Handwritten formula correction and complementation system
By collecting and processing handwriting input signals in the handwriting formula correction and completion system, combining intelligent prediction algorithms and user confirmation, the problem of poor recognition effect of handwriting formulas in the existing technology is solved, and higher recognition accuracy and user participation are achieved sex.
Patent Information
- Application Number
- CN202411333672.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-24
- Publication Date
- 2025-01-07
AI Technical Summary
In the prior art, handwriting formula correction and completion systems rely on high-quality handwriting input. Sloppy handwriting leads to a significant reduction in recognition effect. Long-order complex formulas lead to incorrect or omissions in recognition results due to dense character arrangement and small writing.
A handwriting formula correction and completion system was designed, including data acquisition module, data processing module, model building module, intelligent algorithm module and user interaction module. The handwriting input signal is collected through electronic whiteboard equipment to perform noise cancellation, segmentation processing and identification. , dynamic adjustments are made by combining intelligent prediction algorithms and user historical input modes, and the results are finally confirmed by the user.
Improve the accuracy and reliability of handwritten formula recognition, ensure high-quality data processing, reduce misidentification, allow users to participate in the correction process, and improve the results to meet personal needs and system performance.
Smart Images
Figure CN119274196A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of artificial intelligence and relates to a real-time handwriting recognition technology, in particular to a handwritten formula correction and completion system. Background Art
[0002] Mathematical formulas are an indispensable part of scientific research and technical documents. Building a handwritten formula correction and completion system will help share these materials more widely, especially in online education platforms and collaboration tools. Standardized output formats make cross-platform and cross-software data exchange easier. At the same time, the handwritten formula correction and completion system can help users correct typos or recognition errors in writing through automatic correction functions. For complex formula structures, the handwritten formula correction and completion system can ensure the correct format and avoid format errors caused by manual input. It can also quickly identify handwritten formulas and convert them into editable and searchable digital formats, greatly reducing the time and error rate of manual input and improving the accuracy of formulas.
[0003] In the prior art, when correcting and completing handwritten formulas, high-quality handwriting input is usually relied on, but the recognition effect will be greatly reduced due to the overly sloppy handwriting; for long and complex formulas, since they contain more characters and the characters are arranged densely and written in small sizes when handwritten, errors or omissions may occur in the recognition results.
[0004] The present invention provides a handwritten formula correction and completion system to solve the above technical problems. Summary of the invention
[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a handwritten formula correction and completion system, which is used to solve the problem that in the prior art, when correcting and completing handwritten formulas, high-quality handwriting input is usually relied on, and the recognition effect will be greatly reduced due to the handwriting being too sloppy; for long and complex formulas, since they contain more characters and the characters are arranged densely and written in small sizes when handwritten, errors or omissions in the recognition results will occur.
[0006] To achieve the above-mentioned object, the first aspect of the present invention provides a handwritten formula correction and completion system, comprising: a data acquisition module, a data processing module, a model building module, an intelligent algorithm module, a user interaction module and a terminal processing module;
[0007] Data acquisition module: collects the user's handwriting input signal through the electronic whiteboard device, and determines whether the handwriting input data is valid according to the handwriting input signal; if yes, converts the handwriting input data into a digital signal to form preliminary handwriting input trajectory data; if not, continues to judge;
[0008] Data processing module: pre-process the preliminary handwriting input trajectory data; determine whether the noise level of the pre-processed handwriting input trajectory data is normal; if yes, segment the handwriting input trajectory data; if no, further eliminate the noise;
[0009] Model building module: retrieve and initialize the handwriting recognition model, recognize the segmentation processing result through the handwriting recognition model, and obtain the recognition result; determine whether the user needs to perform manual processing based on the confidence level of the recognition result; if yes, issue a reminder signal for processing and re-judge the processing result; if no, send the recognition result to the intelligent algorithm module;
[0010] Intelligent algorithm module: Introduces intelligent prediction algorithms and dynamically adjusts recognition results based on the user's historical input patterns and common mathematical formula structures;
[0011] User interaction module: The user adjusts the formula and confirms the final result through the electronic whiteboard device; judges whether the correction and completion of the handwritten formula are completed based on the user's operation; if yes, continues to judge; if not, outputs the final result.
[0012] Preferably, the system further includes a terminal processing module: used to implement corresponding processing measures for the received real-time signals; wherein the real-time signals include reminder signals, warning signals, and prompt signals.
[0013] Preferably, collecting the user's handwriting input signal through the electronic whiteboard device includes:
[0014] A sensor distribution map is obtained from a database, and several sensors are installed in the electronic whiteboard device according to the sensor distribution map, and the user's handwriting input signal is collected according to the sensors; wherein the handwriting input signal includes: pen stroke trajectory, pen stroke pressure intensity, and handwriting speed.
[0015] The present invention helps to improve the accuracy of input and reduce misrecognition by capturing the trajectory of pen strokes; by detecting the pressure intensity of pen strokes, it can detect the problem of interruption in the writing process; and by detecting the handwriting speed, it helps to identify the writing habits of users and improve the recognition rate.
[0016] Preferably, judging whether the handwriting input data is valid according to the handwriting input signal includes:
[0017] Retrieving the handwriting input signal, setting condition 1 to determine whether the pen stroke pressure intensity is higher than the set pressure intensity threshold; if yes, the pen stroke pressure intensity is qualified; if no, the pen stroke pressure intensity is unqualified;
[0018] The second setting condition is to judge whether the pen stroke trajectory is within the set pen stroke trajectory range; if yes, the pen stroke trajectory is qualified; if no, the pen stroke trajectory is unqualified;
[0019] The third setting condition is to determine whether the handwriting speed is lower than the set speed threshold; if yes, the handwriting speed is qualified; if no, the handwriting speed is unqualified;
[0020] The judgment results of the three conditions are comprehensively judged to determine whether the handwriting input data is valid; if yes, the handwriting input data is saved; if not, an early warning signal is issued.
[0021] It should be noted that the pressure intensity threshold, pen trajectory range and speed threshold are all set by staff based on actual experience; if the pen pressure intensity is lower than the set pressure intensity threshold, it indicates that the pen has not touched the screen or the contact between the pen and the screen is poor; if the handwriting speed is higher than the set speed threshold, it indicates that the handwriting speed is too fast and the handwritten content cannot be recognized.
[0022] The present invention avoids erroneous operations caused by slight touches by setting a pressure intensity threshold; and prevents data errors caused by accidentally exceeding boundaries by limiting the pen strokes to within a set pen stroke trajectory range.
[0023] Preferably, the comprehensive judgment result of the three conditions to judge whether the handwriting input data is valid includes:
[0024] The judgment results of conditions one, two and three are retrieved to determine whether the judgment results of conditions one, two and three are all qualified. If yes, the handwriting input data is valid; if not, the handwriting input data is invalid.
[0025] The present invention improves the overall quality of the handwritten data by ensuring that the three aspects of pen pressure intensity, pen trajectory and handwriting speed meet the requirements; and reduces data errors caused by blurred handwriting, position deviation or improper speed through strict data screening.
[0026] Preferably, the preprocessing of the preliminary handwriting input trajectory data includes:
[0027] The handwriting input trajectory data is retrieved, the handwriting input trajectory data is smoothed by using the Bezier curve fitting technology, and the noise and error of the handwriting input trajectory data after the smoothing process is eliminated by using Gaussian filtering.
[0028] The present invention generates a smooth curve by using a Bezier curve to approximate the original trajectory points, which helps to remove jitter and irregularity in the handwriting trajectory and make the curve smoother.
[0029] Preferably, the determining whether the noise level of the preprocessed handwriting input trajectory data is normal includes:
[0030] A normal noise level threshold is set, the noise level of the preprocessed handwriting input trajectory data is retrieved, and it is determined whether the noise level of the preprocessed handwriting input trajectory data is lower than the normal noise level threshold; if yes, the noise level of the preprocessed handwriting input trajectory data is normal; if no, the noise level of the preprocessed handwriting input trajectory data is abnormal.
[0031] It should be noted that the normal noise level threshold is set based on the actual experience of the staff.
[0032] The present invention sets a normal noise level threshold so that all processed data are evaluated according to the same standard, thereby ensuring consistency and fairness.
[0033] Preferably, the segmentation process of the handwriting input trajectory data includes:
[0034] The preliminary handwriting input trajectory data is retrieved, and the handwriting input trajectory data is segmented through the trajectory segmentation function built into the electronic whiteboard device to form a number of independent strokes.
[0035] The present invention decomposes the complex overall trajectory into simple parts, which helps the algorithm to more accurately identify each independent stroke; at the same time, it removes unnecessary connections or overlapping parts, greatly reducing the possibility of misrecognition.
[0036] Preferably, the identifying the segmentation processing result by a handwriting recognition model includes:
[0037] Retrieve several independent strokes and obtain common mathematical symbols and letters from the database, recognize several independent strokes, and determine whether they are the corresponding mathematical symbols and letters in the database; if so, combine consecutive characters into a mathematical formula, and build a formula structure recognition model to recognize the structure and components of the formula; if not, continue to judge; wherein, the handwriting recognition model is built based on an artificial intelligence model.
[0038] The present invention can more accurately identify strokes by using mathematical symbols and letters in a database as references; and can enable the system to more quickly understand the components of a formula and their interrelationships by using a formula structure recognition model.
[0039] Preferably, the handwriting recognition model is constructed based on an artificial intelligence model, including:
[0040] The structure and components of the formula are used as the recognition input sequence to obtain standard training data; wherein the standard training data includes standard input data consistent with the content attributes of the recognition input sequence and standard output data consistent with the content attributes of the formula structure recognition;
[0041] The artificial intelligence model is trained using standard training data, and the trained artificial intelligence model is marked as a formula structure recognition model; wherein the artificial intelligence model includes a convolutional neural network model or a long short-term memory neural network model.
[0042] Preferably, judging whether the user needs to perform manual processing according to the confidence level of the recognition result includes:
[0043] The confidence level of the recognition result is obtained from the database, and it is determined whether the confidence level of the recognition result is lower than a normal threshold. If yes, a prompt signal is issued to remind the user to perform manual processing; if no, the determination is continued; wherein the manual processing methods include: re-entering and manual correction.
[0044] It should be noted that confidence is the range of credibility of the interval estimate of the population identification parameter in statistics, which is determined by the confidence level; the confidence level refers to the probability that the value of the population identification parameter falls within a certain range of the identification sample statistical value. The larger the confidence interval, the higher the confidence level.
[0045] The present invention screens out potentially inaccurate results by setting a normal threshold, which helps to improve the accuracy of overall recognition; and manually confirms or corrects results with low confidence levels to avoid directly applying erroneous results to actual scenarios.
[0046] Preferably, the introduction of an intelligent prediction algorithm comprises:
[0047] Mark the given history input as x 1:t , label the model parameters as θ, and give the given history input x 1:t The conditional probability of the model parameters θ is denoted as P(y|x 1:t , θ), the y with the largest conditional probability is taken as the next prediction input and marked as argmaxP(y|x 1:t , θ); where θ is a set of all real numbers and t is a constant greater than 0;
[0048] By formula y t =argmaxP(y|x 1:t , θ) is calculated to get the next prediction input.
[0049] It should be noted that the setting of the model parameter θ during the training process will be affected by factors such as initialization method, regularization technique, gradient clipping, learning rate, and model architecture.
[0050] The present invention can improve the accuracy and reliability of predictions by using historical data and model parameters for predictions; by calculating conditional probabilities, the risks of prediction results can be quantified to support more informed decision-making; and as time goes by, new data can be continuously collected, the model can be continuously optimized, and the prediction performance can be improved.
[0051] Preferably, the dynamically adjusting the recognition result based on the user's historical input mode and common mathematical formula structure includes:
[0052] The user's historical input pattern and common mathematical formula structure are obtained from the database, and the next prediction result calculated according to the formula is retrieved to determine whether the next prediction result conforms to the common mathematical formula structure. If so, the formula is corrected and completed according to the user's historical input pattern; if not, the prediction is made again.
[0053] The present invention can improve the accuracy of prediction results by matching the user's historical input patterns and common mathematical formula structures; based on structured mathematical formulas, the prediction results are made more consistent with mathematical rules and logic, thereby improving accuracy.
[0054] Preferably, the user adjusts the formula and confirms the final result through an electronic whiteboard device, including:
[0055] The formula after recognition, correction and completion is retrieved and displayed on the electronic whiteboard device. The user determines whether there is a matching formula based on the actual required formula; if yes, the corresponding formula is used as the final result; if not, the user further adjusts the formula through operation; wherein the operation methods include: touch and handwriting.
[0056] The present invention further improves recognition accuracy by allowing users to manually adjust and confirm recognition results; users can immediately see the recognized, corrected and completed formulas and quickly determine whether they meet their needs; and touch and handwriting operations allow users to intuitively adjust formulas, improving the convenience and intuitiveness of operations.
[0057] Preferably, judging whether the correction and completion of the handwritten formula is completed according to the user's operation includes:
[0058] Retrieve the adjusted formula, reintroduce the intelligent prediction algorithm, and dynamically adjust the recognition result based on the user's historical input pattern and common mathematical formula structure; determine whether the recognition result meets the user's actual formula requirements; if yes, complete the correction and completion of the handwritten formula; if no, continue to judge.
[0059] The present invention determines again whether the correction and completion of the handwritten formula are completed through the user's operation, so as to ensure the accuracy and consistency of the final result.
[0060] Compared with the prior art, the present invention has the following beneficial effects:
[0061] 1. The present invention can ensure that only high-quality data is further processed by judging the validity of handwritten input data, thereby reducing the burden of subsequent processing; noise elimination in the preprocessing stage can improve the purity of the data and reduce the probability of misrecognition; the accuracy and reliability of recognition can be improved by initializing and training the handwritten formula recognition model; the introduction of intelligent prediction algorithms and dynamic adjustment based on the user's historical input patterns and common mathematical formula structures can further improve the accuracy of recognition.
[0062] 2. Users can participate in the process of adjusting and confirming the formula to ensure that the results meet their personal needs; users can make adjustments based on their personal preferences to provide more flexible services; user adjustments can be used as feedback to optimize the algorithm and improve the performance of the system; by using historical data and model parameters for prediction, the accuracy and reliability of the prediction can be improved; by calculating conditional probabilities, the risks of the prediction results can be quantified to support more informed decision-making; over time, new data can be continuously collected, the model can be continuously optimized, and the performance of the prediction can be improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0064] Figure 1 A schematic diagram of the relationship between modules in the formula correction and completion system of the present invention;
[0065] Figure 2 A schematic diagram of the workflow of the formula correction and completion system of the present invention;
[0066] Figure 3 A schematic diagram of the data processing steps of the present invention;
[0067] Figure 4 It is a schematic diagram of the formula identification and correction process of the present invention. DETAILED DESCRIPTION
[0068] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0069] See also Figure 1-Figure 2, the first aspect of the present invention provides a handwritten formula correction and completion system, including: a data acquisition module, a data processing module, a model building module, an intelligent algorithm module, a user interaction module and a terminal processing module;
[0070] Data acquisition module: collects the user's handwriting input signal through the electronic whiteboard device, and determines whether the handwriting input data is valid according to the handwriting input signal; if yes, converts the handwriting input data into a digital signal to form preliminary handwriting input trajectory data; if not, continues to judge;
[0071] Data processing module: pre-process the preliminary handwriting input trajectory data; determine whether the noise level of the pre-processed handwriting input trajectory data is normal; if yes, segment the handwriting input trajectory data; if no, further eliminate the noise;
[0072] Model building module: retrieve and initialize the handwriting recognition model, recognize the segmentation processing result through the handwriting recognition model, and obtain the recognition result; determine whether the user needs to perform manual processing based on the confidence level of the recognition result; if yes, issue a reminder signal for processing and re-judge the processing result; if no, send the recognition result to the intelligent algorithm module;
[0073] Intelligent algorithm module: Introduces intelligent prediction algorithms and dynamically adjusts recognition results based on the user's historical input patterns and common mathematical formula structures;
[0074] User interaction module: The user adjusts the formula and confirms the final result through the electronic whiteboard device; judges whether the correction and completion of the handwritten formula are completed according to the user's operation; if yes, continues to judge; if not, outputs the final result;
[0075] Terminal processing module: used to implement corresponding processing measures for the received real-time signals; among which, the real-time signals include reminder signals, early warning signals, and prompt signals.
[0076] See also Figure 3 ,Data processing steps, obtain the sensor distribution map from the database, install several sensors in the electronic whiteboard device according to the sensor distribution map, and collect the user's handwriting input signal according to the sensor; wherein the handwriting input signal includes: pen stroke trajectory, pen stroke pressure intensity, and handwriting speed;
[0077] The setting condition 1 is to determine whether the pen stroke pressure intensity is higher than the set pressure intensity threshold; if yes, the pen stroke pressure intensity is qualified; if no, the pen stroke pressure intensity is unqualified;
[0078] The second setting condition is to judge whether the pen stroke trajectory is within the set pen stroke trajectory range; if yes, the pen stroke trajectory is qualified; if no, the pen stroke trajectory is unqualified;
[0079] The third setting condition is to determine whether the handwriting speed is lower than the set speed threshold; if yes, the handwriting speed is qualified; if no, the handwriting speed is unqualified;
[0080] The judgment results of the three conditions are combined to determine whether the handwriting input data is valid; if yes, the handwriting input data is saved; if not, an early warning signal is issued;
[0081] Determine whether the results of the judgment of conditions one, two, and three are all qualified. If yes, the handwriting input data is valid, and the handwriting input data is converted into a digital signal to form preliminary handwriting input trajectory data; if no, the handwriting input data is invalid, and continue to judge;
[0082] Preprocess the preliminary handwriting input trajectory data; set a normal noise level threshold, retrieve the noise level of the preprocessed handwriting input trajectory data, and determine whether the noise level of the preprocessed handwriting input trajectory data is lower than the normal noise level threshold; if yes, the noise level of the preprocessed handwriting input trajectory data is normal, and the handwriting input trajectory data is segmented through the trajectory segmentation function built into the electronic whiteboard device to form several independent strokes; if no, the noise level of the preprocessed handwriting input trajectory data is abnormal, and continue to determine.
[0083] For example, a sensor distribution map is obtained from a database, and several sensors are installed in the electronic whiteboard device according to the sensor distribution map. The three groups of pen stroke trajectories, pen stroke pressure intensity, and handwriting speed of the user are collected according to the sensors and recorded as group A, group B, and group C respectively; the pressure intensity threshold is set to 0.5N, the speed threshold is set to 0.3m / s, and the pen stroke trajectory range is within the electronic whiteboard;
[0084] Among them, the pen stroke pressure strength of group A is 0.3N, which is lower than the pressure strength threshold. Therefore, the pen stroke pressure strength of group A is unqualified, and the handwriting input data is invalid;
[0085] The pen pressure intensity of group B is 1N, and the handwriting speed is 0.5m / s, which is higher than the set speed threshold. The handwriting speed of group B is unqualified, and the handwriting input data is invalid;
[0086] The pen pressure intensity of group C is 1N, the handwriting speed is 0.23m / s, and the pen trajectory range is also within the electronic whiteboard. Therefore, the handwriting input data of group C is valid. The handwriting input data is converted into digital signals to form preliminary handwriting input trajectory data.
[0087] See also Figure 4, formula recognition and correction process, call and initialize the handwriting recognition model, call several independent strokes and obtain common mathematical symbols and letters from the database, recognize several independent strokes, and determine whether they are the corresponding mathematical symbols and letters in the database; if yes, combine the continuous characters into a mathematical formula, and build a formula structure recognition model to recognize the structure and components of the formula; if no, continue to judge;
[0088] Obtain the confidence of the recognition result from the database, and determine whether the confidence of the recognition result is lower than a normal threshold. If yes, a prompt signal is issued to remind the user to perform manual processing; if no, the determination is continued; wherein the manual processing method includes: re-entering and manual correction;
[0089] Mark the given history input as x 1:t , label the model parameters as θ, and give the given history input x 1:t The conditional probability of the model parameters θ is denoted as P(y|x 1:t , θ), the y with the largest conditional probability is taken as the next prediction input and marked as argmaxP(y|x 1:t , θ); where θ is a set of all real numbers and t is a constant greater than 0;
[0090] By formula y t =argmaxP(y|x 1:t ,θ) calculates the next prediction input;
[0091] Obtain the user's historical input pattern and common mathematical formula structure from the database, retrieve the next prediction result calculated according to the formula, and determine whether the next prediction result conforms to the common mathematical formula structure. If yes, correct and complete the formula according to the user's historical input pattern; if no, re-predict;
[0092] The formula after recognition, correction and completion is retrieved and displayed on the electronic whiteboard device, and the user determines whether there is a formula that meets the actual needs; if yes, the corresponding formula is used as the final result; if no, the user further adjusts the formula through operation; wherein the operation methods include: touch and handwriting;
[0093] Retrieve the adjusted formula, reintroduce the intelligent prediction algorithm, and dynamically adjust the recognition result based on the user's historical input pattern and common mathematical formula structure; determine whether the recognition result meets the user's actual formula requirements; if yes, complete the correction and completion of the handwritten formula; if no, continue to judge.
[0094] For example, a user has entered a partial formula by hand, such as Next, the user may want to write "sin(x)" or some other symbol; now suppose that at time t, the user enters The input at time t As history input x 1:t ;
[0095] Now we introduce intelligent prediction algorithm, based on historical input As well as the training parameter θ, the next input y is predicted to be "sin(x)", "cos(x)", or other symbols, where θ is 1 at this moment; the conditional probability P(y|x) corresponding to "sin(x)" is obtained according to the intelligent prediction algorithm. 1:t ,θ)=0.7, the conditional probability P(y|x) corresponding to “cos(x)” 1:t ,θ)=0.2, the total probability of other symbols is 0.1;
[0096] Since the conditional probability P(y|x) corresponding to “sin(x)” 1:t ,θ) is the largest, so the next step is to predict the input y t will be "sin(x)".
[0097] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.
[0098] The working principle of the present invention is as follows: the present invention collects the handwriting input signal of the user through the electronic whiteboard device, and judges whether the handwriting input data is valid according to the handwriting input signal; if yes, the handwriting input data is converted into a digital signal to form preliminary handwriting input trajectory data; if no, continue to judge; pre-process the preliminary handwriting input trajectory data; judge whether the noise level of the pre-processed handwriting input trajectory data is normal; if yes, segment the handwriting input trajectory data; if no, perform further noise elimination; retrieve and initialize the handwriting recognition model, and recognize the segmentation processing result through the handwriting recognition model; judge whether the user needs to perform manual processing according to the confidence of the recognition result; introduce an intelligent prediction algorithm, and dynamically adjust the recognition result in combination with the user's historical input mode and common mathematical formula structure; the user adjusts the formula and confirms the final result through the electronic whiteboard device; judge whether the correction and completion of the handwriting formula are completed according to the user's operation; if yes, continue to judge; if no, output the final result.
[0099] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.
Claims
1. A handwritten formula correction and completion system, characterized in that: include: Data acquisition module, data processing module, model building module, intelligent algorithm module, user interaction module and terminal processing module; Data collection module: collects the user's handwriting input signal through the electronic whiteboard device, and determines whether the handwriting input data is valid according to the handwriting input signal; If yes, the handwriting input data is converted into a digital signal to form preliminary handwriting input trajectory data; If not, continue to judge; Data processing module: pre-processes preliminary handwriting input trajectory data; determines whether the noise level of the pre-processed handwriting input trajectory data is normal; Yes, segment the handwriting input trajectory data; If not, further noise elimination is performed; Model building module: retrieve and initialize the handwriting recognition model, recognize the segmentation processing result through the handwriting recognition model, and obtain the recognition result; determine whether the user needs to perform manual processing based on the confidence level of the recognition result; If yes, a reminder signal is issued for processing and the processing result is re-judged; if no, the recognition result is sent to the intelligent algorithm module; Intelligent algorithm module: Introduces intelligent prediction algorithm to dynamically adjust recognition results based on the user's historical input pattern and common mathematical formula structure; User interaction module: The user adjusts the formula and confirms the final result through the electronic whiteboard device; judges whether the correction and completion of the handwritten formula are completed according to the user's operation; if yes, continues to judge; If not, output the final result.
2. The handwritten formula correction and completion system according to claim 1, characterized in that: The system further includes a terminal processing module: used to implement corresponding processing measures for the received real-time signals; wherein the real-time signals include reminder signals, early warning signals, and prompt signals.
3. The handwritten formula correction and completion system according to claim 1, characterized in that: The collecting of the user's handwriting input signal by the electronic whiteboard device includes: A sensor distribution map is obtained from a database, and several sensors are installed in the electronic whiteboard device according to the sensor distribution map, and the user's handwriting input signal is collected according to the sensors; wherein the handwriting input signal includes: pen stroke trajectory, pen stroke pressure intensity, and handwriting speed.
4. The handwritten formula correction and completion system according to claim 1, characterized in that: The step of judging whether the handwriting input data is valid according to the handwriting input signal comprises: Retrieving the handwriting input signal, setting condition 1 to determine whether the pen stroke pressure intensity is higher than the set pressure intensity threshold; if yes, the pen stroke pressure intensity is qualified; if no, the pen stroke pressure intensity is unqualified; The second setting condition is to judge whether the pen stroke trajectory is within the set pen stroke trajectory range; if yes, the pen stroke trajectory is qualified; if no, the pen stroke trajectory is unqualified; The third setting condition is to determine whether the handwriting speed is lower than the set speed threshold; if yes, the handwriting speed is qualified; if no, the handwriting speed is unqualified; The judgment results of the three conditions are comprehensively judged to determine whether the handwriting input data is valid; if yes, the handwriting input data is saved; if not, an early warning signal is issued.
5. The handwritten formula correction and completion system according to claim 4, characterized in that: The judgment result of the three conditions is used to judge whether the handwriting input data is valid, including: The judgment results of conditions one, two and three are retrieved to determine whether the judgment results of conditions one, two and three are all qualified. If yes, the handwriting input data is valid; if not, the handwriting input data is invalid.
6. The handwritten formula correction and completion system according to claim 1, characterized in that: The preprocessing of the preliminary handwriting input trajectory data includes: The handwriting input trajectory data is retrieved, the handwriting input trajectory data is smoothed by using the Bezier curve fitting technology, and the noise and error of the handwriting input trajectory data after the smoothing process is eliminated by using Gaussian filtering.
7. The handwritten formula correction and completion system according to claim 1, characterized in that: The determining whether the noise level of the preprocessed handwriting input trajectory data is normal includes: A normal noise level threshold is set, the noise level of the preprocessed handwriting input trajectory data is retrieved, and it is determined whether the noise level of the preprocessed handwriting input trajectory data is lower than the normal noise level threshold; if yes, the noise level of the preprocessed handwriting input trajectory data is normal; if no, the noise level of the preprocessed handwriting input trajectory data is abnormal.
8. The handwritten formula correction and completion system according to claim 1, characterized in that: The segmentation process of the handwriting input trajectory data includes: The preliminary handwriting input trajectory data is retrieved, and the handwriting input trajectory data is segmented through the trajectory segmentation function built into the electronic whiteboard device to form a number of independent strokes.
9. The handwritten formula correction and completion system according to claim 1, characterized in that: The identifying the segmentation processing result by the handwriting recognition model includes: Retrieve several independent strokes and obtain common mathematical symbols and letters from the database, recognize several independent strokes, and determine whether they are the corresponding mathematical symbols and letters in the database; if so, combine consecutive characters into a mathematical formula, and build a formula structure recognition model to recognize the structure and components of the formula; if not, continue to judge; wherein, the handwriting recognition model is built based on an artificial intelligence model.
10. The handwritten formula correction and completion system according to claim 9, characterized in that: The handwriting recognition model is constructed based on an artificial intelligence model, including: The structure and components of the formula are used as the recognition input sequence to obtain standard training data; wherein the standard training data includes standard input data consistent with the content attributes of the recognition input sequence and standard output data consistent with the content attributes of the formula structure recognition; The artificial intelligence model is trained using standard training data, and the trained artificial intelligence model is marked as a handwriting recognition model; wherein the artificial intelligence model includes a convolutional neural network model or a long short-term memory neural network model.
11. The handwritten formula correction and completion system according to claim 1, characterized in that: The determining whether the user needs to perform manual processing according to the confidence level of the recognition result includes: The confidence level of the recognition result is obtained from the database, and it is determined whether the confidence level of the recognition result is lower than a normal threshold. If yes, a prompt signal is issued to remind the user to perform manual processing; if no, the determination is continued; wherein the manual processing methods include: re-entering and manual correction.
12. The handwritten formula correction and completion system according to claim 1, characterized in that: The introduction of intelligent prediction algorithm includes: Mark the given history input as x 1:t , label the model training parameters as θ, and give the given history input x 1:t The conditional probability of the model parameters θ is denoted as P(y|x 1:t , θ), the y with the largest conditional probability is taken as the next prediction input and marked as argmaxP(y|x 1:t , θ); where θ is a set of all real numbers and t is a constant greater than 0; By formula y t =argmaxP(y|x 1:t , θ) is calculated to get the next prediction input.
13. The handwritten formula correction and completion system according to claim 1, characterized in that: The dynamically adjusting the recognition result by combining the user's historical input mode and common mathematical formula structure includes: The user's historical input pattern and common mathematical formula structure are obtained from the database, and the next prediction result calculated according to the formula is retrieved to determine whether the next prediction result conforms to the common mathematical formula structure. If so, the formula is corrected and completed according to the user's historical input pattern; if not, the prediction is made again.
14. The handwritten formula correction and completion system according to claim 1, characterized in that: The user adjusts the formula and confirms the final result through the electronic whiteboard device, including: The formula after recognition, correction and completion is retrieved and displayed on the electronic whiteboard device. The user determines whether there is a matching formula based on the actual required formula; if yes, the corresponding formula is used as the final result; if not, the user further adjusts the formula through operation; wherein the operation methods include: touch and handwriting.
15. The handwritten formula correction and completion system according to claim 1, characterized in that: The step of judging whether the correction and completion of the handwritten formula is completed according to the user's operation includes: Retrieve the adjusted formula, reintroduce the intelligent prediction algorithm, and dynamically adjust the recognition result based on the user's historical input pattern and common mathematical formula structure; determine whether the recognition result meets the user's actual formula requirements; if yes, complete the correction and completion of the handwritten formula; if no, continue to judge.