Intelligent Doll Interaction System and Method Based on Context Awareness and AI Emotion Computing

Abstract

This invention belongs to the field of intelligent sensing and control dolls, and discloses an intelligent doll interaction system and method based on context awareness and AI emotion computing. By retrieving facial image data of different individuals, corresponding voice and touch command features generated during interaction, and corresponding historical EEG physiological features, the facial image data is processed and fused to obtain intelligent doll interaction features. Using these features and a pre-defined doll scene interaction method, a doll context-aware interaction model is constructed. This model is then used to process individual interaction needs, resulting in a real-time doll scene interaction method that controls the doll to perform different movements. This invention achieves emotional interaction with specific groups by processing facial images and superimposing physiological signals for these groups, thus enhancing user experience and stimulating interaction. It upgrades doll interaction from traditional user command information recognition to personalized fusion processing of emotionally relevant indicators.

Description

Technical Field

[0001] This invention belongs to the field of intelligent sensing and control doll technology, and particularly relates to an intelligent doll interaction system and method based on context awareness and AI emotion computing. Background Technology

[0002] In recent years, the cross-integration of multimodal perception, affective computing, and personalized human-computer interaction has gradually developed. Traditional smart toys lack a deep understanding of users' emotional states and interactive contexts, making it difficult to achieve natural and adaptive emotional companionship. With the development of flexible electronics and micro-sensing technology, high-precision tactile, motion, and physiological signal acquisition has become possible; at the same time, lightweight edge AI models and multimodal emotion recognition algorithms provide a technological foundation for real-time understanding of user emotions. By integrating computer vision, speech emotion analysis, and behavioral pattern recognition, the system can dynamically construct user contextual profiles and generate interactive responses with emotional resonance using reinforcement learning strategies. This technological approach not only enhances the naturalness and immersion of interaction but also provides innovative tools for social and emotional intervention for special groups, such as children with autism.

[0003] Therefore, existing technologies face challenges in understanding how to leverage image data from different individuals, the corresponding command features generated during interaction, and the unique physiological characteristics of these individuals. Furthermore, they need to preprocess facial image data using existing traditional AI models to fuse these features and construct a context-aware interactive model for dolls. This model would then be more generalized and robust in handling individual interaction needs, enabling real-time doll scene interaction methods to control the dolls' different movements. Finally, there are questions regarding how to process facial images to perform physiological signal superposition processing tailored to specific groups, thereby achieving emotional interaction with these groups, enhancing user experience and stimulating interaction, and upgrading doll interaction from traditional user command information recognition to personalized fusion processing of emotionally relevant indicators. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes an intelligent doll interaction system and method based on context awareness and AI emotion computing.

[0005] In a first aspect of the invention, a method for interacting with an intelligent doll based on context awareness and AI emotion computing is provided, the method comprising:

[0006] S1. Retrieve facial image data of different special individuals, corresponding voice touch command features generated during communication and interaction, and EEG physiological features of historical records to form a dataset for training the doll situational awareness interaction model.

[0007] S2. After performing feature processing on the facial image data, the voice touch command features and the electroencephalogram (EEG) features are fused to obtain intelligent doll interaction features, and the corresponding set doll scene interaction method is obtained.

[0008] S3. Construct the doll context-aware interaction model using the intelligent doll interaction features and the doll scene interaction method. Use the doll context-aware interaction model to process individual interaction needs to obtain real-time intelligent doll interaction features and obtain real-time doll scene interaction method control. Control the doll to perform different dances and voice outputs according to the real-time doll scene interaction method.

[0009] Furthermore, the feature processing of the facial image data is obtained by processing the left and right halves of the face separately from two different facial images of the same specific individual.

[0010] Furthermore, the process of processing the left and right halves of two different facial images of the same individual to obtain a facial image change difference coefficient is obtained by comprehensively processing the facial image change difference coefficients of facial images of the same individual interacting with the doll at different times.

[0011] Furthermore, the voice touch command features are acquired based on the presence or absence of voice control commands and the pressure sensors on the doll.

[0012] Furthermore, the electroencephalographic features are acquired using a wireless electroencephalogram (EEG) acquisition device called NeuroScan. This EEG acquisition device can acquire EEG signals from 256 brain-computer interfaces. The specific brain acquisition areas for the EEG features are C3, C4, CZ, Fz, F3, and F4.

[0013] Furthermore, after feature processing of the facial image data, the voice touch command features and the electroencephalogram (EEG) features are fused to obtain the intelligent doll interaction features. Specifically, the facial image data, the voice touch command features, and the EEG features are spliced ​​together using spatial feature vectors to obtain the intelligent doll interaction features.

[0014] Furthermore, the doll context-aware interaction model adopts a neural network model based on the improved face image change difference coefficient.

[0015] It also provides an intelligent doll interaction system based on context awareness and AI emotion computing. The system includes a facial image retrieval and processing module, a voice touch command feature storage module, an EEG signal retrieval and processing module, an intelligent doll interaction feature generation module, a doll context awareness interaction model construction module, and a doll context awareness interaction module. Its features are:

[0016] The face image retrieval and processing module is used to retrieve face image data of different special individuals;

[0017] The voice touch command feature storage module is used to acquire the corresponding voice touch command features generated during communication and interaction.

[0018] The EEG signal retrieval and processing module is used to acquire historical EEG physiological characteristics.

[0019] The intelligent doll interaction feature generation module: after performing feature processing on the face image data, it fuses the voice touch command features and the electroencephalogram (EEG) physiological features to obtain intelligent doll interaction features, and obtains the corresponding set doll scene interaction methods to form a dataset for training the doll context perception interaction model.

[0020] The doll context-aware interaction model construction module: constructs the doll context-aware interaction model using the intelligent doll interaction features and the doll scene interaction method;

[0021] The doll context-aware interaction module: uses the doll context-aware interaction model to process individual interaction needs to obtain real-time intelligent doll interaction features, and obtains real-time doll scene interaction method control, and controls the doll to perform different dances and voice outputs according to the real-time doll scene interaction method control.

[0022] Furthermore, the feature processing of the facial image data is obtained by processing the left and right halves of two different facial images of the same individual separately. The facial image change difference coefficient is obtained by processing the left and right halves of two different facial images of the same individual separately. The facial image change difference coefficient is obtained by comprehensively processing the facial image change difference coefficients of facial images of the same individual interacting with the doll at different times.

[0023] The feature processing of facial image data of special individuals in this invention is based on the following objective facts and the needs of model construction. Generally speaking, special individuals, especially children with autism, have a lower frequency and intensity of spontaneous facial expressions, especially social smiles, and a smaller range of expression changes. In terms of facial coordination, the synchronicity and coordination of various facial areas, such as eyebrows, eyes, and lips, are weak when expressing complex emotions, appearing unnatural or fragmented.

[0024] This invention calculates the difference coefficient of facial image changes by dividing the face into left and right sides for comprehensive calculation to obtain the difference coefficient of facial image changes of the same individual at different interaction times. This enables accurate feature representation of facial expression changes of specific individuals and eliminates the need to calculate the difference of each pixel, greatly reducing the amount of computational data and improving the accuracy of the doll interaction method output for subsequent model construction.

[0025] This invention targets specific populations. It retrieves facial image data from different individuals, analyzes corresponding voice and touch command features generated during interaction, and incorporates historical EEG physiological features. After feature processing of the facial image data, the voice and touch command features and EEG physiological features are fused to obtain intelligent doll interaction features. Using these intelligent doll interaction features and doll scene interaction methods, a group big data model of a doll context-aware interaction model is constructed. The doll context-aware interaction model processes the interaction needs of the same individual at different times, generating real-time intelligent doll interaction features to control the doll to perform different forms of dance. This invention achieves emotional interaction for specific populations by processing facial images and superimposing physiological signals, thus enhancing user experience and stimulating interaction. It upgrades doll interaction from traditional user command information recognition to the fusion processing of emotionally related indicators. This not only integrates emotion and commands to achieve personalized dance interaction but also particularly helps improve the emotional experience and sense of companionship for specific populations. Attached Figure Description

[0026] Figure 1 This is a flowchart of the intelligent doll interaction method based on context awareness and AI emotion computing of the present invention;

[0027] Figure 2 This is a schematic diagram of the intelligent doll interaction system based on context awareness and AI emotion computing of the present invention;

[0028] Figure 3 This is a schematic diagram of the left and right half-face processing in this invention;

[0029] Figure 4 This is a schematic diagram of the application of the intelligent doll interaction method based on context awareness and AI emotion computing in an embodiment of the present invention.

[0030] Figure 5 This is a schematic diagram of the electronic device structure for implementing the method of the present invention in an embodiment of the present invention. Detailed Implementation

[0031] The invention will now be further described in conjunction with the accompanying drawings and specific embodiments. The improved neural network model used in this invention is different from the traditional neural network model. It is an improved neural network model that is adapted to the training set data. This model is a specific modification of the model in the context-aware and AI emotion computing-based intelligent doll interaction scenario. Specifically, it improves the activation function of the neural network model by using the face image change coefficients that have been innovatively preprocessed in the training set, so as to obtain a neural network model with more accurate classification probability values.

[0032] The intelligent doll interaction system and method based on context awareness and AI emotion computing of the present invention are used in the manufacturing of consumer devices for intelligent sensing and control of dolls. It belongs to the manufacturing of other intelligent consumer devices such as intelligent sensing and control devices, and therefore belongs to the field of artificial intelligence industry.

[0033] In a first aspect of the invention, a method for interacting with an intelligent doll based on context awareness and AI emotion computing is provided, the method comprising:

[0034] S1. Retrieve facial image data of different special individuals, corresponding voice touch command features generated during communication and interaction, and EEG physiological features of historical records to form a dataset for training the doll situational awareness interaction model.

[0035] S2. After performing feature processing on the facial image data, the voice touch command features and the electroencephalogram (EEG) features are fused to obtain intelligent doll interaction features, and the corresponding set doll scene interaction method is obtained.

[0036] S3. Construct the doll context-aware interaction model using the intelligent doll interaction features and the doll scene interaction method. Use the doll context-aware interaction model to process individual interaction needs to obtain real-time intelligent doll interaction features and obtain real-time doll scene interaction method control. Control the doll to perform different dances and voice outputs according to the real-time doll scene interaction method.

[0037] Existing electric intelligent dolls are only suitable for children in the general population, and are not specifically designed for children with disabilities. General-purpose dolls pursue a "cool" experience, often featuring sudden flashes of light, high-frequency music, rapid and unpredictable movements, and complex and varied voices. For autistic children who are sensitive to their senses, this is not entertainment, but a sensory bombardment, which can easily trigger anxiety, screaming, or avoidance behaviors. Therefore, there is an urgent need for interactive intelligent dolls with adjustable, gentle, and predictable sensory output based on context awareness and AI emotional computing.

[0038] The specific individuals mentioned in this invention refer in particular to people with autism who crave targeted doll companionship.

[0039] Furthermore, the feature processing of the facial image data involves processing the facial image change difference coefficient of the same individual interacting with the doll at different times. The resulting facial image change difference coefficient, along with the voice touch command features and the electroencephalographic features, yields the intelligent doll interaction features. The formula for calculating the change difference coefficient of one of the first facial images between different facial images of the same individual in the training set is as follows:

[0040]

[0041] In the formula, To train a set of different face images of the same individual, the variation coefficient of one of the first face images is used. The total grayscale value of the left half of the face image data of the first face image is given. The total grayscale value of the left half of the face image data of the second face image. The average grayscale value of the left half of the face in the first face image data and the second face image data is the total grayscale value. This represents the total grayscale value of the right half of the face image in the first face image data. This represents the total grayscale value of the right half of the face image from the second face image data. The average grayscale value of the right half of the face is the sum of the first face image data and the second face image data. In addition, in order to obtain the face image change difference coefficient of the whole face, the face image change difference coefficients calculated for the left and right halves of the face are averaged.

[0042] The feature processing of facial image data of special individuals in this invention is based on the following objective facts and the needs of model construction. Generally speaking, special individuals, especially children with autism, have a lower frequency and intensity of spontaneous facial expressions, especially social smiles, and a smaller range of expression changes. In terms of facial coordination, the synchronicity and coordination of various facial areas, such as eyebrows, eyes, and lips, are weak when expressing complex emotions, appearing unnatural or fragmented.

[0043] This invention, when calculating the facial image variation difference coefficient, divides the face into left and right sides for comprehensive calculation to obtain the facial image variation difference coefficient of the same individual at different interaction times. This allows for accurate feature representation of facial expression changes in specific individuals and eliminates the need to calculate the difference of each individual pixel, significantly reducing the amount of computational data and improving the accuracy of the doll interaction method output for subsequent model construction. The division of the left and right sides of the face is determined by... Figure 3 As shown.

[0044] Furthermore, the facial image variation difference coefficient is obtained by processing the facial image variation difference coefficients of the same specific individual interacting with the doll at different times. The comprehensive calculation formula for the facial image variation difference coefficient is as follows:

[0045]

[0046] In the formula, To train the coefficients of variation of facial images of the same individual in the same dataset. Let n be the number of combinations of two different facial images of the same individual, and n be the number of facial images of the same individual. Let be the difference coefficient of the i-th face image in the i-th combination.

[0047] In this embodiment, after calculating the difference coefficient of the change in one of the left and right half of the face image, the sum of the difference coefficients of the face images during multiple doll interactions can be calculated to more comprehensively acquire big data of face images and represent the emotional changes of doll interactions for special groups. This provides a basis for subsequent optimization of data selection and targeted expression for special groups in building a context-aware and AI emotion computing model.

[0048] Furthermore, the voice touch command features are acquired based on the presence or absence of voice control commands and the pressure sensors on the doll.

[0049] In this embodiment, if a person with autism issues a voice command and touches or presses a specific part of the doll, such as the doll's head or waist, generally speaking, people with autism are less sensitive to pressure than typical children and will apply greater pressure. The feature vector representing the corresponding voice-touch command is then:

[0050]

[0051] In this case, 1 represents that the individual has produced speech, while 0 represents that no speech has been produced. The force of the individual's touch is 8N. The above feature vectors are preprocessed by removing dimensions to facilitate subsequent model calculations.

[0052] Furthermore, the electroencephalographic features are acquired using a wireless electroencephalogram (EEG) acquisition device called NeuroScan. This EEG acquisition device can acquire EEG signals from 256 brain-computer interfaces. The specific brain acquisition areas for the EEG features are C3, C4, CZ, Fz, F3, and F4.

[0053] Furthermore, the formula for calculating the feature vector value of the aforementioned electroencephalographic features is as follows:

[0054]

[0055] In the formula, These are brain electrophysiological characteristics. , where m is the gender difference coefficient for a specific population, and m is the number of EEG sampling sites. This refers to the amplitude of the electroencephalogram (EEG) characteristics at the j-th EEG acquisition site of a specific individual. This represents the average amplitude of the electroencephalographic characteristics of the j-th EEG acquisition site in a typical individual, in units of... .

[0056] In this embodiment, the EEG acquisition sites are the EEG signal amplitude measured over a period of time at six brain-computer interfaces: C3, C4, CZ, Fz, F3, and F4.

[0057] Furthermore, after feature processing of the facial image data, the voice touch command features and the electroencephalogram (EEG) features are fused to obtain the intelligent doll interaction features. Specifically, the facial image data, the voice touch command features, and the EEG features are spliced ​​together using spatial feature vectors to obtain the intelligent doll interaction features.

[0058] Furthermore, the doll context-aware interaction model employs a neural network model improved based on the coefficient of variation of facial images. The activation function calculation formula for the neural network model improved based on the coefficient of variation of facial images is as follows:

[0059]

[0060] In the formula, The activation function value is the value of the neural network model improved based on the difference coefficient of facial image variations. To obtain the minimum face image variation difference coefficient in the dataset used to train the doll context-aware interaction model, The dataset used to train the doll context-aware interaction model contains the maximum face image variation coefficient. The feature values ​​of the intelligent doll interaction features are input into the activation function and linearly transformed by weights and biases. Due to the non-interpretability of neural networks, activation functions can solve the defects of nonlinear expression in the model and are used to solve linearly inseparable problems. However, when it comes to the output of doll scene interaction methods, general activation functions have certain limitations in handling nonlinear changes and cannot be well applied to the output of doll scene interaction methods. In particular, the facial expression changes of different individuals during doll interaction can largely determine the output of the doll scene interaction method. By taking the minimum and maximum face image change difference coefficients in the training set data to adjust the activation function, the model's generalization performance and robustness in nonlinear problems can be improved in the subsequent acquisition of doll scene interaction methods.

[0061] In this embodiment, the corresponding doll scene interaction method is determined based on the final output value of the neural network. In one embodiment of this application, if the final output value of the neural network is 0.125, the doll scene interaction method is determined to be the doll performing a first type of dance and outputting a first type of voice. If the final output value of the neural network is 0.764, the doll scene interaction method is determined to be the doll performing a second type of dance and outputting a second type of voice. The corresponding doll dance and voice output are set by the pediatric experts of this invention. The above are the corresponding results obtained after training the model in this application, and the determination of the doll scene interaction method in this application is not limited to this result. Figure 4 An example diagram of a doll using the method of the present invention.

[0062] It also provides an intelligent doll interaction system based on context awareness and AI emotion computing. The system includes a facial image retrieval and processing module, a voice touch command feature storage module, an EEG signal retrieval and processing module, an intelligent doll interaction feature generation module, a doll context awareness interaction model construction module, and a doll context awareness interaction module.

[0063] The face image retrieval and processing module is used to retrieve face image data of different special individuals;

[0064] The voice touch command feature storage module is used to acquire the corresponding voice touch command features generated during communication and interaction.

[0065] The EEG signal retrieval and processing module is used to acquire historical EEG physiological characteristics.

[0066] The intelligent doll interaction feature generation module: after performing feature processing on the face image data, it fuses the voice touch command features and the electroencephalogram (EEG) physiological features to obtain intelligent doll interaction features, and obtains the corresponding set doll scene interaction methods to form a dataset for training the doll context perception interaction model.

[0067] The doll context-aware interaction model construction module: constructs the doll context-aware interaction model using the intelligent doll interaction features and the doll scene interaction method;

[0068] The doll context-aware interaction module: uses the doll context-aware interaction model to process individual interaction needs to obtain real-time intelligent doll interaction features, and obtains real-time doll scene interaction method control, and controls the doll to perform different dances and voice outputs according to the real-time doll scene interaction method control.

[0069] Furthermore, the feature processing of the facial image data involves processing the facial image change difference coefficient of the same individual interacting with the doll at different times. The resulting facial image change difference coefficient, along with the voice touch command features and the electroencephalographic features, yields the intelligent doll interaction features. The formula for calculating the change difference coefficient of one of the first facial images between different facial images of the same individual in the training set is as follows:

[0070]

[0071] In the formula, To train a set of different face images of the same individual, the variation coefficient of one of the first face images is used. The total grayscale value of the left half of the face image data of the first face image is given. The total grayscale value of the left half of the face image data of the second face image. The average grayscale value of the left half of the face in the first face image data and the second face image data is the total grayscale value. This represents the total grayscale value of the right half of the face image in the first face image data. This represents the total grayscale value of the right half of the face image from the second face image data. The average grayscale value of the right half of the face is the sum of the first face image data and the second face image data. In addition, in order to obtain the face image change difference coefficient of the whole face, the face image change difference coefficients calculated for the left and right halves of the face are averaged.

[0072] The feature processing of facial image data of special individuals in this invention is based on the following objective facts and the needs of model construction. Generally speaking, special individuals, especially children with autism, have a lower frequency and intensity of spontaneous facial expressions, especially social smiles, and a smaller range of expression changes. In terms of facial coordination, the synchronicity and coordination of various facial areas, such as eyebrows, eyes, and lips, are weak when expressing complex emotions, appearing unnatural or fragmented.

[0073] This invention calculates the difference coefficient of facial image changes by dividing the face into left and right sides for comprehensive calculation to obtain the difference coefficient of facial image changes of the same individual at different interaction times. This enables accurate feature representation of facial expression changes of specific individuals and eliminates the need to calculate the difference of each pixel, greatly reducing the amount of computational data and improving the accuracy of the doll interaction method output for subsequent model construction.

[0074] This invention targets specific populations. It retrieves facial image data from different individuals, analyzes corresponding voice and touch command features generated during interaction, and incorporates historical EEG physiological features. After feature processing of the facial image data, the voice and touch command features and EEG physiological features are fused to obtain intelligent doll interaction features. Using these intelligent doll interaction features and doll scene interaction methods, a group big data model of a doll context-aware interaction model is constructed. The doll context-aware interaction model processes the interaction needs of the same individual at different times, generating real-time intelligent doll interaction features to control the doll to perform different forms of dance. This invention achieves emotional interaction for specific populations by processing facial images and superimposing physiological signals, thus enhancing user experience and stimulating interaction. It upgrades doll interaction from traditional user command information recognition to the fusion processing of emotionally related indicators. This not only integrates emotion and commands to achieve personalized dance interaction but also particularly helps improve the emotional experience and sense of companionship for specific populations.

[0075] The combination of multiple embodiments of the present invention can achieve all the above effects, but it is not required that each embodiment of the present invention achieve all the above advantages and effects, because each embodiment of the present invention can constitute a separate technical solution and make one or more contributions to the prior art.

[0076] For any module structures not specifically defined in this invention, the existing technical specifications shall prevail. The existing technical specifications mentioned in the foregoing background and specific embodiments sections are considered part of this invention and are used to understand the meaning of certain technical features or parameters. The scope of protection of this invention is determined by the actual contents of the claims.

Claims

1. A method for smart doll interaction based on context awareness and AI emotional computing, characterized in that, The method includes: S1. Retrieve facial image data of different special individuals, corresponding voice touch command features generated during communication and interaction, and EEG physiological features of historical records to form a dataset for training the doll situational awareness interaction model. S2. After performing feature processing on the facial image data, the voice touch command features and the electroencephalogram (EEG) features are fused to obtain intelligent doll interaction features, and the corresponding set doll scene interaction method is obtained. S3. Construct the doll context-aware interaction model using the intelligent doll interaction features and the doll scene interaction method. Use the doll context-aware interaction model to process individual interaction needs to obtain real-time intelligent doll interaction features and obtain real-time doll scene interaction method control. Control the doll to perform different dances and voice outputs according to the real-time doll scene interaction method.

2. The intelligent doll interaction method based on context awareness and AI emotion computing as described in claim 1, characterized in that: The feature processing of the facial image data is obtained by processing the left and right halves of the face separately from two different facial images of the same specific individual.

3. The intelligent doll interaction method based on context awareness and AI emotion computing as described in claim 2, characterized in that: The method involves processing the left and right halves of two different facial images of the same individual separately to obtain the facial image change difference coefficient. The facial image change difference coefficient is obtained by comprehensively processing the facial image change difference coefficients of the same individual interacting with the doll at different times.

4. The intelligent doll interaction method based on context awareness and AI emotion computing as described in claim 1, characterized in that: The voice touch command features are obtained based on the presence or absence of voice control commands and the pressure sensors on the doll.

5. The intelligent doll interaction method based on context awareness and AI emotion computing as described in claim 1, characterized in that: The electroencephalogram (EEG) characteristics were acquired using a wireless EEG acquisition device called NeuroScan. The EEG acquisition device can acquire EEG signals from 256 brain-computer interfaces. The specific brain acquisition areas for the EEG characteristics are C3, C4, CZ, Fz, F3, and F4.

6. The intelligent doll interaction method based on context awareness and AI emotion computing as described in claim 5, characterized in that: The calculation of the feature vector value of the electroencephalogram (EEG) characteristics is related to the gender difference coefficient of the special population, the number of EEG acquisition sites, and the amplitude of the EEG characteristics at the j-th EEG acquisition site of the special individual.

7. The intelligent doll interaction method based on context awareness and AI emotion computing as described in claim 3, 4, or 6, characterized in that: After performing feature processing on the facial image data, the voice touch command features and the electroencephalogram (EEG) features are fused to obtain the intelligent doll interaction features. Specifically, the facial image data, the voice touch command features, and the EEG features are spliced ​​together as spatial feature vectors to obtain the intelligent doll interaction features.

8. The intelligent doll interaction method based on context awareness and AI emotion computing as described in claim 7, characterized in that: The doll context-aware interaction model adopts a neural network model based on the improved face image change difference coefficient.

9. An intelligent doll interaction system based on context awareness and AI emotion computing, the system comprising a face image retrieval and processing module, a voice touch command feature storage module, an EEG signal retrieval and processing module, an intelligent doll interaction feature generation module, a doll context awareness interaction model construction module, and a doll context awareness interaction module, characterized in that: The face image retrieval and processing module is used to retrieve face image data of different special individuals; The voice touch command feature storage module is used to acquire the corresponding voice touch command features generated during communication and interaction. The EEG signal retrieval and processing module is used to acquire historical EEG physiological characteristics. The intelligent doll interaction feature generation module: after performing feature processing on the face image data, it fuses the voice touch command features and the electroencephalogram (EEG) physiological features to obtain intelligent doll interaction features, and obtains the corresponding set doll scene interaction methods to form a dataset for training the doll context perception interaction model. The doll context-aware interaction model construction module: constructs the doll context-aware interaction model using the intelligent doll interaction features and the doll scene interaction method; The doll context-aware interaction module: uses the doll context-aware interaction model to process individual interaction needs to obtain real-time intelligent doll interaction features, and obtains real-time doll scene interaction method control, and controls the doll to perform different dances and voice outputs according to the real-time doll scene interaction method control.

10. The intelligent doll interaction system based on context awareness and AI emotion computing as described in claim 9, characterized in that: The feature processing of the facial image data is obtained by processing the left and right halves of two different facial images of the same individual separately. The facial image change difference coefficient is obtained by processing the left and right halves of two different facial images of the same individual separately. The facial image change difference coefficient is obtained by comprehensively processing the facial image change difference coefficients of facial images of the same individual interacting with the doll at different times.

Images