A personalized learning path recommendation system for English majors

Abstract

The present invention discloses a personalized learning path recommendation system for English majors, and relates to the technical field of personalized learning path recommendation for English majors. The system comprises an English personalized learning acquisition module, an English personalized learning processing module, an English personalized learning analysis module, and an English personalized learning recommendation module. The English personalized learning acquisition module is used to obtain users' English major learning data, wherein the English major learning data is divided into behavioral feature data and image data; the English personalized learning processing module is divided into an elderly learning path recommendation unit and an image emotion recognition unit, which respectively use algorithms to construct a personalized learning path recommendation model for the elderly and an image emotion recognition model; the English personalized learning analysis module analyzes the user's age data and real-time dynamic facial images, generates analysis results, and transmits them to the recommendation module; and the English personalized learning path recommendation module receives the analysis results and makes corresponding recommendations.

Description

Technical Field

[0001] The present invention relates to the technical field of personalized English learning path recommendation, and in particular to a personalized English learning path recommendation system. Background Art

[0002] The personalized learning path recommendation system for English majors integrates advanced technologies from multiple fields. Its core lies in personalized recommendation algorithms. These algorithms can deeply analyze users' learning behaviors, interests, preferences, and learning outcomes, thereby accurately pushing English learning resources that meet users' needs. At the same time, intelligent language processing technology also plays a key role. It can deeply analyze text content and extract key information, providing strong support for the recommendation system. In addition, big data analysis technology is also widely used in the system. By mining and analyzing massive learning data, the system can continuously optimize recommendation strategies and improve the accuracy and personalization of recommendations. The combined application of these technologies enables the personalized learning path recommendation system for English majors to provide users with more efficient and accurate learning path recommendation services.

[0003] Although the existing personalized learning path recommendation system for English majors has made great progress, there are still some problems that need to be optimized. Most elderly people have poorer English learning ability than young people. Using the same recommendation system will make the elderly uncomfortable in the English learning process. In addition, the elderly have limited acceptance and operation capabilities of modern technological products. The complex system interface may hinder the learning process of the elderly group. Secondly, when users learn with negative emotions, the user's psychological pressure increases, resulting in a decrease in learning motivation, resulting in the resources originally recommended by the system not being fully utilized. Summary of the Invention

[0004] In order to solve the above problems, the present invention provides an English major personalized learning path recommendation system, which includes an English personalized learning acquisition module, an English personalized learning processing module, an English personalized learning analysis module and an English personalized learning recommendation module;

[0005] The English personalized learning acquisition module is used to obtain the user's English major learning data, wherein the English major learning data is divided into behavioral feature data and image data;

[0006] The English personalized learning processing module is divided into an elderly learning path recommendation unit and an image emotion recognition unit. The elderly learning path recommendation unit is used to extract elderly learning preference data, generate elderly electronic device proficiency, and build an elderly personalized learning path recommendation model; the image emotion recognition unit is used to extract face position, user facial deep features, and build an image emotion recognition model.

[0007] The English personalized learning analysis module analyzes the user's age data and real-time dynamic facial images, and transmits the analysis results to the English personalized learning path recommendation module;

[0008] The English personalized learning path recommendation module makes corresponding recommendations based on the received analysis results

[0009] Furthermore, the behavioral characteristic data includes the user's age, learning period, number of views of English articles and videos, and the results of a questionnaire on proficiency in electronic device use. The acquisition process includes:

[0010] Utilize the user's English learning device to collect the user's age data, and obtain the user's learning time period data, and the number of times the English article and English video have been viewed;

[0011] A related questionnaire on the operation of English learning equipment is preset, wherein the questionnaire includes multiple-choice questions, true-or-false questions and actual operation proficiency scores on the use of English learning equipment. Based on the questionnaire survey results, the user's proficiency scores for English learning equipment are extracted.

[0012] Furthermore, the image data is a real-time dynamic image of the user's face collected by a high frame rate wide dynamic camera, and the acquisition process includes:

[0013] After the user clicks "Start Learning", the high-frame-rate dynamic camera equipped on the user's English learning device is automatically started, and the camera's frame rate and resolution are initialized;

[0014] The high frame rate dynamic camera after initialization is used to collect user facial expression image data at a high frequency to obtain real-time dynamic images of the user's face.

[0015] Furthermore, in the elderly learning path recommendation unit, the process of extracting elderly learning preference data and generating elderly electronic device proficiency includes:

[0016] The user's age, learning time period, and number of views of English articles and videos are encoded and converted into a form that can be recognized by the CNN model. The CNN model is constructed, which consists of convolutional layers, pooling layers, and fully connected layers. The encoded data is input into the CNN model. Through the alternating action of convolutional layers and pooling layers, the daily English learning time period and English learning method of the elderly are extracted from the user's age, learning time period, and number of views of English articles and videos. The extracted features are then integrated using a fully connected layer to output the elderly's learning preference data.

[0017] The results of the electronic device use proficiency questionnaire were pre-processed and three levels of electronic device use proficiency were set: the accuracy of the primary questionnaire was 30%, the accuracy of the intermediate questionnaire was 60%, and the accuracy of the advanced questionnaire was 90%. Users aged 60 and above were also classified as elderly.

[0018] The questionnaire accuracy and user age are selected as features to construct a decision tree. The results of the electronic device proficiency questionnaire are used as a data set, and the data set is divided into a training set and a test set. The training set is used to build a decision tree, and the test set is used to verify the decision tree performance. The questionnaire results are then input into the trained decision tree model, traversing from the root node until a leaf node is reached. The proficiency of electronic devices for the elderly is output according to the category corresponding to the leaf node.

[0019] Furthermore, in the elderly learning path recommendation unit, the process of constructing a personalized learning path recommendation model for the elderly includes:

[0020] Construct an LSTM model and define the model's input layer, hidden layer, and output layer. The elderly's learning preferences include their daily English learning time and English learning methods. The elderly's learning preference data and electronic device proficiency are organized into a serialized form and input into the LSTM model for model training. The constructed LSTM model is used to identify the elderly's learning preference data and electronic device proficiency. Instructions are issued based on actual conditions to control the personalized learning path recommendation system for English majors to recommend content related to the elderly's learning preferences, adjust the display mode of the interface, and provide operation guides. The LSTM model adjusts model parameters through the optimizer and evaluates the optimization model performance. After the model optimization is completed, a personalized learning path recommendation model for the elderly is generated. The LSTM model collects feedback data from the elderly after receiving the recommendation results in real time, and adjusts parameters and updates the optimization model in real time based on the feedback data.

[0021] Furthermore, in the image emotion recognition unit, the process of extracting the face position and deep features of the user's face includes:

[0022] Perform preprocessing operations such as grayscale conversion, denoising, and image enhancement on the real-time dynamic image of the user's face. In the preprocessed image, based on Haar-like features and cascade classifiers, slide rectangular windows of different sizes and positions to generate Haar features. The difference between the sum of pixels in different rectangular windows is calculated to obtain the feature value reflecting the grayscale change of the area. The weak classifier parameters are adjusted, and the cascade classifier is optimized. The image to be tested is converted into a grayscale image, and the optimized cascade classifier is used to detect and identify the location of the face area.

[0023] Preprocess the real-time dynamic image of the user's face, build a CNN model, and input the preprocessed image into the CNN model. The convolution layer extracts local facial features, including eyes, eyebrows, lips, facial muscles, and facial contours, and generates a feature map. The pooling layer downsamples the feature map to reduce the size of the feature map while retaining important features. The fully connected layer integrates the important features output by the pooling layer and outputs deep facial features of the user, including changes in facial eyes, eyebrows, lips, facial muscles, and facial contours.

[0024] Furthermore, in the image emotion recognition unit, the process of constructing the image emotion recognition model includes:

[0025] A1. Setting negative emotion labels, wherein the negative emotion labels are composed of anxiety, depression, anger, frustration, and low self-esteem;

[0026] A2. Using a random forest algorithm and setting random forest parameters, the facial region locations and deep facial features extracted from the real-time dynamic image of the user's face are associated with corresponding emotion labels. Anxiety corresponds to furrowed brows, wandering eyes, tightly closed lips, and tense facial muscles. Depression corresponds to hollow eyes, drooping mouth corners, a dull facial expression, and sagging facial contours. Anger corresponds to sunken brows, wide eyes, tightly closed and protruding lips, and tense facial muscles. Frustration corresponds to dull eyes, drooping mouth corners, and a tired and weak face. Inferiority corresponds to wandering eyes, tightly closed lips, and a tense face.

[0027] A3. Use the facial region locations and deep facial features extracted from the real-time dynamic images of the user's face and the corresponding emotion labels to form a dataset. The dataset is divided into a training set, a validation set, and a test set. Multiple subsets are extracted from the training set, and a decision tree is trained for each subset. During the generation process of each decision tree, some features are randomly selected for node splitting to construct a decision tree. The above process is repeated until the predetermined number of decision trees is reached.

[0028] A4. Use the validation set to evaluate the trained random forest model. Adjust the model fit by changing the parameters. Then input the test set into the adjusted random forest model to evaluate the model performance and obtain the image emotion recognition model.

[0029] Furthermore, the process of analyzing the user's age data and real-time dynamic facial image in the English personalized learning analysis module includes:

[0030] S1. Analyze the user's age data. If the user is younger than 60 years old, the user is determined to be a non-elderly person. If the user is older than 60 years old, the user is determined to be an elderly person.

[0031] S2, using the image emotion recognition model, identifies the emotions in the user's facial dynamic image in real time and outputs the corresponding emotion label;

[0032] S3. Wirelessly transmit the above analysis results to the English personalized learning path recommendation module.

[0033] Furthermore, the process of the English personalized learning path recommendation module making corresponding recommendations based on the received user age analysis results includes:

[0034] Receive the analysis results of the user's age. When a signal is received that the user is an elderly person, switch the personalized learning path recommendation system for English majors to the personalized learning path recommendation mode for the elderly, adjust the interface display font to a large size, simplify the learning interface, and provide corresponding operation guides and tutorials. Through the personalized learning path recommendation model for the elderly in the above mode, adjust in real time to the English learning path that suits the learning preferences of the elderly.

[0035] Furthermore, the process of the English personalized learning path recommendation module making corresponding recommendations based on the analysis results of the received user emotion tags includes:

[0036] Based on the user's emotional labels, for anxiety, we recommend English movies and songs and adjust the user's learning goals; for depression, we recommend positive English articles; for anger, we recommend meditation audio and soothing English movies; for frustration, we recommend interesting learning content and share other users' learning experiences; for inferiority complex, we recommend basic grammar learning and vocabulary memorization suitable for the user's level.

[0037] The beneficial effects of the present invention are as follows: compared with the traditional personalized learning path recommendation system for English majors, the model construction technology in the system of the present invention is closely integrated with modern information technology. The system adopts a combination of hardware integration and software integration, integrates data entry technology and high frame rate wide dynamic camera technology, accurately collects the user's age, learning time period, number of views of English articles and English videos, electronic equipment proficiency questionnaire results and real-time dynamic image of the face, extracts the features of the user's age, learning time period, number of views of English articles and English videos through convolutional neural network CNN to generate elderly learning preference data, uses the electronic equipment proficiency questionnaire results to generate elderly electronic equipment proficiency through decision tree algorithm, and then uses LSTM algorithm to construct a personalized learning path recommendation model for the elderly; extracts face position and deep features of the user's face through Haar features and convolutional neural network CNN, and then uses random sampling to extract the features of the user's age, learning time period, number of views of English articles and English videos, and generates elderly learning preference data; uses random sampling to extract the features of the user's age, learning time period, number of views of English articles and English videos, and generates elderly electronic equipment proficiency through decision tree algorithm, and then uses LSTM algorithm to construct a personalized learning path recommendation model for the elderly; extracts face position and deep features of the user's face through Haar features and convolutional neural network CNN, and then uses random sampling to extract the features of the user's age, learning time period, number of views of English articles and English videos, and generates elderly learning preference data Machine forest algorithm, builds an image emotion recognition model, and uses real-time user age data and real-time dynamic facial images, combined with the constructed model, to achieve real-time monitoring of user age and emotional changes. The above technology solves the problem that most elderly people's ability to learn English is worse than that of young people, and using the same recommendation system will make the elderly uncomfortable in the English learning process. In addition, the elderly have limited acceptance and operation capabilities of modern scientific and technological products, and complex system interfaces may hinder the learning process of the elderly group; secondly, when users learn with negative emotions, the user's psychological pressure increases, resulting in a decrease in learning motivation, resulting in the system's originally recommended resources not being fully utilized, ensuring that the method in the present invention can refine the dynamic monitoring standards of the personalized learning path recommendation system for English majors within a more precise range, so that the monitored data becomes a more accurate indicator under the same conditions. The research and development and application of this method significantly enhance the level of intelligence in the working process of the personalized learning path recommendation system for English majors. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A block diagram of a personalized learning path recommendation system for English majors. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] like Figure 1As shown, the present invention provides a technical solution: a personalized learning path recommendation system for English majors, including an English personalized learning acquisition module, an English personalized learning processing module, an English personalized learning analysis module and an English personalized learning recommendation module;

[0041] The English personalized learning acquisition module is used to obtain the user's English major learning data, where the English major learning data is divided into behavioral feature data and image data;

[0042] The English personalized learning processing module is divided into a learning path recommendation unit for the elderly and an image emotion recognition unit. The learning path recommendation unit is used to extract elderly learning preference data, generate elderly electronic device proficiency levels, and build a personalized learning path recommendation model for the elderly; the image emotion recognition unit is used to extract face location, deep facial features of users, and build an image emotion recognition model.

[0043] The English personalized learning analysis module analyzes the user's age data and real-time dynamic facial images, and transmits the analysis results to the English personalized learning path recommendation module;

[0044] English personalized learning path recommendation module, making corresponding recommendations based on the received analysis results

[0045] Behavioral characteristic data includes the user's age, learning period, number of views of English articles and videos, and the results of a questionnaire on electronic device proficiency. The acquisition process includes:

[0046] Utilize the user's English learning device to collect the user's age data, and obtain the user's learning time period data, and the number of times the English article and English video have been viewed;

[0047] A related questionnaire on the operation of English learning equipment is preset, wherein the questionnaire includes multiple-choice questions, true-or-false questions and actual operation proficiency scores on the use of English learning equipment. Based on the questionnaire survey results, the user's proficiency scores for English learning equipment are extracted.

[0048] The image data is a real-time dynamic image of the user's face captured by a high-frame-rate wide dynamic camera. The acquisition process includes:

[0049] After the user clicks "Start Learning", the high-frame-rate dynamic camera equipped on the user's English learning device is automatically started, and the camera's frame rate and resolution are initialized;

[0050] The high frame rate dynamic camera after initialization is used to collect user facial expression image data at a high frequency to obtain real-time dynamic images of the user's face.

[0051] In the elderly learning path recommendation unit, the process of extracting elderly learning preference data and generating elderly electronic device proficiency levels includes:

[0052] The user's age, learning time period, and number of views of English articles and videos are encoded and converted into a form that can be recognized by the CNN model. The CNN model is constructed, which consists of convolutional layers, pooling layers, and fully connected layers. The encoded data is input into the CNN model. Through the alternating action of convolutional layers and pooling layers, the daily English learning time period and English learning method of the elderly are extracted from the user's age, learning time period, and number of views of English articles and videos. The extracted features are then integrated using a fully connected layer to output the elderly's learning preference data.

[0053] The results of the electronic device use proficiency questionnaire were pre-processed and three levels of electronic device use proficiency were set: the accuracy of the primary questionnaire was 30%, the accuracy of the intermediate questionnaire was 60%, and the accuracy of the advanced questionnaire was 90%. Users aged 60 and above were also classified as elderly.

[0054] The questionnaire accuracy and user age are selected as features to construct a decision tree. The results of the electronic device proficiency questionnaire are used as a data set, and the data set is divided into a training set and a test set. The training set is used to build a decision tree, and the test set is used to verify the decision tree performance. The questionnaire results are then input into the trained decision tree model, traversing from the root node until a leaf node is reached. The proficiency of electronic devices for the elderly is output according to the category corresponding to the leaf node.

[0055] In the elderly learning path recommendation unit, the process of building a personalized learning path recommendation model for the elderly includes:

[0056] Construct an LSTM model and define the model's input layer, hidden layer, and output layer. The elderly's learning preferences include their daily English learning time and English learning methods. The elderly's learning preference data and electronic device proficiency are organized into a serialized form and input into the LSTM model for model training. The constructed LSTM model is used to identify the elderly's learning preference data and electronic device proficiency. Instructions are issued based on actual conditions to control the personalized learning path recommendation system for English majors to recommend content related to the elderly's learning preferences, adjust the display mode of the interface, and provide operation guides. The LSTM model adjusts model parameters through the optimizer and evaluates the optimization model performance. After the model optimization is completed, a personalized learning path recommendation model for the elderly is generated. The LSTM model collects feedback data from the elderly after receiving the recommendation results in real time, and adjusts parameters and updates the optimization model in real time based on the feedback data.

[0057] In the image emotion recognition unit, the process of extracting face location and deep facial features of the user includes:

[0058] Perform preprocessing operations such as grayscale conversion, denoising, and image enhancement on the real-time dynamic image of the user's face. In the preprocessed image, based on Haar-like features and cascade classifiers, slide rectangular windows of different sizes and positions to generate Haar features. The difference between the sum of pixels in different rectangular windows is calculated to obtain the feature value reflecting the grayscale change of the area. The weak classifier parameters are adjusted, and the cascade classifier is optimized. The image to be tested is converted into a grayscale image, and the optimized cascade classifier is used to detect and identify the location of the face area.

[0059] Preprocess the real-time dynamic image of the user's face, build a CNN model, and input the preprocessed image into the CNN model. The convolution layer extracts local facial features, including eyes, eyebrows, lips, facial muscles, and facial contours, and generates a feature map. The pooling layer downsamples the feature map to reduce the size of the feature map while retaining important features. The fully connected layer integrates the important features output by the pooling layer and outputs deep facial features of the user, including changes in facial eyes, eyebrows, lips, facial muscles, and facial contours.

[0060] In the image emotion recognition unit, the process of building an image emotion recognition model includes:

[0061] A1. Setting negative emotion labels, wherein the negative emotion labels are composed of anxiety, depression, anger, frustration, and low self-esteem;

[0062] A2. Using a random forest algorithm and setting random forest parameters, the facial region locations and deep facial features extracted from the real-time dynamic image of the user's face are associated with corresponding emotion labels. Anxiety corresponds to furrowed brows, wandering eyes, tightly closed lips, and tense facial muscles. Depression corresponds to hollow eyes, drooping mouth corners, a dull facial expression, and sagging facial contours. Anger corresponds to sunken brows, wide eyes, tightly closed and protruding lips, and tense facial muscles. Frustration corresponds to dull eyes, drooping mouth corners, and a tired and weak face. Inferiority corresponds to wandering eyes, tightly closed lips, and a tense face.

[0063] A3. Use the facial region locations and deep facial features extracted from the real-time dynamic images of the user's face and the corresponding emotion labels to form a dataset. The dataset is divided into a training set, a validation set, and a test set. Multiple subsets are extracted from the training set, and a decision tree is trained for each subset. During the generation process of each decision tree, some features are randomly selected for node splitting to construct a decision tree. The above process is repeated until the predetermined number of decision trees is reached.

[0064] A4. Use the validation set to evaluate the trained random forest model. Adjust the model fit by changing the parameters. Then input the test set into the adjusted random forest model to evaluate the model performance and obtain the image emotion recognition model.

[0065] The English personalized learning analysis module analyzes user age data and real-time dynamic facial images. The process includes:

[0066] S1. Analyze the user's age data. If the user is younger than 60 years old, the user is determined to be a non-elderly person. If the user is older than 60 years old, the user is determined to be an elderly person.

[0067] S2, using the image emotion recognition model, identifies the emotions in the user's facial dynamic image in real time and outputs the corresponding emotion label;

[0068] S3. Wirelessly transmit the above analysis results to the English personalized learning path recommendation module.

[0069] The English personalized learning path recommendation module makes recommendations based on the user's age analysis results. The process includes:

[0070] Receive the analysis results of the user's age. When a signal is received that the user is an elderly person, switch the personalized learning path recommendation system for English majors to the personalized learning path recommendation mode for the elderly, adjust the interface display font to a large size, simplify the learning interface, and provide corresponding operation guides and tutorials. Through the personalized learning path recommendation model for the elderly in the above mode, adjust in real time to the English learning path that suits the learning preferences of the elderly.

[0071] The English personalized learning path recommendation module makes corresponding recommendations based on the analysis results of the received user emotion tags. The process includes:

[0072] Based on the user's emotional labels, for anxiety, we recommend English movies and songs and adjust the user's learning goals; for depression, we recommend positive English articles; for anger, we recommend meditation audio and soothing English movies; for frustration, we recommend interesting learning content and share other users' learning experiences; for inferiority complex, we recommend basic grammar learning and vocabulary memorization suitable for the user's level.

[0073] The personalized learning path recommendation system for English majors of the present invention integrates an English personalized learning acquisition module, an English personalized learning processing module, an English personalized learning analysis module and an English personalized learning recommendation module. The functions of these four modules work together to ensure the normal implementation of the system functions. First, an interface is provided for users to fill in their age, learning time period, number of views of English articles and English videos, and a questionnaire survey on the use of electronic devices with multiple-choice questions, judgment questions and actual operation demonstrations. The user's age, learning time period, number of views of English articles and English videos, and the results of the electronic device proficiency questionnaire are collected, and a high-frame rate wide dynamic camera is turned on when the personal device starts the system to collect real-time dynamic images of the user's face; secondly, a convolutional neural network (CNN) is used to extract the user's age, learning time period, number of views of English articles and English videos, and the results of the electronic device proficiency questionnaire. The learning preference data of the elderly are collected, and the decision tree algorithm uses the results of the electronic device proficiency questionnaire to generate the electronic device proficiency of the elderly. Then, the LSTM algorithm is used to build a personalized learning path recommendation model for the elderly; the face position and deep facial features of the user are extracted through Haar features and convolutional neural network CNN, and then the random forest algorithm is used to build an image emotion recognition model; then, the user's age data and real-time dynamic image of the face are analyzed, and the analysis results are transmitted to the English personalized learning path recommendation module; finally, the English personalized learning path recommendation module switches to the personalized learning path recommendation mode for the elderly learning group based on the received analysis results, adjusts the interface display font to large, simplifies the learning interface and provides corresponding operation guides and tutorials. For different emotional learning groups, the system recommends different content to mobilize user enthusiasm.

[0074] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include," "comprise," or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations. The phrase "includes an element defined by..." does not exclude the presence of other identical elements in the process, method, article, or device that includes the element.

[0075] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A personalized learning path recommendation system for English majors, characterized by: The English personalized learning acquisition module is used to obtain the user's English major learning data, where the English major learning data is divided into behavioral feature data and image data; The English personalized learning processing module is divided into a learning path recommendation unit for the elderly and an image emotion recognition unit. The learning path recommendation unit is used to extract elderly learning preference data, generate elderly electronic device proficiency levels, and build a personalized learning path recommendation model for the elderly; the image emotion recognition unit is used to extract face location, deep facial features of users, and build an image emotion recognition model. The English personalized learning analysis module analyzes the user's age data and real-time dynamic facial images, and transmits the analysis results to the English personalized learning path recommendation module; English personalized learning path recommendation module, which makes corresponding recommendations based on the received analysis results; The process of constructing a personalized learning path recommendation model for the elderly includes: constructing an LSTM model, defining the input layer, hidden layer and output layer of the model, the elderly's learning preferences include daily English learning time and English learning methods, organizing the elderly's learning preference data and electronic device proficiency into a serialized form, inputting the data into the LSTM model for model training, using the constructed LSTM model to identify the elderly's learning preference data and electronic device proficiency, issuing instructions according to actual conditions, controlling the personalized learning path recommendation system for English majors to recommend content related to the elderly's learning preferences, adjusting the display mode of the interface and providing operation guides, the LSTM model adjusts model parameters through an optimizer, evaluates and optimizes the model performance, and generates a personalized learning path recommendation model for the elderly after the model optimization is completed, the LSTM model collects feedback data from the elderly after receiving the recommendation results in real time, and adjusts parameters and updates the optimization model in real time according to the feedback data; The process of making corresponding recommendations based on the received analysis results includes: receiving the analysis results of the user's age, and when receiving a signal that the user is an elderly person, switching the personalized learning path recommendation system for English majors to the personalized learning path recommendation mode for the elderly, adjusting the interface display font to large, simplifying the learning interface and providing corresponding operation guides and tutorials, and adjusting in real time to an English learning path that adapts to the learning preferences of the elderly through the personalized learning path recommendation model for the elderly.

2. The personalized learning path recommendation system for English majors according to claim 1 is characterized by: The behavioral characteristic data includes the user's age, learning period, number of views of English articles and videos, and the results of a questionnaire on electronic device proficiency. The acquisition process includes: Utilize the user's English learning device to collect the user's age data, and obtain the user's learning time period data, and the number of times the English article and English video have been viewed; A related questionnaire on the operation of English learning equipment is preset, wherein the questionnaire includes multiple-choice questions, true-or-false questions and actual operation proficiency scores on the use of English learning equipment. Based on the questionnaire survey results, the user's proficiency scores for English learning equipment are extracted.

3. The personalized learning path recommendation system for English majors according to claim 2 is characterized by: The image data is a real-time dynamic image of the user's face collected by a high frame rate wide dynamic camera, and the acquisition process includes: After the user clicks "Start Learning", the high-frame-rate dynamic camera equipped on the user's English learning device is automatically started, and the camera's frame rate and resolution are initialized; The high frame rate dynamic camera after initialization is used to collect user facial expression image data at a high frequency to obtain real-time dynamic images of the user's face.

4. The personalized learning path recommendation system for English majors according to claim 3 is characterized by: In the elderly learning path recommendation unit, the process of extracting elderly learning preference data and generating elderly electronic device proficiency includes: The user's age, learning time period, and number of views of English articles and videos are encoded and converted into a form that can be recognized by the CNN model. The CNN model is then constructed and the encoded data is input into the CNN model. Through alternating convolutional and pooling layers, the CNN model extracts the elderly's daily English learning time and English learning methods from the user's age, learning time period, and number of views of English articles and videos. A fully connected layer is then used to integrate the extracted features to output the elderly's learning preference data. The results of the electronic device use proficiency questionnaire were pre-processed and three levels of electronic device use proficiency were set: the accuracy of the primary questionnaire was 30%, the accuracy of the intermediate questionnaire was 60%, and the accuracy of the advanced questionnaire was 90%. Users aged 60 and above were also classified as elderly. The questionnaire accuracy and user age are selected as features to construct a decision tree. The results of the electronic device proficiency questionnaire are used as a data set, and the data set is divided into a training set and a test set. The training set is used to build a decision tree, and the test set is used to verify the decision tree performance. The questionnaire results are then input into the trained decision tree model, traversing from the root node until a leaf node is reached. The proficiency of the elderly in electronic devices is output according to the category corresponding to the leaf node.

5. The personalized learning path recommendation system for English majors according to claim 4 is characterized by: In the image emotion recognition unit, the process of extracting the face position and deep features of the user's face includes: Perform preprocessing operations such as grayscale conversion, denoising, and image enhancement on the real-time dynamic image of the user's face. In the preprocessed image, based on Haar-like features and cascade classifiers, slide rectangular windows of different sizes and positions to generate Haar features. The difference between the sum of pixels in different rectangular windows is calculated to obtain the feature value reflecting the grayscale change of the region. The weak classifier parameters are adjusted and the cascade classifier is optimized. The image to be tested is converted into a grayscale image. The optimized cascade classifier is used to detect and identify the location of the facial region. Preprocess the real-time dynamic image of the user's face, build a CNN model, and input the preprocessed image into the CNN model. The convolution layer extracts local facial features, including eyes, eyebrows, lips, facial muscles, and facial contours, and generates a feature map. The pooling layer downsamples the feature map to reduce the size of the feature map while retaining important features. The fully connected layer integrates the important features output by the pooling layer and outputs deep facial features of the user, including changes in facial eyes, eyebrows, lips, facial muscles, and facial contours.

6. The personalized learning path recommendation system for English majors according to claim 5 is characterized by: In the image emotion recognition unit, the process of constructing the image emotion recognition model includes: A1. Setting negative emotion labels, wherein the negative emotion labels are composed of anxiety, depression, anger, frustration, and low self-esteem; A2. Using the random forest algorithm, set the random forest parameters and associate the facial region locations and deep facial features extracted from the real-time dynamic image of the user's face with the corresponding emotion labels; A3. Use the facial region locations and deep facial features extracted from the real-time dynamic images of the user's face and the corresponding emotion labels to form a dataset. The dataset is divided into a training set, a validation set, and a test set. Multiple subsets are extracted from the training set, and a decision tree is trained for each subset. During the generation process of each decision tree, some features are randomly selected for node splitting to construct a decision tree. The above process is repeated until the predetermined number of decision trees is reached. A4. Use the validation set to evaluate the trained random forest model. Adjust the model fit by changing the parameters. Then input the test set into the adjusted random forest model to evaluate the model performance and obtain the image emotion recognition model.

7. The personalized learning path recommendation system for English majors according to claim 6 is characterized by: The process of analyzing the user's age data and real-time dynamic facial image in the English personalized learning analysis module includes: S1. Analyze the user's age data. If the user is younger than 60 years old, the user is considered a non-elderly person. If the user is older than 60 years old, the user is considered an elderly person. S2, using the image emotion recognition model, identifies the emotions in the user's facial dynamic image in real time and outputs the corresponding emotion label; S3. Wirelessly transmit the analysis results to the English personalized learning path recommendation module.

8. The personalized learning path recommendation system for English majors according to claim 7 is characterized by: The process of the English personalized learning path recommendation module making corresponding recommendations based on the analysis results of the received user emotion tags includes: Based on the user's emotional labels, for anxiety, we recommend English movies and songs and adjust the user's learning goals; for depression, we recommend positive English articles; for anger, we recommend meditation audio and soothing English movies; for frustration, we recommend interesting learning content and share other users' learning experiences; for inferiority complex, we recommend basic grammar learning and vocabulary memorization suitable for the user's level.

Images