Desk system based on artificial intelligence and correction auxiliary learning method

Abstract

The invention discloses a desk system based on artificial intelligence and a correction auxiliary learning method. The system comprises an intelligent desk, an intelligent seat and a WIFIamp. The system comprises a USB module, a power supply module, a communication module, a mechanical execution module, a data processing module, a data transmission module and a signal control module. The intelligent desk comprises an intelligent desk vertical desktop, a camera assembly arranged on the intelligent desk vertical desktop, a first circuit control assembly, a data storage assembly, an intelligent terminal, an intelligent desk horizontal desktop and a first intelligent adjusting and supporting assembly arranged on the intelligent desk horizontal desktop. The intelligent seat comprises a second intelligent adjusting supporting assembly arranged on a seat back of the intelligent seat, a second circuit control assembly and a gravity sensing assembly, wherein the second circuit control assembly and the gravity sensing assembly are arranged on the intelligent seat. According to the invention, the sitting posture of the student can be monitored and corrected in real time, the healthy learning habit is effectively promoted, the advanced artificial intelligence technology is integrated, and automatic homework correction and personalized learning tutoring are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of smart desks, and in particular to an artificial intelligence-based desk system and a correction-assisted learning method. Background Art

[0002] Traditional desks are made of wood or plastic and only have a simple supporting function. With the improvement of science and technology and people's pursuit of high-quality life, the style and material of desks have also undergone earth-shaking changes, including variable height desks (which can be manually adjusted as students rise to meet the needs of students from elementary school to high school), various storage and telescopic devices, integrated plug boards, desk lamps, etc. However, these desks are only used as ordinary learning aids.

[0003] When students are studying and doing homework, they need their parents to provide guidance, correct their homework and explain the wrong questions. There is a saying that often circulates on the Internet: "If you don't help with homework, the father is kind and the son is filial; if you help with homework, there will be chaos." Parents are prone to conflicts with students during the process of helping with homework, causing family disharmony. In addition, when doing homework, students often make various small movements, such as lying on the desk or being absent, which seriously affects the enthusiasm and patience of parents, causing more "fierce" conflicts on the spot, which greatly affects family harmony.

[0004] The main problems with existing desks and parent-assisted teaching are: first, parents are inefficient in correcting homework, and their emotions are easily affected by the students' homework situation, resulting in low overall efficiency and effectiveness; second, the various homework assignments are relatively scattered, especially wrong questions are scattered in various books, which is not conducive to later review; third, parents need to pay attention to students' sitting posture, pen-holding posture, etc., and correct them during the homework tutoring process, which can easily lead to conflicts and cause family disharmony. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the purpose of the present invention is to provide an artificial intelligence-based desk system and a corrective assisted learning method. The system has the function of real-time monitoring and correction of students' sitting posture, can effectively promote healthy learning habits, and through the integration of advanced artificial intelligence technology, realize automatic correction of homework and personalized learning guidance, significantly improving learning efficiency and quality. In addition, the present invention can also automatically record key moments in the learning process, generate a growth manual, enhance the interaction and understanding between parents and children, which not only optimizes the family learning environment and reduces the tutoring burden of parents, but also promotes students' independent learning and all-round development through intelligent technology.

[0006] In a first aspect, the present invention provides a desk system based on artificial intelligence, comprising: a smart desk, a smart chair, a WIFI&USB module, a power supply module, a communication module, a mechanical execution module, a data processing module, a data transmission module and a signal control module; the data processing module is used for automatic correction of homework and classification of wrong questions;

[0007] The smart desk comprises a camera component, a first circuit control component, a data storage component, a smart terminal, a horizontal desktop of the smart desk, and a first smart adjustment support component disposed on the horizontal desktop of the smart desk;

[0008] The smart chair includes a second intelligent adjustment support component arranged on the back of the smart chair and a second circuit control component and a gravity sensing component arranged on the seat of the smart chair; the camera component and the gravity sensing component are used for sitting posture detection, and the first intelligent adjustment support component and the second intelligent adjustment support component are used for sitting posture correction.

[0009] The desk system effectively promotes healthy learning habits by real-time monitoring and correcting students' sitting postures, and integrates advanced artificial intelligence technology to achieve automated homework correction and personalized learning guidance. Furthermore, it also includes a binocular scanning device on the vertical desktop of the smart desk, which can be used to record the three-dimensional data of students' postures / expressions when studying, and output them into standard models for 3D printing, so as to establish a three-dimensional data model library that records students' growth.

[0010] Furthermore, it also includes a desk lamp assembly arranged on the vertical desktop of the smart desk, and the desk lamp assembly can automatically control the brightness of the desk lamp according to the ambient brightness conditions.

[0011] Furthermore, it also includes a distance measuring sensor arranged on the vertical desktop of the smart desk, and the distance measuring sensor can be used for sitting posture detection.

[0012] In a second aspect, the present invention provides a correction-assisted learning method, which is implemented by using the above-mentioned artificial intelligence-based desk system, comprising:

[0013] S100: Performing sitting posture detection on students through distance measurement, angle measurement and gravity distribution status, and calculating posture angle score, head distance score and gravity distribution score. The sitting posture detection is completed based on the comprehensive sitting posture score calculated according to the posture angle score, head distance score and gravity distribution score, and the first intelligent adjustment support component (16) and the second intelligent adjustment support component (21) are used for sitting posture correction.

[0014] S200: Construct a deep learning model to automatically correct and coach homework, and classify the correct and incorrect questions according to categories and save them in the data storage component (14), transmit them to the smart terminal through the USB / WIFI module, or directly print them through a printer peripheral for students to review.

[0015] Furthermore, step S100 includes:

[0016] S101, perform posture angle detection:

[0017] θ=ω1x+ω2y+b

[0018] In the formula, θ represents the posture angle, x and y represent the coordinates of the key points extracted from the camera image, ω1 represents the first weight coefficient, ω2 represents the second weight coefficient, and b represents the bias term;

[0019] S102, performing head distance detection:

[0020] d=Dd min

[0021] Where d is the distance between the distance sensor and the head, D is the sensor reading, and d min Indicates the reference minimum distance;

[0022] S103, perform gravity distribution detection:

[0023]

[0024] In the formula, n represents the number of gravity sensing points, and G represents the average gravity intensity.

[0025] S104, respectively calculating the posture angle score, the head distance score and the gravity distribution score:

[0026]

[0027] Where f(θ), g(d) and Respectively represent the posture angle score, head distance score and gravity distribution score; α, β, γ represent the third weight coefficient, the fourth weight coefficient and the fifth weight coefficient; θ ideal , d ideal , Respectively represent the angle, distance and gravity distribution value of the target posture;

[0028] S105, calculating the comprehensive scoring function:

[0029]

[0030] In the formula, S represents the comprehensive score of sitting posture;

[0031] Further, including:

[0032] S106, defining a comprehensive score threshold T, and when the sitting posture comprehensive score S is less than the sitting posture comprehensive score threshold T, performing sitting posture correction;

[0033] S107, controlling the first adjustment support assembly and the second adjustment support assembly to perform sitting posture correction until the sitting posture comprehensive score S is greater than the sitting posture comprehensive score threshold T.

[0034] Further, step S200 includes:

[0035] S201, using a pre-trained model as a basic feature extractor, and adding several fully connected layers on top of the pre-trained model for classification or regression tasks;

[0036] S202. Build a multi-task learning framework, in which one sub-network is used for question recognition and another sub-network is used for answer parsing, and use a shared underlying feature extractor to reduce the amount of computation and improve the generalization ability of the model;

[0037] S203, use the cross entropy loss function to measure the difference between the classification result and the true label. For multi-classification problems, the first loss function is expressed as:

[0038]

[0039] Where, L classification represents the first loss function; n represents the number of training samples; m represents the number of questions; tij represents the true label of the jth question in the ith sample, and pij represents the predicted probability of the jth question in the ith sample;

[0040] S204, using the mean square error loss function to measure the difference between the answer parsing result and the true answer, the second loss function is expressed as:

[0041]

[0042] In the formula, n represents the number of training samples, s i represents the true answer of the i-th sample, Represents the predicted answer of the i-th sample;

[0043] Furthermore, step S200 further includes:

[0044] S205, the total loss function is expressed as:

[0045] L total =αL classification +βL regression

[0046] In the formula, α and β represent weighting factors;

[0047] S206, using the optimizer to update the weight of the model, and using the learning rate decay strategy to improve the convergence speed of the model;

[0048] S207, perform hyperparameter adjustment and model evaluation and tuning to complete the construction of the deep learning model.

[0049] Furthermore, step S200 includes: the data processing module classifies the correct and incorrect questions according to categories and stores them in the data storage component (14), and then transmits them to the smart terminal through the USB / WIFI module, or directly prints them through a printer peripheral for students to review.

[0050] Beneficial effects of the present invention:

[0051] 1. The present invention has the function of real-time monitoring and correction of students' sitting posture, effectively promoting healthy learning habits, and by integrating advanced artificial intelligence technology, it realizes automatic correction of homework and personalized learning guidance, which can significantly improve learning efficiency and quality.

[0052] 2. The present invention can automatically record key moments in the learning process, generate a growth manual, and enhance the interaction and understanding between parents and children. It not only optimizes the home learning environment and reduces the tutoring burden of parents, but also promotes students' independent learning and all-round development through intelligent technology.

[0053] 3. The present invention can automatically control the brightness of the desk lamp according to the gravity sensing component, and can automatically control the brightness of the desk lamp according to the brightness of the surrounding environment, so as to effectively protect students' eyesight.

[0054] Additional aspects and advantages of the present application will be partially given in the following description, which will become apparent from the following description, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0056] Figure 1 is a schematic diagram of the hardware composition of a smart desk system in an embodiment of the present invention;

[0057] Figure 2 It is a schematic diagram of the software composition of a smart desk system in an embodiment of the present invention;

[0058] Figure 3 is a schematic diagram of the change of the angle of the student's back posture in an embodiment of the present invention;

[0059] Figure 4is a schematic diagram of the change in the distance between the student's head and the back panel of the smart desk in an embodiment of the present invention;

[0060] Figure 5 is a flow chart of a correction-assisted learning method in an embodiment of the present invention;

[0061] Figure 6 is a flow chart of an automatic homework correction method in an embodiment of the present invention;

[0062] In all the drawings, the same figure marks represent the same technical features, specifically: 1-smart desk, 10-smart desk vertical desktop, 11-desk lamp assembly, 12-camera assembly, 13-first circuit control assembly, 14-data storage assembly, 15-smart terminal, 16-first adjustment support assembly, 17-smart desk horizontal desktop; 2-smart chair, 21-second adjustment support assembly, 22-second circuit control assembly, 23-gravity sensing assembly. DETAILED DESCRIPTION

[0063] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0064] In the description of the present invention, unless otherwise clearly specified and limited, the terms "connected", "connected", and "fixed" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0065] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0066] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as the general understanding of those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless specifically defined as in the embodiments of this application.

[0067] The present invention provides an artificial intelligence-based desk system and a correction and auxiliary learning method. The system has the function of real-time monitoring and correction of students' sitting posture, can effectively promote healthy learning habits, and through the integration of advanced artificial intelligence technology, realizes automatic correction of homework and personalized learning guidance, significantly improving learning efficiency and quality. In addition, the present invention can also automatically record key moments in the learning process, generate a growth manual, and enhance the interaction and understanding between parents and children. It not only optimizes the family learning environment and reduces the tutoring burden of parents, but also promotes students' independent learning and all-round development through intelligent technology.

[0068] Example 1

[0069] like Figure 1 Figure 2 As shown, an embodiment of the present invention provides an artificial intelligence-based desk system, including: a smart desk 1, a smart seat 2, a WIFI&USB module, a power supply module, a communication module, a mechanical execution module, a data processing module, a data transmission module and a signal control module; the smart desk 1 includes a smart desk vertical desktop 10, a camera component 12 arranged on the smart desk vertical desktop 10, a first circuit control component 13, a data storage component 14, a smart terminal 15 and a smart desk horizontal desktop 17 and a first smart adjustment support component 16 arranged on the smart desk horizontal desktop 17;

[0070] The smart chair 2 includes a second smart adjustment support component 21 arranged on the back of the smart chair 2, a second circuit control component 22 and a gravity sensing component 23 arranged on the seat of the smart chair 2; the desk system effectively promotes healthy learning habits by real-time monitoring and correcting students' sitting posture, and integrates advanced artificial intelligence technology to realize automatic correction of homework and personalized learning guidance. Among them, the WIFI / USB module is mainly used to connect with peripherals such as printers and smart terminals; the power supply module is mainly used to power various hardware of the system; the communication module is used for communication between the smart desk and the smart chair on the one hand, and for communication between the peripherals and the smart terminal on the other hand; the mechanical execution module is mainly used for the adjustment module of the posture correction device, which is responsible for adjusting the student's posture; the data processing module is mainly used for automatic correction of homework, including the recognition of homework questions, answer analysis, automatic classification of wrong questions, archiving and typesetting, etc.; the data transmission module is used to transmit data / pictures to peripherals and smart terminals, etc.; the signal control module is used for command control of the mechanical actuators of smart desks and smart chairs, and real-time dynamic adjustment of sitting posture and position. Among them, the smart terminal includes electronic devices such as mobile phones and tablets.

[0071] Furthermore, it also includes a binocular scanning device arranged on the vertical desktop 10 of the smart desk. The binocular scanning device can be used to record the three-dimensional data of the student's posture / expression when studying, and output it into a standard model for 3D printing, so as to establish a three-dimensional data model library that records the student's growth. It also includes a desk lamp component 11 arranged on the vertical desktop 10 of the smart desk. The desk lamp component 11 can automatically control the brightness of the desk lamp according to the ambient brightness. It also includes a distance sensor arranged on the vertical desktop 10 of the smart desk. The distance sensor can be used for sitting posture detection, and record the distance between the distance sensor and the head when the student is in the correct sitting posture.

[0072] Example 2

[0073] An embodiment of the present invention proposes a sitting posture correction method, which is implemented by using an artificial intelligence-based desk system as described above, and includes: performing sitting posture detection by distance measurement or angle measurement or a combination of the two.

[0074] Furthermore, if Figure 3 As shown, the angle measurement detection includes: photographing the student's sitting posture through the camera component 12, calculating the angle between the student's back plane and the desktop, and giving a voice reminder when the angle is reduced to 75°, while controlling the first intelligent adjustment support component 16 and the second intelligent adjustment support component 21 to correct the student's posture.

[0075] Furthermore, if Figure 4As shown, the distance measurement detection includes: recording the distance H between the distance sensor and the head of the student in the correct sitting posture in advance through the distance sensor; when the actual detection distance H' of the distance sensor differs from the distance H in the correct sitting posture by 15%, that is, (H'-H) / H>15%, controlling the first intelligent adjustment support component 16 and the second intelligent adjustment support component 21 to correct the student's posture.

[0076] Example 3

[0077] like Figure 5 As shown, this embodiment provides a correction assisted learning method, which is implemented by using the desk system based on artificial intelligence in the above embodiment, including:

[0078] S100: Performing sitting posture detection on students through distance measurement, angle measurement and gravity distribution state, and calculating posture angle score, head distance score and gravity distribution score, completing sitting posture detection based on the comprehensive sitting posture score calculated according to the posture angle score, head distance score and gravity distribution score, and using the first intelligent adjustment support component 16 and the second intelligent adjustment support component 21 for sitting posture correction;

[0079] S200: Construct a deep learning model to automatically correct and coach homework, and classify the correct and incorrect questions according to categories and save them in the data storage component 14, transmit them to the smart terminal through the USB / WIFI module, or directly print them through a printer peripheral for students to review.

[0080] Wherein, the step S100 includes:

[0081] S101, perform posture angle detection:

[0082] θ=ω1x+ω2y+b

[0083] In the formula, θ represents the posture angle, x and y represent the coordinates of the key points extracted from the camera image, ω1 represents the first weight coefficient, ω2 represents the second weight coefficient, and b represents the bias term;

[0084] S102, performing head distance detection:

[0085] d=Dd min

[0086] Where d is the distance between the distance sensor and the head, D is the sensor reading, and d min Indicates the reference minimum distance;

[0087] S103, perform gravity distribution detection:

[0088]

[0089] Where n represents the number of gravity sensing points, Represents the average gravity strength.

[0090] S104, respectively calculating the posture angle score, the head distance score and the gravity distribution score:

[0091]

[0092] Where f(θ), g(d) and Respectively represent the posture angle score, head distance score and gravity distribution score; α, β, γ represent the third weight coefficient, the fourth weight coefficient and the fifth weight coefficient; θ ideal , d ideal , Respectively represent the angle, distance and gravity distribution value of the target posture;

[0093] S105, calculating the comprehensive scoring function:

[0094]

[0095] In the formula, S represents the comprehensive score of sitting posture;

[0096] S106, defining a comprehensive score threshold T, and when the sitting posture comprehensive score S is less than the sitting posture comprehensive score threshold T, performing sitting posture correction;

[0097] S107, controlling the first adjustment support assembly (16) and the second adjustment support assembly (21) to perform sitting posture correction until the sitting posture comprehensive score S is greater than the sitting posture comprehensive score threshold T.

[0098] Furthermore, the step S200 includes:

[0099] S201, using a pre-trained model as a basic feature extractor, and adding several fully connected layers on top of the pre-trained model for classification or regression tasks;

[0100] S202. Build a multi-task learning framework, in which one sub-network is used for question recognition and another sub-network is used for answer parsing, and use a shared underlying feature extractor to reduce the amount of computation and improve the generalization ability of the model;

[0101] S203, use the cross entropy loss function to measure the difference between the classification result and the true label. For multi-classification problems, the first loss function is expressed as:

[0102]

[0103] Where, L classificationrepresents the first loss function; n represents the number of training samples; m represents the number of questions; tij represents the true label of the jth question in the ith sample, and pij represents the predicted probability of the jth question in the ith sample;

[0104] S204, using the mean square error loss function to measure the difference between the answer parsing result and the true answer, the second loss function is expressed as:

[0105]

[0106] In the formula, n represents the number of training samples, s i represents the true answer of the i-th sample, Represents the predicted answer of the i-th sample;

[0107] S205, the total loss function is expressed as:

[0108] L total =αL classification +βL regression

[0109] In the formula, α and β represent weighting factors;

[0110] S206, selecting an optimizer to update the weight of the model, and using a learning rate decay strategy to improve the convergence speed of the model;

[0111] S207, perform hyperparameter adjustment and model evaluation and tuning to complete the construction of the deep learning model.

[0112] Wherein, the step S200 also includes: the data processing module classifies the correct and incorrect questions according to the categories and stores them in the data storage component (14), and then transmits them to the smart terminal through the USB / WIFI module, or directly prints them through a printer peripheral for students to review.

[0113] Example 4

[0114] After the students complete their homework, the system will automatically grade the homework and explain and correct the wrong questions on the smart terminal (tablet / mobile phone, etc.)15.

[0115] First, the camera component 12 is used to capture the student's paper homework, and the content is automatically analyzed and the answers are corrected. There are two ways to correct homework, mainly as follows:

[0116] The first method is the built-in database. Each page of the paper document has a QR code. The homework is photographed by the camera component. After the QR code is recognized, the current homework is compared with the standard homework in the database (including answers). The differences are found to be wrong questions, and the wrong questions are correctly analyzed.

[0117] The second method is to use image text recognition and deep learning technology. First, the question is analyzed through the text recognition module, and then the deep learning module is used to automatically answer the assignment. Finally, the automatic answer result is compared with the recognized answer to determine whether the assignment is correct.

[0118] like Figure 6 As shown, an embodiment of the present invention proposes an automatic homework correction method, which is implemented by using an artificial intelligence-based desk system as described above, including:

[0119] S100, using the camera assembly to photograph a student's paper homework screen to obtain a homework image;

[0120] S200, preprocessing the collected operation image to improve the accuracy of text recognition;

[0121] S300, using optical character recognition technology to recognize the text in the image and convert it into an editable text format;

[0122] S400, parsing the question according to the built-in database, identifying the question type and the corresponding answer;

[0123] S500, if the recognition is successful, the recognized answer is compared with the standard answer to find out the difference, determine the wrong question, and generate explanation content according to the preset answer.

[0124] S600: If the recognition fails, the question is parsed through the deep learning model to identify the question type and the corresponding answer, determine the wrong question, and generate explanation content based on the analysis provided by the deep learning model;

[0125] Example 5

[0126] An embodiment of the present invention provides a method for classifying wrong questions, which is implemented using the above-mentioned artificial intelligence-based desk system, wherein the data processing module can classify the wrong questions according to categories and save them in the data storage system 14, including wrong question categories, time and other information, and can be transmitted to the smart terminal through the USB / WIFI module, or directly printed through a printer peripheral for students to review.

[0127] Specifically include:

[0128] Step 1: After the student completes the homework, he / she uses the camera component 12 of the smart desk to take a picture of the homework page. The system first converts the text in the image into text data through OCR technology.

[0129] Step 2: The system's built-in NLP engine analyzes the text data, identifies the question type (such as multiple-choice questions, fill-in-the-blank questions, true-or-false questions, and essay questions), and marks the question type in the data.

[0130] Step 3: The system associates the student's answer with the question type. For multiple-choice questions and true-false questions, the system identifies the selected option or the right or wrong judgment; for fill-in-the-blank questions and essay questions, the system analyzes the keywords and structure of the answer.

[0131] Step 4: The system matches the student's answer with the built-in standard answer library, where the standard answer library contains the correct answer to each question and its analysis process.

[0132] Step 5: If the student's answer does not match the standard answer, the system will judge the question as wrong and extract relevant question and answer information.

[0133] Step 6: The system categorizes the wrong questions according to question type and knowledge point. For example, all algebra questions in mathematics are classified into one category, and geometry questions are classified into another category.

[0134] Step 7: The system uses deep learning algorithms to further intelligently label wrong questions, such as "easy to make mistakes", "key questions", "extension questions", etc., to facilitate students' identification and review.

[0135] Step 8: All wrong questions and their classification information are stored in the data storage system 14 and indexed for quick retrieval and access.

[0136] Furthermore, the system recommends review plans and related learning resources to students, such as video explanations, additional exercises, etc., based on the classification and labels of wrong questions.

[0137] Furthermore, parents or students can access the wrong question information through the smart terminal 15, and choose to print out the wrong question book, or transfer it to other devices via USB / WIFI.

[0138] Furthermore, the system provides a user-friendly interface that displays wrong question classification, labels and review suggestions. Students and parents can view, manage and print wrong questions through the interface.

[0139] Furthermore, the system will track students' learning progress and review of wrong questions, and dynamically adjust learning resources and review plans based on the review results.

[0140] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0141] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A desk system based on artificial intelligence, characterized in that: include: A smart desk (1), a smart chair (2) and a smart control system; The intelligent control system includes a WIFI&USB module, a power supply module, a communication module, a mechanical execution module, a data processing module, a data transmission module and a signal control module; wherein the data processing module is used for automatic correction of homework and classification of wrong questions; The smart desk (1) comprises a smart desk vertical desktop (10), a camera component (12) arranged on the smart desk vertical desktop (10), a first circuit control component (13), a data storage component (14), a smart terminal (15), a smart desk horizontal desktop (17), and a first smart adjustment support component (16) arranged on the smart desk horizontal desktop (17); The intelligent chair (2) comprises a second intelligent adjustment support component (21), a second circuit control component (22) and a gravity sensing component (23); wherein the camera component (12) and the gravity sensing component (23) are used for sitting posture detection, and the first intelligent adjustment support component (16) and the second intelligent adjustment support component (21) are used for sitting posture correction according to the sitting posture detection result.

2. The desk system based on artificial intelligence according to claim 1, characterized in that: It also includes a binocular scanning device arranged on the vertical desktop (10) of the smart desk, and the binocular scanning device is used to record the three-dimensional data of the student's posture / expression when studying, and output it into a standard model for 3D printing, so as to establish a three-dimensional data model library that records the student's growth.

3. The desk system based on artificial intelligence according to claim 2, characterized in that: It also includes a desk lamp assembly (11) arranged on the vertical desktop (10) of the smart desk, and the desk lamp assembly (11) automatically controls the brightness of the desk lamp according to the ambient brightness.

4. The desk system based on artificial intelligence according to claim 3, characterized in that: It also includes a distance measuring sensor arranged on the vertical desktop (10) of the smart desk, and the distance measuring sensor is used for sitting posture detection.

5. A correction-assisted learning method, characterized in that: The application is implemented by an artificial intelligence-based desk system as claimed in any one of claims 1 to 4, comprising: S100: Performing sitting posture detection on students through distance measurement, angle measurement and gravity distribution status, and calculating posture angle score, head distance score and gravity distribution score. The sitting posture detection is completed based on the comprehensive sitting posture score calculated according to the posture angle score, head distance score and gravity distribution score, and the first intelligent adjustment support component (16) and the second intelligent adjustment support component (21) are used for sitting posture correction. S200: Construct a deep learning model to automatically correct and coach homework, and classify the correct and incorrect questions according to categories and save them in the data storage component (14), transmit them to the smart terminal through the USB / WIFI module, or directly print them through a printer peripheral for students to review.

6. A correction-assisted learning method according to claim 5, characterized in that: The step S100 includes: S101, perform posture angle detection: θ=ω1x+ω2y+b In the formula, θ represents the posture angle, x and y represent the coordinates of the key points extracted from the camera image, ω1 represents the first weight coefficient, ω2 represents the second weight coefficient, and b represents the bias term; S102, performing head distance detection: d=D-d min Where d is the distance between the distance sensor and the head, D is the sensor reading, and d min Indicates the reference minimum distance; S103, perform gravity distribution detection: In the formula, n represents the number of gravity sensing points, represents the average gravity intensity; S104, respectively calculating the posture angle score, the head distance score and the gravity distribution score: Where f(θ), g(d) and Respectively represent the posture angle score, head distance score and gravity distribution score; α, β, γ represent the third weight coefficient, the fourth weight coefficient and the fifth weight coefficient; θ ideal , d ideal , Respectively represent the angle, distance and gravity distribution value of the target posture; S105, calculating the comprehensive scoring function: Where S represents the comprehensive score of sitting posture.

7. A correction-assisted learning method according to claim 6, characterized in that: The step S100 further includes: S106, defining a comprehensive score threshold T, and when the sitting posture comprehensive score S is less than the sitting posture comprehensive score threshold T, performing sitting posture correction; S107, controlling the first adjustment support assembly (16) and the second adjustment support assembly (21) to perform sitting posture correction until the sitting posture comprehensive score S is greater than the sitting posture comprehensive score threshold T.

8. A correction-assisted learning method according to claim 5, characterized in that: The step S200 includes: S201, using a pre-trained model as a basic feature extractor, and adding several fully connected layers on top of the pre-trained model for classification or regression tasks; S202. Build a multi-task learning framework, in which one sub-network is used for question recognition and another sub-network is used for answer parsing, and use a shared underlying feature extractor to reduce the amount of computation and improve the generalization ability of the model; S203, use the cross entropy loss function to measure the difference between the classification result and the true label. For multi-classification problems, the first loss function is expressed as: Where, L classification represents the first loss function; n represents the number of training samples; m represents the number of questions; tij represents the true label of the jth question in the ith sample, and pij represents the predicted probability of the jth question in the ith sample; S204, using the mean square error loss function to measure the difference between the answer parsing result and the true answer, the second loss function is expressed as: In the formula, n represents the number of training samples, s i represents the true answer of the i-th sample, Represents the predicted answer of the i-th sample.

9. The method for constructing an automatic homework correction model according to claim 8, characterized in that: The step S200 further includes: S205, the total loss function is expressed as: L total =αL classification +βL regression In the formula, α and β represent weighting factors; S206, selecting an optimizer to update the weight of the model, and using a learning rate decay strategy to improve the convergence speed of the model; S207: Perform hyperparameter adjustment, model evaluation and tuning to complete the construction of the deep learning model.

10. The desk system based on artificial intelligence according to claim 7 or 8, characterized in that: The step S200 also includes: the data processing module classifies the correct and incorrect questions according to the categories and stores them in the data storage component (14), and then transmits them to the smart terminal through the USB / WIFI module, or directly prints them through a printer peripheral for students to review.

Images