fold cross validation as the fitness function;

[0071] X3. Update the speed and position of particles: Update the local optimal position of each particle, calculate the fitness of each particle and compare it with P ibest Compare, if its value is better than P ibest , then update to the current position of the particle, otherwise remain unchanged;

[0072] X4, update the global optimal position, compare the updated P ibest With P gbest, if better than P gbest If yes, update it, otherwise keep it unchanged;

[0073] X5. Determine whether the termination condition is met. If so, output the optimal parameter combination; otherwise, go to step X3.

[0074] In a specific embodiment, in step X5, the termination condition is that the maximum number of iterations is reached or the obtained fitness value meets the specified accuracy.

[0075] In a specific embodiment, in step S23, the Gaussian kernel function and the polynomial kernel function are linearly combined to obtain a modified kernel function, such as Figure 2 shown.

[0076] An embodiment of the present invention provides a multimodal emotion recognition system for classical music for full-scenario teaching applications. The system is used to implement the above method, and the system includes:

[0077] A training sample and test sample construction unit, used to construct artistic emotion feature training samples and artistic emotion feature test samples of classical music;

[0078] An improved SVM classification model construction and training unit is used to construct an improved SVM classification model, train the model using artistic emotion feature training samples, and obtain the optimal parameters of the improved SVM classification model;

[0079] The emotion feature recognition unit uses an improved SVM classification model to perform multimodal artistic emotion feature recognition on classical music data.

[0080] In a specific embodiment, the system further comprises:

[0081] The classical music professional teaching curriculum system optimization unit identifies the artistic emotion characteristics of classical music professional teaching data and classical music related teaching data, and optimizes the classical music professional teaching curriculum system that integrates professional teaching data and related teaching data based on the results of artistic emotion feature identification;

[0082] The learning matching unit matches the classical music professional teaching curriculum system according to the learner's professional foundation and emotional level, and obtains the learner's classical music professional teaching curriculum.

[0083] The present invention pre-processes the online classical music education professional teaching resources formed by the structuring (normalization / systematization) and knowledge-based (disciplinary integration / professional classification) processing of multi-source, heterogeneous, and cross-platform art professional course data (graphics / audio / video) of (professional art colleges), realizes multi-modal art emotion information endpoint detection, and extracts multi-modal art emotion feature parameters of online classical music education professional teaching resources; then, dimensionality reduction processing is performed on the multi-modal art emotion features to eliminate (merge) redundant features; further, by reconstructing the online classical music education hybrid knowledge graph (knowledge graph fusion), the knowledge fusion of professional teaching resources and related teaching resources is realized; at the same time, multi-modal art emotion feature training samples and test samples of online classical music education professional teaching resources are constructed, emotion classification is performed on the multi-modal art emotion features, and the professional knowledge and related knowledge are determined. The recognition, classification, sorting, continuous optimization and continuous refinement are carried out to realize the reorganization and automatic updating of multimodal resources, forming an "online classical music education professional course system"; finally, based on the improved support vector machine (ISVM) algorithm, multimodal art emotion feature extraction is realized, and then particle swarm optimization is introduced into SVM for pre-training of multimodal art emotion feature samples of online art education professional teaching resources to realize multimodal art emotion recognition - the improved support vector machine (ISVM) algorithm with the introduction of kernel function has good learning ability for small sample, high dimensional, nonlinear multimodal art emotion feature samples in online art education professional teaching resources. When the number of samples of online art education professional teaching resources is small, the ISVM algorithm can still accurately identify the test samples and obtain the optimal solution. It can be used for small sample pattern classification and is suitable for multimodal art emotion recognition applications in online art education.

[0084] Support vector machines (SVM) are a classification algorithm developed based on statistical learning, outperforming traditional classifiers such as KNN, ANN, and HMM. Originally designed to solve linear classification problems between two classes of samples, traditional SVM algorithms have limitations. In actual nonlinear classification, they are prone to generating a large number of misclassified samples, resulting in poor classification results. For multimodal samples such as text, audio, and video in online art education resources, traditional SVM algorithms are not suitable for extracting multimodal artistic sentiment features.

[0085] In this invention, the SVM uses a nonlinear function to alter the distribution of input professional teaching resource samples, fundamentally improving their linear separability. The improved support vector machine (ISVM) algorithm, which incorporates a kernel function, maps the original linearly inseparable input samples of online art education professional teaching resources into a high-dimensional feature space, making them linearly separable. The SVM algorithm is then applied to this high-dimensional feature space to find the optimal hyperplane to separate the two classes of samples.

[0086] In addition, since the training results of the support vector machine for online classical music education professional teaching resource samples are only related to the support vector, this allows multimodal (graphics / audio / video) art emotion feature extraction of online classical music education professional teaching resource samples with fewer professional teaching resource (course) samples, and then obtains the multimodal art emotion feature set of online classical music art education professional teaching resource training samples and the multimodal art emotion feature set of test samples.

[0087] Traditional training methods easily cause the training results to fall into local optimal solutions. The present invention adopts the particle swarm optimization algorithm to make the results as global as possible. The momentum particle swarm optimization algorithm is an algorithm that optimizes the combination of hybrid kernel SVM parameters and obtains the optimal position by updating the fitness function.

[0088] Although several embodiments of the present invention have been described herein, those skilled in the art will appreciate that modifications may be made to the embodiments herein without departing from the spirit of the present invention. The above embodiments are merely exemplary and should not be used as limitations on the scope of the present invention.

Claims

1. A multimodal emotion recognition method for classical music for full-scenario teaching applications, characterized by: The method comprises: S1. Constructing training samples and testing samples of artistic emotion features of classical music; S2. Constructing an improved SVM classification model, wherein the improved SVM classification model is used to extract multimodal emotion feature parameters of the artistic emotion feature training sample; training the model using the multimodal emotion feature parameters to obtain optimal parameters of the improved SVM classification model; S3, using the improved SVM classification model obtained in step S2, testing the artistic emotion feature test sample in step S1; S4. Use the improved SVM classification model tested in step S3 to perform multimodal artistic emotion feature recognition on classical music data.

2. The multimodal emotion recognition method for classical music for full-scenario teaching applications according to claim 1 is characterized in that: The method further comprises: Structuring and knowledge-based processing of multi-source and heterogeneous classical music course data to form classical music professional teaching data; Structural and knowledge-based processing of multi-source, heterogeneous classical music Internet big data to form classical music-related teaching data; Using the tested improved SVM classification model, the artistic emotion feature recognition is performed on the classical music professional teaching data and the classical music related teaching data, and based on the artistic emotion feature recognition results, a classical music professional teaching course system that integrates the professional teaching data and the related teaching data is constructed and optimized; According to the learner's professional foundation and emotional level, the classical music professional teaching curriculum system is matched to obtain the learner's classical music professional teaching curriculum.

3. The multimodal emotion recognition method for classical music for full-scenario teaching applications according to claim 1 is characterized in that: In step S1, the classical music professional teaching data is preprocessed to eliminate redundant data.

4. The multimodal emotion recognition method for classical music for full-scenario teaching applications according to claim 1 is characterized in that: Step S2 is specifically as follows: S21, constructing a multimodal art emotion feature set {(x1, y1), (x2, y2), ... (x n ,y n )},x i Represents the multimodal artistic emotion feature vector of the i-th training sample, y i Represents the multimodal artistic emotional feature category label of the i-th training sample; S22, calculate the multimodal artistic emotional feature weights of the training samples according to the information gain method d = (d1, d2, ... d m ) T ; S23, introducing the multimodal artistic emotion feature weights of step S22 into the Gaussian kernel function and the polynomial kernel function for correction, and using the two corrected kernel functions to form a hybrid kernel function; S24. Calculate the optimal parameters of the improved SVM classification model.

5. The multimodal emotion recognition method for classical music for full-scenario teaching applications according to claim 4 is characterized in that: In step S24, the momentum particle swarm optimization algorithm is used to find the optimal parameters of the improved SVM classification model.

6. The multimodal emotion recognition method for classical music for full-scenario teaching applications according to claim 5 is characterized in that: The momentum particle swarm optimization algorithm is used to find the optimal parameters of the improved SVM classification model, specifically: X1. Set parameters: population size m, number of iterations, inertia weight w, acceleration constants C1, C2 and SVM parameter solution space; X2. Initialization: Randomly initialize the speed and position of m particles in the search space, calculate the value of the initial fitness of the particles, and the P of each particle ibest Initialized to the current position of the particle, P gbest Initialized to the local optimal position of the particle f with the best fitness value among the moving particles, the algorithm uses the SVM classification accuracy under H-fold cross validation as the fitness function; X3. Update the speed and position of particles: Update the local optimal position of each particle, calculate the fitness of each particle and compare it with P ibest Compare, if its value is better than P ibest , then update to the current position of the particle, otherwise remain unchanged; X4, update the global optimal position, compare the updated P ibest With P gbest , if better than P gbest If yes, update it, otherwise keep it unchanged; X5. Determine whether the termination condition is met. If so, output the optimal parameter combination; otherwise, go to step X3.

7. The multimodal emotion recognition method for classical music for full-scenario teaching applications according to claim 6 is characterized in that: In step X5, the termination condition is that the maximum number of iterations is reached or the obtained fitness value meets the specified accuracy.

8. The multimodal emotion recognition method for classical music for full-scenario teaching applications according to claim 4 is characterized in that: In step S23, the Gaussian kernel function and the polynomial kernel function are linearly combined to obtain a modified kernel function.

9. A multimodal emotion recognition system for classical music for all-scenario teaching applications, characterized by: The system is used to implement the method according to any one of claims 1 to 8, and the system includes: A training sample and test sample construction unit, used to construct artistic emotion feature training samples and artistic emotion feature test samples of classical music; An improved SVM classification model construction and training unit is used to construct an improved SVM classification model, train the model using artistic emotion feature training samples, and obtain the optimal parameters of the improved SVM classification model; The emotion feature recognition unit uses an improved SVM classification model to perform multimodal artistic emotion feature recognition on classical music data.

10. The multimodal emotion recognition system for classical music for full-scenario teaching applications according to claim 9, characterized in that: The system further comprises: The classical music professional teaching curriculum system optimization unit identifies the artistic emotion characteristics of classical music professional teaching data and classical music related teaching data, and optimizes the classical music professional teaching curriculum system that integrates professional teaching data and related teaching data based on the results of artistic emotion feature identification; The learning matching unit matches the classical music professional teaching curriculum system according to the learner's professional foundation and emotional level, and obtains the learner's classical music professional teaching curriculum.

Images