Emotion recognition method and system based on artificial intelligence

Abstract

The invention discloses an emotion recognition method and system based on artificial intelligence, and relates to the technical field of emotion recognition, and the method comprises the steps: obtaining user data, carrying out the preprocessing and feature extraction, and constructing an emotion recognition model for user emotion recognition; matching a personalized emotion adjustment strategy based on the user emotion recognition result, and performing user emotion prediction according to the emotion adjustment strategy; and optimizing a personalized emotion adjustment strategy of the user according to the emotion prediction result of the user, and storing the emotion prediction result in a database. According to the method, the user emotion feature potential field is extracted by collecting the user data, the emotion recognition model is constructed for user emotion recognition, the accuracy of user emotion recognition is greatly improved, the user emotion regulation strategy is matched through the user emotion recognition result and optimized, the pertinence of the emotion regulation strategy is effectively improved, and the user emotion recognition accuracy is improved. Therefore, the user is helped to carry out accurate emotion regulation.

Description

Technical Field

[0001] The present invention relates to the technical field of emotion recognition, and in particular to an emotion recognition method and system based on artificial intelligence. Background Art

[0002] With the rapid development of artificial intelligence technology, affective computing has become one of the important research directions in the field of artificial intelligence. Emotion recognition technology judges the user's emotional state in real time by analyzing multimodal data such as facial expressions, voice, text and physiological signals, thereby realizing the intelligent and emotional human-computer interaction. In recent years, emotion recognition methods based on deep learning have gradually become the mainstream technical route due to their efficient feature extraction capabilities and accurate classification performance. For example, convolutional neural networks (CNN) and recurrent neural networks (RNN) are widely used in the extraction and analysis of facial expressions and voice emotion features; natural language processing (NLP) technology realizes the deep analysis of text emotion semantics through pre-trained models (such as BERT). In addition, the progress of multimodal data fusion technology has further improved the adaptability of emotion recognition systems to complex scenarios. The application areas of these technologies have covered multiple industries such as intelligent customer service, education and training, mental health, and gaming and entertainment, showing broad application prospects. However, existing technologies still have defects. Most existing technologies use multimodal data for emotion recognition, but existing multimodal fusion algorithms are often limited by model complexity and computing resource bottlenecks, making it difficult to achieve breakthroughs in real-time and accuracy. They also lack dynamic prediction of emotion change trends and personalized matching of regulation strategies, and cannot fully meet the actual needs of users. Summary of the invention

[0003] In view of the problems existing in the above-mentioned existing artificial intelligence-based emotion recognition methods and systems, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is that the accuracy of emotion recognition is insufficient, and there is a lack of dynamic prediction of emotion change trends and personalized matching of regulation strategies.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: an emotion recognition method based on artificial intelligence, which includes obtaining user data for preprocessing and feature extraction, using a complex superpotential function to perform feature mapping to generate a feature potential field based on the extracted features, and constructing an emotion recognition model to perform user emotion recognition; extracting an emotion feature vector based on the user emotion recognition result, using a Gaussian potential energy function to calculate the coarse-grained variables of the emotion feature vector potential energy mapping, and converting the coarse-grained variables into fine-grained emotion features, matching personalized emotion regulation strategies based on the coarse-grained variables and the fine-grained emotion features, and predicting user emotions based on the emotion regulation strategies; optimizing the user's personalized emotion regulation strategy based on the user emotion prediction results, and storing the emotion prediction results in a database.

[0006] As a preferred solution of the emotion recognition method based on artificial intelligence described in the present invention, wherein: the acquisition of user data for preprocessing and feature extraction, and the use of complex hyperpotential functions for feature mapping to generate a feature potential field according to the extracted features refer to collecting user facial images, voice data and text data for preprocessing, graying the user facial images and performing histogram equalization, and using convolutional neural networks to extract facial features. , and use complex hyperpotential functions to process facial features :

[0007] in For facial features The complex superpotential function, k and q are potential field parameters, i is an imaginary unit; Use MFCC to extract speech features of user voice data , describing speech features through complex hyperpotential functions Frequency changes:

[0008] in For voice features The complex superpotential function of Segment user text data into text segments and extract text features of user text data through BERT , through the complex superpotential function, the text features To map:

[0009] in Text features The complex superpotential function of n is the total number of text fragments, is the weight of the i-th text segment; Generate characteristic potential field based on complex superpotential function:

[0010] in is the characteristic potential field, is the complex superpotential function, is the first-order derivative of the complex superpotential function; According to facial features , voice features And text features Generate facial features using complex superpotential functions , voice features And text features The characteristic potential field , as well as , and then fused to construct a unified characteristic potential field :

[0011] in is the i-th characteristic potential field, including , as well as , is the fusion weight of the i-th characteristic potential field.

[0012] As a preferred solution of the emotion recognition method based on artificial intelligence described in the present invention, wherein: the construction of the emotion recognition model for user emotion recognition refers to constructing a convolutional neural network including an input layer, a convolution layer, a pooling layer and a fully connected layer, setting the input layer input to a unified feature potential field, and outputting the user emotion recognition result; Use the training set to train the convolutional neural network, and use the cross entropy loss function and Adam optimizer to iteratively optimize the convolutional neural network parameters; The user's unified characteristic potential field Input the trained convolutional neural network to obtain the user emotion recognition results.

[0013] As a preferred solution of the emotion recognition method based on artificial intelligence described in the present invention, wherein: the emotion feature vector is extracted based on the user emotion recognition result, the coarse-grained variable is mapped using the Gaussian potential energy function to calculate the potential energy of the emotion feature vector, and the coarse-grained variable is converted into a fine-grained emotion feature, and the personalized emotion regulation strategy is matched according to the coarse-grained variable and the fine-grained emotion feature, which means extracting the probability of various emotion recognitions from the user emotion recognition result to form an emotion feature vector E, and obtaining historical user data to extract historical emotion feature vectors ; The Gaussian potential energy function is used to calculate the potential energy U of the emotion feature vector, and the emotion feature vector is mapped to the coarse-grained variable s according to the potential energy U:

[0014] in is the reciprocal of the thermodynamic temperature, is the Dirac function; Define the MLP multilayer perceptron as the generator, set the initial parameters of the generator and the random variable z, where z conforms to the standard normal distribution, and convert the coarse-grained variable s into a fine-grained emotional feature R based on the generator:

[0015] Define the generative distribution of R based on the coarse-grained variable s and the fine-grained emotional feature R :

[0016] in is the prior distribution of the random variable z; Using historical sentiment feature vectors Calculate historical coarse-grained variables and convert them into historical fine-grained sentiment features , calculate historical fine-grained sentiment features The generated distribution and KL discretization are used to measure the historical fine-grained sentiment features The difference between the generated distribution and the true distribution is used as the loss, the generator parameters are iteratively optimized through the gradient descent method, and the loss is iteratively calculated, and the generator parameters that minimize the loss are selected as the optimal parameters; Generate fine-grained emotion features using the optimized generator based on the emotion feature vector E ; Collect emotion regulation activity data, define each emotion regulation activity as an emotion regulation strategy A, obtain the use records of emotion regulation strategies, extract the fine-grained emotion features before and after regulation, and calculate the fine-grained emotion features Cosine similarity with the fine-grained emotion features before adjustment , and calculate the cosine similarity of the fine-grained emotional features before and after adjustment ; Define the loss function L as cosine similarity and The ratio of , calculates the probability p(A) of the emotion regulation strategy according to the loss function:

[0017] in is a set of emotion regulation strategies, is the emotion regulation strategy in the emotion regulation strategy set, and They are emotion regulation strategies A and The loss function of The emotion regulation strategy that maximizes p(A) is taken as the final emotion regulation strategy.

[0018] As a preferred solution of the emotion recognition method based on artificial intelligence of the present invention, wherein: the prediction of user emotions according to the emotion regulation strategy refers to constructing a DNN network, setting the input layer as the emotion regulation strategy and the emotion feature vector, the output is the predicted user emotion recognition result, and using the training set to train the DNN network; The final emotion regulation strategy and user emotion feature vector E are input into the trained DNN network to obtain the predicted user emotion recognition results.

[0019] As a preferred embodiment of the artificial intelligence-based emotion recognition method described in the present invention, wherein: the optimization of the user's personalized emotion regulation strategy based on the user emotion prediction result refers to calculating the absolute value of the difference between the probability of emotion recognition of each category based on the user emotion recognition result and the predicted user emotion recognition result, performing weighted summation to obtain an emotion regulation value, setting an adjustment threshold, and if the emotion regulation value is greater than the adjustment threshold, recording the user's emotion regulation strategy and implementing it; if the emotion regulation value is less than or equal to the adjustment threshold, re-optimizing the generator to select the emotion regulation strategy until the emotion regulation value is greater than the adjustment threshold.

[0020] As a preferred solution of the artificial intelligence-based emotion recognition method described in the present invention, wherein: storing the emotion prediction results in a database refers to storing the predicted user emotion recognition results and the user emotion recognition results in the database, synchronously obtaining the user data after the implementation of the user emotion regulation strategy for storage, and the database regularly uploads the stored data to the cloud for backup.

[0021] Another object of the present invention is to provide an artificial intelligence-based emotion recognition system for performing emotion recognition based on any of the artificial intelligence-based emotion recognition methods as described above, wherein the artificial intelligence-based emotion recognition system at least comprises: a data acquisition end, a data calculation and storage end, and a data storage end, wherein the data calculation and storage end is configured with an edge computing box, and the edge computing box is configured with an AI algorithm for calculating emotion recognition; wherein: The data collection end is used to obtain user data, preprocess and extract features of the user data, and build an emotion recognition model to perform user emotion recognition; The data calculation and storage end is used to match the personalized emotion regulation strategy based on the user emotion recognition result, and perform user emotion prediction according to the emotion regulation strategy; The data storage end is used to optimize the user's personalized emotion regulation strategy according to the user emotion prediction result, and store the emotion prediction result in a preset database.

[0022] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of the above-mentioned artificial intelligence-based emotion recognition method when executing the computer program.

[0023] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned emotion recognition method based on artificial intelligence.

[0024] The beneficial effects of the present invention are as follows: the present invention collects user data to extract user emotion characteristic potential fields, and constructs an emotion recognition model to perform user emotion recognition, thereby greatly improving the accuracy of user emotion recognition. The present invention matches and optimizes user emotion regulation strategies through user emotion recognition results, thereby effectively improving the pertinence of emotion regulation strategies, thereby helping users to accurately regulate their emotions. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0026] Figure 1 It is a flowchart of an emotion recognition method based on artificial intelligence; Figure 2 Schematic diagram of the structure of the emotion recognition system based on artificial intelligence. DETAILED DESCRIPTION

[0027] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0028] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0029] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0030] Example 1 Reference Figure 1, which is the first embodiment of the present invention, and which provides an emotion recognition method based on artificial intelligence. The emotion recognition method based on artificial intelligence includes: S1. Obtain user data for preprocessing and feature extraction, use complex hyperpotential function to perform feature mapping based on the extracted features to generate a feature potential field, and build an emotion recognition model to perform user emotion recognition; Specifically, obtaining user data for preprocessing and feature extraction, using complex hyperpotential function for feature mapping to generate feature potential field according to the extracted features means collecting user facial images, voice data and text data for preprocessing, graying the user facial images and performing histogram equalization, and extracting facial features using principal component analysis. , and use complex hyperpotential functions to process facial features :

[0031] in For facial features The complex superpotential function, k and q are potential field parameters, i is an imaginary unit; Describes the periodic damping phenomenon, which shows rapid oscillation and gradually weakens. Represents periodic spikes in dynamic amplitude, It has a symmetrical oscillation characteristic with rapid decay, highlights the change of peak amplitude, is suitable for capturing local features, is sensitive to high-frequency features, and is suitable for capturing rapid fluctuations. Facial expression features usually contain local rapid changes, such as raised eyebrows, twitching of the corners of the mouth, etc. These features require high-frequency capture capabilities. Can effectively reflect the subtle features of facial expression changes; Use MFCC to extract speech features of user voice data , describing speech features through complex hyperpotential functions Frequency changes:

[0032] in For voice features The complex superpotential function of Describes asymmetric amplitude growth, suitable for capturing gradually increasing dynamic changes. Indicates the amplitude increment of periodic fluctuations, enhancing the response to high-frequency features. It has a gradually increasing asymmetric oscillation characteristic, highlights the gradual amplitude change, is suitable for capturing the overall trend, is more sensitive to low-frequency features, and is suitable for describing slow-changing features. Speech data is mainly manifested as continuous changes in the spectrum, including gradually increasing or decreasing characteristics. The smooth change of its dynamic amplitude requires low-frequency response capabilities. Can better describe the continuous changes in speech frequency; Segment user text data into text segments and extract text features of user text data through BERT , through the complex superpotential function, the text features To map:

[0033] in Text features The complex superpotential function of n is the total number of text fragments, is the weight of the i-th text segment; pass Smooth the emotion weights to avoid the interference of the sudden change of a single segment weight on the overall emotion, complex superpotential function The weighted summation feature on the components can naturally support multi-category sentiment analysis, and enhance the ability to capture sentiment trends through the emotional weight synthesis of semantic segments. It has the advantages of stability and significance in sentiment feature extraction; Generate characteristic potential field based on complex superpotential function:

[0034] in is the characteristic potential field, is the complex superpotential function, is the first-order derivative of the complex superpotential function; According to facial features , voice features And text features Generate facial features using complex superpotential functions , voice features And text features The characteristic potential field , as well as , and then fused to construct a unified characteristic potential field :

[0035] in is the i-th characteristic potential field, including , as well as , is the fusion weight of the i-th characteristic potential field, which is determined empirically.

[0036] The feature potential field maps the local dynamic changes of the complex hyperpotential function into the global potential field, further improving the dimension and dynamic characteristics of feature expression, and generating independent potential field representations for facial features, speech features and text features respectively, which is convenient for the subsequent fusion of multimodal emotion features. Through weighted fusion, the fusion weight of the feature potential field ensures the reasonable distribution of the contribution of different modal features to the unified feature potential field. For example, when the weight of facial features is higher, facial expressions have a greater influence in emotion recognition, and a unified feature potential field is generated to provide efficient and dynamic emotion feature input for the emotion recognition model. The emotion recognition model based on the unified feature potential field can more accurately identify user emotions, especially the significant complementary effect between modalities under complex emotional states.

[0037] Furthermore, constructing an emotion recognition model to perform user emotion recognition refers to constructing a convolutional neural network including an input layer, a convolution layer, a pooling layer, and a fully connected layer, setting the input layer input as a unified feature potential field, and outputting the user emotion recognition result; Use the training set to train the convolutional neural network, and use the cross entropy loss function and Adam optimizer to iteratively optimize the convolutional neural network parameters; The user's unified characteristic potential field The user emotion recognition results are obtained by inputting the trained convolutional neural network. On the basis of multimodal fusion, the unified feature potential field reduces the information redundancy between modalities and improves the input quality of the model. CNN realizes the accurate classification of high-dimensional input feature vectors to emotion categories through multi-layer feature extraction and nonlinear mapping. The output emotion recognition results have high real-time and accuracy and can be directly applied to emotion regulation and prediction tasks. The multimodal emotion features are integrated into a unified feature potential field, which significantly improves the input quality of the emotion recognition model and the correlation of multimodal information. Combined with the cross entropy loss function and Adam optimizer, fast and efficient model parameter optimization is achieved, and the real-time performance of emotion recognition is enhanced. Through the hierarchical feature extraction of CNN, the ability of emotion dynamic analysis from local to global is constructed to adapt to the needs of diverse emotional scenarios. The unified feature potential field is used as a high-dimensional representation of multimodal emotion features, which retains the global dynamic information of emotions.

[0038] S2. Extract emotion feature vectors based on the user emotion recognition results, use Gaussian potential energy function to calculate the potential energy mapping of emotion feature vectors to coarse-grained variables, and convert the coarse-grained variables into fine-grained emotion features, match personalized emotion regulation strategies according to the coarse-grained variables and the fine-grained emotion features, and predict user emotions according to the emotion regulation strategies; Specifically, based on the user emotion recognition results, the emotion feature vector is extracted, the Gaussian potential function is used to calculate the potential energy mapping of the emotion feature vector, and the coarse-grained variables are converted into fine-grained emotion features. The personalized emotion regulation strategy is matched according to the coarse-grained variables and the fine-grained emotion features. The probability of various emotion recognitions is extracted from the user emotion recognition results to form an emotion feature vector E, and the historical user data is obtained to extract the historical emotion feature vector ; Use the Gaussian potential energy function to calculate the emotional feature vector potential U:

[0039] in is the recognition probability of the i-th emotion in the emotion feature vector, is the mean recognition probability of all emotions in the emotion feature vector, and m is the emotion type of the emotion feature vector; The emotion feature vector is mapped to the coarse-grained variable s according to the potential energy U:

[0040] in is the reciprocal of the thermodynamic temperature, is the Dirac function; Define the MLP multilayer perceptron as the generator, set the initial parameters of the generator and the random variable z, where z conforms to the standard normal distribution, and convert the coarse-grained variable s into a fine-grained emotional feature R based on the generator:

[0041] Define the generative distribution of R based on the coarse-grained variable s and the fine-grained emotional feature R :

[0042] in is the prior distribution of the random variable z; Using historical sentiment feature vectors Calculate historical coarse-grained variables and convert them into historical fine-grained sentiment features , calculate historical fine-grained sentiment features The generated distribution and KL discretization are used to measure the historical fine-grained sentiment features The difference between the generated distribution and the true distribution is used as the loss, the generator parameters are iteratively optimized through the gradient descent method, and the loss is iteratively calculated, and the generator parameters that minimize the loss are selected as the optimal parameters; Generate fine-grained emotion features using the optimized generator based on the emotion feature vector E ; Collect emotion regulation activity data, define each emotion regulation activity as an emotion regulation strategy A, obtain the use records of emotion regulation strategies, extract the fine-grained emotion features before and after regulation, and calculate the fine-grained emotion features Cosine similarity with the fine-grained emotion features before adjustment , and calculate the cosine similarity of the fine-grained emotional features before and after adjustment ; Define the loss function L as cosine similarity and The ratio of , calculates the probability p(A) of the emotion regulation strategy according to the loss function:

[0043] in is a set of emotion regulation strategies, is the emotion regulation strategy in the emotion regulation strategy set, and They are emotion regulation strategies A and The loss function of The emotion regulation strategy that maximizes p(A) is taken as the final emotion regulation strategy.

[0044] The emotion feature vector retains the multi-category probability distribution of user emotion recognition, providing basic data for subsequent emotion regulation strategy matching. After extracting the emotion feature vector, the Gaussian potential energy function is combined to quantify the emotion state, providing support for diverse emotion regulation needs. The Gaussian potential energy function can intuitively reflect the complexity and dynamics of emotions by quantifying the discrete degree of the probability distribution of the emotion feature vector. The calculation of the potential energy value provides a scientific basis for the generation of coarse-grained variables, ensuring that the generated variables have biological and emotional significance. The coarse-grained variables are calculated by combining the potential energy function and the inverse of thermodynamics, combined with the Dirac function, which can convert the emotion characteristics into The complex information of the quantity is compressed into a low-dimensional representation. The mapped coarse-grained variables retain the main characteristics of the emotional features, laying the foundation for generating high-precision fine-grained emotional features. The KL divergence measures the difference between the generated distribution and the true distribution, and provides an objective function for the optimization of the generator parameters. The generator parameters are iteratively updated by the gradient descent method, which can gradually reduce the distribution difference and improve the generation accuracy of fine-grained emotional features. By calculating the cosine similarity of fine-grained emotional features, the actual effect of the emotion regulation strategy on the user's emotional state is quantified. The similarity calculation of the features before and after adjustment can directly reflect the effectiveness of the strategy, providing a scientific basis for strategy selection.

[0045] Furthermore, predicting user emotions according to the emotion regulation strategy refers to constructing a DNN network, setting the input layer as the emotion regulation strategy and the emotion feature vector, and the output as the predicted user emotion recognition result, and using the training set to train the DNN network; The final emotion regulation strategy and user emotion feature vector E are input into the trained DNN network to obtain the predicted user emotion recognition results.

[0046] Combining emotion regulation strategies with emotion feature vectors, the model can capture the complex relationship between the user's emotional state and regulation strategies. The multi-layer neural network gradually extracts high-dimensional features from the data and constructs an abstract representation of changes in emotional states. By outputting emotion recognition results, the model provides a high-precision, multi-category classification scheme for emotion state prediction. By using large-scale historical training data, the DNN model's adaptability to different emotion regulation strategies and user emotional characteristics is improved. Through emotion prediction, the system can judge the trend of emotion changes in advance and provide a basis for further emotion intervention. The efficient computing power of DNN ensures that emotion prediction results can be generated in real time to meet users' dynamic emotion management needs.

[0047] S3, optimizing the user's personalized emotion regulation strategy according to the user's emotion prediction results, and storing the emotion prediction results in a database; Specifically, optimizing the user's personalized emotion regulation strategy according to the user emotion prediction result means calculating the absolute value of the difference between the probability of emotion recognition of each category according to the user emotion recognition result and the predicted user emotion recognition result, performing weighted summation to obtain the emotion regulation value, setting the regulation threshold, and if the emotion regulation value is greater than the regulation threshold, recording the user's emotion regulation strategy and implementing it; if the emotion regulation value is less than or equal to the regulation threshold, re-optimizing the generator to select the emotion regulation strategy until the emotion regulation value is greater than the regulation threshold.

[0048] The emotion regulation value reflects the deviation between the user's emotional state and the predicted state through difference calculation, and has a clear physical meaning. By setting the adjustment threshold, unnecessary emotion regulation can be reduced and the system efficiency can be improved. The threshold judgment mechanism ensures timely response to changes in user emotions. The optimization process of the generator is combined with user feedback to ensure that the emotion regulation strategy can accurately improve the emotional state, provide an efficient strategy generation mechanism, and provide intelligent support for user emotion management. Real-time monitoring of the regulation effect provides data support for further optimization of the generator.

[0049] Furthermore, storing the emotion prediction results in the database means storing the predicted user emotion recognition results and the user emotion recognition results in the database, synchronously acquiring the user data after the implementation of the user emotion regulation strategy for storage, and the database regularly uploading the stored data to the cloud for backup.

[0050] Example 2 Reference Figure 2, which is the second embodiment of the present invention, which is different from the previous embodiment, and provides an artificial intelligence-based emotion recognition system for performing emotion recognition based on any artificial intelligence-based emotion recognition method as described above, wherein the artificial intelligence-based emotion recognition system at least includes: a data acquisition end, a data calculation and storage end, and a data storage end, wherein the data calculation and storage end is configured with an edge computing box, and the edge computing box is configured with an AI algorithm for calculating emotion recognition; wherein: The data collection end is used to obtain user data, preprocess and extract features of the user data, and build an emotion recognition model to perform user emotion recognition; The data calculation and storage end is used to match the personalized emotion regulation strategy based on the user emotion recognition result, and perform user emotion prediction according to the emotion regulation strategy; The data storage end is used to optimize the user's personalized emotion regulation strategy according to the user emotion prediction result, and store the emotion prediction result in a preset database.

[0051] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0052] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0053] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0054] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit with a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit with a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0055] Example 3 In order to verify the beneficial effects of the method of the present invention, experimental data are used for demonstration, and the experimental data are shown in Table 1 and Table 2: Table 1: Emotion recognition effect display table of the present invention method and the prior art

[0056] Table 2: Display of the adjustment effect of the method of the present invention and the prior art

[0057] Through the above experimental data, it can be concluded that the method of the present invention has obvious advantages over the existing technical content in terms of emotion recognition and emotion regulation. In practical applications, the method of the present invention can effectively improve the quality and effect of emotion regulation and provide a more advanced solution for personalized emotion management.

[0058] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An emotion recognition method based on artificial intelligence, characterized in that: include, Obtain user data for preprocessing and feature extraction, use complex hyperpotential function to perform feature mapping based on the extracted features to generate a feature potential field, and build an emotion recognition model to perform user emotion recognition; Extract emotion feature vectors based on the results of user emotion recognition, use Gaussian potential energy function to calculate the potential energy mapping of emotion feature vectors to coarse-grained variables, and convert the coarse-grained variables into fine-grained emotion features. Match personalized emotion regulation strategies based on coarse-grained variables and fine-grained emotion features, and predict user emotions based on the emotion regulation strategies. The user's personalized emotion regulation strategy is optimized according to the user's emotion prediction results, and the emotion prediction results are stored in the database.

2. The emotion recognition method based on artificial intelligence as claimed in claim 1, characterized in that: The method of obtaining user data for preprocessing and feature extraction, and using complex hyperpotential function to perform feature mapping to generate feature potential field according to the extracted features refers to collecting user facial images, voice data and text data for preprocessing, graying and histogram equalization of user facial images, and extracting facial features using principal component analysis. , and use complex hyperpotential functions to process facial features : ； in For facial features The complex superpotential function, k and q are potential field parameters, i is an imaginary unit; Use MFCC to extract speech features of user voice data , describing speech features through complex hyperpotential functions Frequency changes: ； in For voice features The complex superpotential function of Segment user text data into text segments and extract text features of user text data through BERT , through the complex superpotential function, the text features To map: ； in Text features The complex superpotential function of n is the total number of text fragments, is the weight of the i-th text segment; Generate characteristic potential fields based on complex superpotential functions: ； in is the characteristic potential field, is the complex superpotential function, is the first-order derivative of the complex superpotential function; According to facial features , voice features And text features Generate facial features using complex superpotential functions , voice features And text features The characteristic potential field , as well as , and then fused to construct a unified characteristic potential field : ； in is the i-th characteristic potential field, including , as well as , is the fusion weight of the i-th characteristic potential field.

3. The emotion recognition method based on artificial intelligence as claimed in claim 2, characterized in that: The construction of the emotion recognition model for user emotion recognition refers to constructing a convolutional neural network including an input layer, a convolution layer, a pooling layer, and a fully connected layer, setting the input layer input as a unified feature potential field, and outputting the user emotion recognition result; Use the training set to train the convolutional neural network, and use the cross entropy loss function and Adam optimizer to iteratively optimize the convolutional neural network parameters; The user's unified characteristic potential field Input the trained convolutional neural network to obtain the user emotion recognition results.

4. The emotion recognition method based on artificial intelligence as claimed in claim 3, characterized in that: The method extracts the emotion feature vector based on the user emotion recognition result, uses the Gaussian potential energy function to calculate the emotion feature vector potential map coarse-grained variables, and converts the coarse-grained variables into fine-grained emotion features. The personalized emotion regulation strategy is matched according to the coarse-grained variables and the fine-grained emotion features. It refers to extracting various emotion recognition probabilities from the user emotion recognition result to form an emotion feature vector E, and obtaining historical user data to extract historical emotion feature vectors. ; The Gaussian potential energy function is used to calculate the potential energy U of the emotion feature vector, and the emotion feature vector is mapped to the coarse-grained variable s according to the potential energy U: ； in is the reciprocal of the thermodynamic temperature, is the Dirac function; Define the MLP multilayer perceptron as the generator, set the initial parameters of the generator and the random variable z, where z conforms to the standard normal distribution, and convert the coarse-grained variable s into a fine-grained emotional feature R based on the generator: ； Define the generative distribution of R based on the coarse-grained variable s and the fine-grained emotional feature R : ； in is the prior distribution of the random variable z; Using historical sentiment feature vectors Calculate historical coarse-grained variables and convert them into historical fine-grained sentiment features , calculate historical fine-grained sentiment features The generated distribution and KL discretization are used to measure the historical fine-grained sentiment features The difference between the generated distribution and the true distribution is used as the loss, the generator parameters are iteratively optimized through the gradient descent method, and the loss is iteratively calculated, and the generator parameters that minimize the loss are selected as the optimal parameters; Generate fine-grained emotion features using the optimized generator based on the emotion feature vector E ; Collect emotion regulation activity data, define each emotion regulation activity as an emotion regulation strategy A, obtain the use records of emotion regulation strategies, extract the fine-grained emotion features before and after regulation, and calculate the fine-grained emotion features Cosine similarity with the fine-grained emotion features before adjustment , and calculate the cosine similarity of the fine-grained emotional features before and after adjustment ; Define the loss function L as cosine similarity and The ratio of , calculates the probability p(A) of the emotion regulation strategy according to the loss function: ； in is a set of emotion regulation strategies. is the emotion regulation strategy in the emotion regulation strategy set, and They are emotion regulation strategies A and The loss function of The emotion regulation strategy that maximizes p(A) is taken as the final emotion regulation strategy.

5. The emotion recognition method based on artificial intelligence as claimed in claim 4, characterized in that: The user emotion prediction according to the emotion regulation strategy refers to constructing a DNN network, setting the input layer as the emotion regulation strategy and the emotion feature vector, the output as the predicted user emotion recognition result, and using the training set to train the DNN network; The final emotion regulation strategy and user emotion feature vector E are input into the trained DNN network to obtain the predicted user emotion recognition results.

6. The emotion recognition method based on artificial intelligence as claimed in claim 5, characterized in that: The optimizing the personalized emotion regulation strategy of the user according to the user emotion prediction result refers to calculating the absolute value of the difference between the probability of emotion recognition of each category according to the user emotion recognition result and the predicted user emotion recognition result, performing weighted summation to obtain the emotion regulation value, setting the regulation threshold, and if the emotion regulation value is greater than the regulation threshold, recording the user emotion regulation strategy and implementing it; if the emotion regulation value is less than or equal to the regulation threshold, re-optimizing the generator to select the emotion regulation strategy until the emotion regulation value is greater than the regulation threshold.

7. The emotion recognition method based on artificial intelligence as claimed in claim 6, characterized in that: Storing the emotion prediction results in the database refers to storing the predicted user emotion recognition results and the user emotion recognition results in the database, synchronously acquiring the user data after the user emotion regulation strategy is implemented for storage, and the database regularly uploads the stored data to the cloud for backup.

8. An emotion recognition system based on artificial intelligence, characterized by: Used for performing emotion recognition based on the emotion recognition method based on artificial intelligence according to any one of claims 1 to 7, the emotion recognition system based on artificial intelligence comprises at least: a data acquisition end, a data calculation and storage end, and a data storage end, wherein the data calculation and storage end is configured with an edge computing box, and the edge computing box is configured with an AI algorithm for calculating emotion recognition; wherein: The data collection end is used to obtain user data, preprocess and extract features of the user data, and build an emotion recognition model to perform user emotion recognition; The data calculation and storage end is used to match the personalized emotion regulation strategy based on the user emotion recognition result, and perform user emotion prediction according to the emotion regulation strategy; The data storage end is used to optimize the user's personalized emotion regulation strategy according to the user emotion prediction result, and store the emotion prediction result in a preset database.

9. A computer device comprising: Memory and processor; The memory stores a computer program, characterized in that when the processor executes the computer program, the steps of the emotion recognition method based on artificial intelligence described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the emotion recognition method based on artificial intelligence described in any one of claims 1 to 7 are implemented.

Images