Multidimensional touch data color matching system of electronic whiteboard

Abstract

The invention relates to the field of electronic whiteboards, in particular to a multi-dimensional touch data color matching system of an electronic whiteboard, which comprises a data acquisition module, a data analysis module, a dynamic adjustment module and an application display module which are in communication connection, and is characterized in that the data acquisition module is used for acquiring data of writing speed, angle, pressure and width through a sensor; the data analysis module is used for determining a color value through a comprehensive weight according to the data acquired in real time and a color matching model; the dynamic adjustment module is used for dynamically calculating a current comprehensive weight value according to the data acquired in real time; and the application display module is used for applying the calculated color value to handwriting display. The system can provide richer writing experience, enhance teaching and demonstration effects, improve the accuracy and adaptability of color matching, and achieve a real-time visible writing effect.

Description

Technical Field

[0001] The invention relates to the field of electronic whiteboards, and in particular to a multi-dimensional touch data color matching system for electronic whiteboards. Background Art

[0002] In traditional electronic whiteboard technology, writing functions are often relatively simple, and the handwriting color is usually fixed or switched in a limited number of ways, lacking flexibility and intuitiveness. For example, during teaching, teachers can only use a few preset colors for writing and marking, and it is difficult to adjust the handwriting color in real time to highlight key content according to different teaching situations and needs. In corporate training and conference presentations, speakers also face the same problem, unable to better convey key information and attract audience attention by changing the handwriting color.

[0003] In addition, the sensing dimensions of traditional electronic whiteboards for writing are relatively limited, and most of them only focus on the position information of writing, while ignoring factors such as writing speed, angle, pressure and width. These factors play an important role in the actual writing and creation process, and can provide users with richer means of expression and more personalized experience. Therefore, the existing electronic whiteboard technology has deficiencies in color matching and touch sensing, and a new technical solution is needed to solve these problems in order to meet users' higher demands in teaching, demonstration and creative expression. Summary of the invention

[0004] To solve the above problems, the present invention provides a multi-dimensional touch data color matching system for an electronic whiteboard, which can provide a richer writing experience, enhance teaching and demonstration effects, improve the accuracy and adaptability of color matching, and achieve real-time visible writing effects.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] A multi-dimensional touch data color matching system for an electronic whiteboard includes: a communication-connected data acquisition module, a data analysis module, a dynamic adjustment module, and an application display module, wherein:

[0007] The data acquisition module is used to obtain data on writing speed, angle, pressure and width through sensors;

[0008] The data analysis module is used to determine the color value through comprehensive weights based on the real-time acquired data and the color matching model;

[0009] The dynamic adjustment module is used to dynamically calculate the current comprehensive weight value based on the real-time acquired data;

[0010] The application display module is used to apply the calculated color value to the handwriting display.

[0011] Furthermore, the method of obtaining data of writing speed, angle, pressure and width through a sensor comprises the following steps:

[0012] A1: Calibrate the sensors that obtain writing speed, angle, pressure, width and environmental data one by one, check the connection between the sensors and the electronic whiteboard system, and initialize the software configuration;

[0013] A2: When the user writes or touches the electronic whiteboard, the sensor monitors the digital and analog events in real time and transmits and obtains the touch data at the same time;

[0014] A3: Filter the collected data to remove noise and normalize the data.

[0015] Furthermore, the acquisition of touch data includes: obtaining writing speed by calculating the change in touch point position per unit time; obtaining writing angle by the tangent direction of the touch track; obtaining touch pressure by the pressure value output by the sensor; and obtaining touch width by the touch width value output by the sensor.

[0016] Furthermore, obtaining the writing speed and the writing angle includes: detecting the touch position by electromagnetic induction technology, and optimizing the coordinate accuracy by multiple sampling and averaging methods;

[0017] The writing speed is calculated based on the position change and the time interval, and the obtained speed is smoothed by the moving average method;

[0018] The coordinates of the touch points are continuously recorded to form a touch trajectory. The touch trajectory is expressed as a function according to the curve fitting algorithm. The tangent direction of each point on the touch trajectory is calculated by numerical differentiation. The writing angle is then calculated by the inverse tangent function, and the angle is smoothed using a low-pass filter.

[0019] Furthermore, the color matching model includes: a color matching model formed by combining a fuzzy logic algorithm and a color matching algorithm based on machine learning through a distributed framework; and the color value is specifically a GRB color value.

[0020] Furthermore, the method of determining the color value by comprehensive weight according to the data acquired in real time and the color matching model comprises the following steps:

[0021] B1: Build a distributed framework based on Spark core components and Hadoop distributed file system, use HDFS for data storage, and monitor the running status of Spark jobs;

[0022] B2: Fuzzify the touch data of writing speed, angle, pressure and width through fuzzy logic algorithm, use the touch data as input variable of fuzzy logic algorithm, and color value as output variable. Determine fuzzy rules and make inferences based on experience and knowledge domain, and convert the fuzzy set of output variables into specific color value through maximum membership method.

[0023] B3: According to the data processing of the color matching algorithm based on machine learning, the fuzzy logic algorithm and the color matching algorithm based on machine learning are assigned comprehensive weights according to the specific requirements of the application scenario, and the color values ​​of the two algorithms are merged by weighted average according to the comprehensive weights;

[0024] B4: Continuously monitor touch data and environmental changes, regularly evaluate and optimize color matching models, analyze algorithm performance and effects, and identify potential problems and areas for improvement.

[0025] Furthermore, the processing of data according to the color matching algorithm based on machine learning comprises the following steps:

[0026] B31: Collect touch data samples and corresponding expected color values ​​to form a training data set, and divide the data set into a training set, a validation set, and a test set;

[0027] B32: Extract statistical features from the touch data set, perform feature dimensionality reduction on the statistical features using the principal component analysis method, train the support vector machine model using the training set, and adjust the model parameters using the cross-validation method to minimize the prediction error;

[0028] B33: Evaluate the trained model through the validation set. When new touch data is input, the trained model is used to predict and obtain the corresponding color value.

[0029] Furthermore, the current comprehensive weight value is dynamically calculated based on the data acquired in real time, including: adjusting the weight of the color matching model as a reinforcement learning problem according to the deep Q network reinforcement learning algorithm, constructing and training the deep Q network by defining the state space, action space and reward function, thereby realizing online weight adjustment.

[0030] Furthermore, the state space includes: touch data of the current writing speed, angle, pressure and width, environmental data of the light intensity of the electronic whiteboard, and the weight value at the previous moment; the action space is specifically the adjustment amount of the weight; the reward function is specifically a reward function used to evaluate the effect of taking an action in the current state, which is designed based on the accuracy, stability and user satisfaction of color matching.

[0031] Furthermore, the construction and training of the deep Q network to achieve online weight adjustment includes the following steps:

[0032] C1: Based on the PyTorch deep learning framework, a deep neural network is constructed to approximate and represent the Q function. Specifically, the input of the network is the state vector, and the output is the Q value corresponding to each action. In the network structure, the number of neurons in the input layer is the same as the dimension of the state space, and the number of neurons in the output layer is the same as the size of the action space.

[0033] C2: The electronic whiteboard randomly takes actions in different writing scenarios, and records the state, action reward and next state to form a training data set. The collected experience data is stored in a playback buffer, and small batches of data are randomly extracted from the buffer for training. Then, the stochastic gradient descent optimization algorithm is used to minimize the loss function, and the parameters of the current Q network are copied to a target network at regular intervals.

[0034] C3: Determine the optimal weight adjustment amount based on the current state and the trained deep Q network, update the weight of the color matching model based on the weight adjustment amount, continuously monitor the state changes, and adjust the optimal weight adjustment amount again based on the new state.

[0035] The beneficial effects of the present invention are:

[0036] 1. Dynamic adjustment of handwriting color is achieved through changes in writing speed, angle, pressure and width, bringing users a richer and more vivid writing experience. Users are no longer limited to a single color choice. Users can naturally produce different color effects according to their writing habits and creative needs, which increases the fun and artistry of writing. Different users can write according to their own preferences and styles. The system can adapt to the writing characteristics of different users and achieve personalized color presentation. For example, some users may be accustomed to writing with greater pressure, and the handwriting color will change accordingly to highlight the sense of strength of their writing; while other users may pay more attention to writing speed, and the change of color when writing quickly can reflect their agility of thinking.

[0037] 2. In teaching and demonstration scenarios, teachers and speakers can use color changes to highlight key points, such as increasing the writing pressure to make the handwriting color more eye-catching, or changing the writing speed to attract the audience's attention when explaining key concepts. This method can convey information more effectively and improve the audience's attention and understanding. For students and audiences, this dynamic color change can increase their sense of participation. During the teaching process, students can better understand the knowledge points by observing the changes in handwriting color, and can also actively participate in classroom interactions. In demonstration scenarios, the audience can more intuitively feel the speaker's ideas and key points, improving the effectiveness and attractiveness of the demonstration.

[0038] 3. The system combines fuzzy logic algorithms and machine learning-based color matching algorithms, and dynamically adjusts weights through a deep Q network reinforcement learning algorithm, taking into full account multi-dimensional touch data such as writing speed, angle, pressure, width, and environmental factors. This comprehensive consideration can improve the accuracy of color matching and make the handwriting color more consistent with the actual writing situation. By dynamically adjusting the weights, the system can automatically optimize the color matching strategy according to different application scenarios. For example, in teaching scenarios, more attention is paid to color stability and readability; while in artistic creation scenarios, more attention may be paid to color diversity and innovation. The system can be adaptively adjusted according to the needs of the scene to improve its applicability in different scenarios.

[0039] 4. When users write on the electronic whiteboard, the color of the handwriting can be adjusted in real time according to the changes in touch data, and users can immediately see the effect of their writing. This instant feedback can help users better grasp the strength, speed and angle of writing, etc., and improve the quality and efficiency of writing. The real-time visible writing effect provides users with a smooth interactive experience. Users can adjust their writing methods at any time during the writing process and observe the changes in color to better express their ideas and creativity. At the same time, this real-time nature also makes electronic whiteboards more practical and efficient in scenarios such as teaching, meetings and presentations.

[0040] 5. The system integrates a variety of advanced technologies, including sensor technology, fuzzy logic algorithms, machine learning algorithms, and reinforcement learning algorithms. With the continuous development of technology, the system can be further optimized and expanded. It can combine more sensor data, such as the tilt angle and rotation angle of the stylus, to achieve richer color change effects. At the same time, it can also be deeply integrated with other teaching software and tools to provide users with more intelligent teaching and demonstration solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a module schematic diagram of a multi-dimensional touch data color matching system of an electronic whiteboard.

[0042] Figure 2 It is a method flow chart of the data analysis module.

[0043] Figure 3 It is a flow chart of the method for dynamically adjusting modules. DETAILED DESCRIPTION

[0044] See also Figures 1 to 3 As shown, the present invention relates to a multi-dimensional touch data color matching system for an electronic whiteboard, comprising: a data acquisition module, a data analysis module, a dynamic adjustment module and an application display module connected in communication, wherein:

[0045] The data acquisition module is used to obtain data on writing speed, angle, pressure and width through sensors;

[0046] The data analysis module is used to determine the color value through comprehensive weights based on the real-time acquired data and the color matching model;

[0047] The dynamic adjustment module is used to dynamically calculate the current comprehensive weight value based on the real-time acquired data;

[0048] The application display module is used to apply the calculated color value to the handwriting display.

[0049] Specifically, in the application display module, touch data from the sensor is continuously received, including information such as writing speed, angle, pressure and width, as well as light intensity in environmental data, etc., and RGB color values ​​are dynamically calculated based on real-time touch data through a deep Q network reinforcement learning algorithm or a color matching model calculation method. The calculated RGB color value and the corresponding touch position information (such as coordinates) are stored in a data buffer area, and the content in the data buffer area is updated in time whenever new touch data and color value calculations are completed. The electronic whiteboard software obtains the touch position information and RGB color value in the data buffer area, and determines the starting point of the handwriting and the coordinates of the current point according to the touch position information. Use the graphics drawing function in the software to draw a line segment in the display area of ​​the electronic whiteboard based on the coordinates of the current point, with the determined starting point as the starting point. Set the color of the line segment to the calculated RGB color value, and use the color setting function provided by the graphics library, such as in the graphics drawing library, use a function similar to "setColor(r,g,b)" to set the color, where r, g, and b are the values ​​of the red, green, and blue channels, respectively. As the user writes on the electronic whiteboard, the above drawing process is repeated continuously to update the display of the handwriting in real time. Ensure that the drawing frame rate is high enough to achieve real-time visibility of the writing effect. By optimizing the algorithm and graphics drawing process, unnecessary calculations and delays are reduced, and the drawing efficiency is improved. In order to make the handwriting display smoother, the drawn line segments are smoothed, such as using linear interpolation or other smoothing algorithms to insert some intermediate points between adjacent touch points to make the handwriting look more coherent and natural.

[0050] Provide visual feedback, such as displaying the current color value at the edge or corner of the electronic whiteboard so that users can understand the color of the handwriting, and provide options to adjust the color display effect, such as brightness or contrast, according to user needs. When the user performs some specific operations, such as pausing writing, selecting different tools or adjusting parameters, the electronic whiteboard software can respond in time and make corresponding adjustments. For example, when the user pauses writing, the real-time display update of the handwriting is stopped until the user resumes writing.

[0051] Furthermore, the method of obtaining data of writing speed, angle, pressure and width through a sensor comprises the following steps:

[0052] A1: Calibrate the sensors that obtain writing speed, angle, pressure, width and environmental data one by one, check the connection between the sensors and the electronic whiteboard system, and initialize the software configuration;

[0053] A2: When the user writes or touches the electronic whiteboard, the sensor monitors the digital and analog events in real time and transmits and obtains the touch data at the same time;

[0054] A3: Filter the collected data to remove noise and normalize the data.

[0055] Specifically, in step A1, for pressure calibration: place an object of known weight in the sensing area of ​​the electronic whiteboard, the system records the pressure value detected by the sensor at this time, compares the actual weight of the object with the pressure value detected by the sensor, and adjusts the sensitivity parameters of the sensor so that the detected pressure value matches the actual weight. Repeatedly use objects of different weights for multiple calibrations to ensure the accuracy of the pressure sensor in different pressure ranges. For angle calibration, use a calibration tool with a specific angle, such as an angle ruler, place it in the sensing area of ​​the electronic whiteboard and place it at different angles. The system records the angle value detected by the sensor and compares it with the actual angle of the calibration tool. By adjusting the parameters of the angle detection algorithm, the angle value detected by the sensor is as close to the actual angle as possible. For speed and width calibration, write tests are performed on the electronic whiteboard at different speeds and widths, record the speed and width values ​​detected by the sensor, compare them with the known standard speed and width, adjust the corresponding sensor parameters and algorithms, and ensure the accuracy of the detection results. For environmental data, it is agreed to test the known environmental data for accuracy comparison. Check whether the connection between the sensor and the electronic whiteboard system is normal, ensure that there is no looseness or poor contact, test whether the power supply of the sensor is stable, and ensure that the sensor can work normally. Configure the sampling frequency, resolution and other parameters of the sensor to adapt to different application scenarios and requirements, conduct preliminary tests, perform simple touch operations on the electronic whiteboard, and check whether the touch data acquisition module can work normally.

[0056] In step A2, the pressure sensor detects the pressure applied by the user when writing; the angle sensor detects the angle between the writing tool and the electronic whiteboard; the speed sensor detects the moving speed of the writing; and the width sensor detects the width of the writing handwriting. The sensor transmits the detected touch data to the processor of the electronic whiteboard system via wired or wireless means to ensure the stability and timeliness of data transmission and avoid data loss or delay.

[0057] In step A3, since the sensor may be disturbed by environmental factors during operation, such as electromagnetic interference, vibration, etc., resulting in noise in the collected data, digital filtering technology is used to remove the noise in the data and improve the quality of the data. Different sensors have different measurement ranges and units. In order to facilitate subsequent data processing and analysis, the touch data needs to be normalized. Map data such as pressure, speed, angle and width to a unified range so that data of different dimensions are comparable. If the amount of touch data collected is large, it will affect the processing speed and storage efficiency of the system. Use the lossless compression algorithm in the data compression algorithm to compress the touch data to reduce the storage space and transmission bandwidth of the data.

[0058] Furthermore, the acquisition of touch data includes: obtaining writing speed by calculating the change in touch point position per unit time; obtaining writing angle by the tangent direction of the touch track; obtaining touch pressure by the pressure value output by the sensor; and obtaining touch width by the touch width value output by the sensor.

[0059] Furthermore, obtaining the writing speed and the writing angle includes: detecting the touch position by electromagnetic induction technology, and optimizing the coordinate accuracy by multiple sampling and averaging methods;

[0060] The writing speed is calculated based on the position change and the time interval, and the obtained speed is smoothed by the moving average method;

[0061] The coordinates of the touch points are continuously recorded to form a touch trajectory. The touch trajectory is expressed as a function according to the curve fitting algorithm. The tangent direction of each point on the touch trajectory is calculated by numerical differentiation. The writing angle is then calculated by the inverse tangent function, and the angle is smoothed using a low-pass filter.

[0062] Specifically, the touch position is detected by electromagnetic induction technology. When the user performs a touch operation on the electronic whiteboard, the sensor starts to detect the position of the touch point. The sensor converts the position of the touch point on the electronic whiteboard plane into a digital signal, which is expressed in the form of two-dimensional coordinates (x, y). The sensor transmits the collected coordinate data to the processor of the electronic whiteboard system, and the processor can update the touch position on the display screen in real time. In order to improve the accuracy of the coordinates, multiple sampling and averaging methods are used, such as collecting the coordinates of the touch point multiple times in a short period of time, and calculating the average value as the final coordinate value. When the user starts to touch the electronic whiteboard, the system records the current time t1. When the user moves the touch point on the electronic whiteboard, the system continues to record the time, using the system's clock function or a special timer to record the time. Within a time interval Δt, the system records the position change of the touch point. For example, at time t2, the position coordinates (x2, y2) of the touch point are recorded, compared with the previous position coordinates (x1, y1), and the position change Δx = x2-x1, Δy = y2-y1 is calculated. To avoid sudden changes in speed, the calculated speed is smoothed. The current speed value is averaged with several previous speed values ​​through the moving average method to obtain a smoother speed value.

[0063] When the user writes on the electronic whiteboard, the system continuously records the position coordinates of the touch point to form a touch trajectory, and the trajectory is represented as a function through a curve fitting algorithm. For each point on the touch trajectory, the tangent direction is calculated, and the slope of the tangent at the point is approximated by numerical differentiation, such as calculating the slope between two adjacent points. The angle θ is calculated according to the tangent slope, and the angle is calculated by the inverse tangent function arctan. For example, if the tangent slope is m, the angle θ = arctan (m). Since the touch trajectory may have noise and discontinuity, the calculated angle will have large fluctuations. The angle is smoothed, and the current angle value is weighted averaged with the previous angle values ​​through a low-pass filter to obtain a smoother angle value. Ensure that the pressure sensor equipped with the electronic whiteboard works normally and is stably connected to the system, and adjust the parameters of the sensor so that the output pressure value is as consistent as possible with the actual pressure. When the user writes on the electronic whiteboard, the pressure sensor detects the pressure of the touch point and converts the pressure value into a digital signal and transmits it to the system processor. The processor updates the pressure indication on the display screen in real time, reflecting the pressure by changing the thickness and color of the handwriting. The width sensor is calibrated so that the output width value is as consistent as possible with the actual width. When the user writes on the electronic whiteboard, the width sensor detects the width of the touch handwriting and converts the width value into a digital signal and transmits it to the system processor. The processor can adjust the display width of the handwriting according to the width value, or use different colors, styles, etc. to represent different width ranges.

[0064] Furthermore, the color matching model includes: a color matching model formed by combining a fuzzy logic algorithm and a color matching algorithm based on machine learning through a distributed framework; and the color value is specifically a GRB color value.

[0065] Furthermore, the method of determining the color value by comprehensive weight according to the data acquired in real time and the color matching model comprises the following steps:

[0066] B1: Build a distributed framework based on Spark core components and Hadoop distributed file system, use HDFS for data storage, and monitor the running status of Spark jobs;

[0067] B2: Fuzzify the touch data of writing speed, angle, pressure and width through fuzzy logic algorithm, use the touch data as input variable of fuzzy logic algorithm, and color value as output variable. Determine fuzzy rules and make inferences based on experience and knowledge domain, and convert the fuzzy set of output variables into specific color value through maximum membership method.

[0068] B3: According to the data processing of the color matching algorithm based on machine learning, the fuzzy logic algorithm and the color matching algorithm based on machine learning are assigned comprehensive weights according to the specific requirements of the application scenario, and the color values ​​of the two algorithms are merged by weighted average according to the comprehensive weights;

[0069] B4: Continuously monitor touch data and environmental changes, regularly evaluate and optimize color matching models, analyze algorithm performance and effects, and identify potential problems and areas for improvement.

[0070] Specifically, in step B1, configure the master node and slave nodes of Spark, store the collected data in HDFS, continuously write real-time data to HDFS, and submit the written Spark application to the Spark cluster for execution. Monitor the running status of the Spark job, including task progress, resource usage, and performance indicators. Perform optimization based on the monitoring results, including adjusting the parallelism, memory allocation, and data partitioning of the job to improve the efficiency and performance of the job.

[0071] In step B2, several fuzzy sets are defined for each input variable (writing speed, angle, pressure, width, etc.), and there may be fuzzy sets such as "very slow", "slow", "medium", "fast", and "very fast" for writing speed. Through the triangle membership function, the value of the input variable is mapped to the corresponding fuzzy set to determine the degree to which each input variable belongs to each fuzzy set. Based on experience and domain knowledge, a series of fuzzy rules are determined to describe the relationship between the input variable and the output color value, such as "if the writing speed is fast and the pressure is high, the handwriting color is bright red". According to the fuzzification results of the input variables and the fuzzy rules in the rule base, the fuzzy set of the output variable (color value) is obtained by reasoning through the Mamdani reasoning method. The fuzzy set of the output variable is converted into a specific color value, and a preliminary color value is obtained by defuzzification methods such as the maximum membership method.

[0072] Furthermore, the processing of data according to the color matching algorithm based on machine learning comprises the following steps:

[0073] B31: Collect touch data samples and corresponding expected color values ​​to form a training data set, and divide the data set into a training set, a validation set, and a test set;

[0074] B32: Extract statistical features from the touch data set, perform feature dimensionality reduction on the statistical features using the principal component analysis method, train the support vector machine model using the training set, and adjust the model parameters using the cross-validation method to minimize the prediction error;

[0075] B33: Evaluate the trained model through the validation set. When new touch data is input, the trained model is used to predict and obtain the corresponding color value.

[0076] Specifically, a large number of touch data samples and corresponding expected color values ​​are collected to form a training data set D = {(X1, Y1), (X2, Y2), ..., (X n , Y n )}, where X i =(v i ,θ i , P i , W i , I i ) is the input feature vector, Y i =(r i , g i , b i ) is the corresponding RGB color value. Calculate statistical features, such as the mean of speed: variance Etc. You can also perform principal component analysis (PCA) for dimensionality reduction. Assuming the original data matrix is ​​X, calculate the covariance matrix Perform eigenvalue decomposition on the covariance matrix and obtain eigenvalues ​​λ1, λ2, …, λ m and the corresponding eigenvectors e1, e2, …, e m , select the eigenvectors corresponding to the first k largest eigenvalues ​​to form the projection matrix P = (e1, e2, ..., e k ), project the original data into the low-dimensional space to obtain a new feature vector Z = XP.

[0077] The goal of the support vector machine is to find a hyperplane w T x+b=0, so that samples of different categories can be correctly separated and the interval is maximized. For the linearly separable case, the optimization goal is subject to you i (w T x i +b)≥1, where y i ∈{-1, 1} represents the category label of the sample. By solving this optimization problem, we can get the optimal parameters w and b. For nonlinear cases, by introducing the kernel function K(x i , x j ) maps the data to a high-dimensional space and then performs linear classification in the high-dimensional space. When there is a new input feature vector X new When , the trained support vector machine model is used for prediction. For the linear support vector machine, the prediction function is f(X new )=sign(w T X new +b); for nonlinear support vector machine, the prediction function is where α i is the Lagrange multiplier.

[0078] Through a large number of experiments in different scenarios, the performance of fuzzy logic algorithm and machine learning-based color matching algorithm is evaluated, and the performance of the algorithm is measured using indicators such as accuracy, recall, and mean square error. The weights are determined based on the evaluation results. If the fuzzy logic algorithm performs more stably in some scenarios, and the machine learning algorithm is more accurate in other scenarios, the weights can be assigned according to the specific situation. Let the weight of the fuzzy logic algorithm be w f , the weight of the color matching algorithm based on machine learning is w m , and w f +w m =1. Let the color value output by the fuzzy logic algorithm be C f =(r f , g f , b f ), the color value output by the color matching algorithm based on machine learning is C m =(r m , gm , b m ), then the fused color value C = (r, g, b), where: r = w f r f +w m r m ; g = w f g f +w m g m ; b = w f b f +w m b m The fused color value is sent as the final output to a display device or application for display.

[0079] Furthermore, the current comprehensive weight value is dynamically calculated based on the data acquired in real time, including: adjusting the weight of the color matching model as a reinforcement learning problem according to the deep Q network reinforcement learning algorithm, constructing and training the deep Q network by defining the state space, action space and reward function, thereby realizing online weight adjustment.

[0080] Furthermore, the state space includes: touch data of the current writing speed, angle, pressure and width, environmental data of the light intensity of the electronic whiteboard, and the weight value at the previous moment; the action space is specifically the adjustment amount of the weight; the reward function is specifically a reward function used to evaluate the effect of taking an action in the current state, which is designed based on the accuracy, stability and user satisfaction of color matching.

[0081] Specifically, the action space is the amount of adjustment of the weight, and several discrete adjustment amounts are determined, such as Δw = {-0.1, 0, 0.1}, which represents weight adjustment actions of reducing by 0.1, keeping unchanged, and increasing by 0.1 respectively. The reward function is used to evaluate the effect of taking a certain action in the current state. The reward function is designed based on factors such as the accuracy and stability of color matching and user satisfaction. If the gap between the current color and the expected color is reduced, or the color change is smoother and more stable, a positive reward can be given; if the color change is too drastic or the gap with the expected color increases, a negative reward can be given.

[0082] Furthermore, the construction and training of the deep Q network to achieve online weight adjustment includes the following steps:

[0083] C1: Based on the PyTorch deep learning framework, a deep neural network is constructed to approximate and represent the Q function. Specifically, the input of the network is the state vector, and the output is the Q value corresponding to each action. In the network structure, the number of neurons in the input layer is the same as the dimension of the state space, and the number of neurons in the output layer is the same as the size of the action space.

[0084] C2: The electronic whiteboard randomly takes actions in different writing scenarios, and records the state, action reward and next state to form a training data set. The collected experience data is stored in a playback buffer, and small batches of data are randomly extracted from the buffer for training. Then, the stochastic gradient descent optimization algorithm is used to minimize the loss function, and the parameters of the current Q network are copied to a target network at regular intervals.

[0085] C3: Determine the optimal weight adjustment amount based on the current state and the trained deep Q network, update the weight of the color matching model based on the weight adjustment amount, continuously monitor the state changes, and adjust the optimal weight adjustment amount again based on the new state.

[0086] Specifically, the output is the Q value corresponding to each action, which represents the expected long-term reward of taking the action in the current state. The network structure may include an input layer, several hidden layers, and an output layer. In step C2, the experience replay mechanism is used to break the correlation between data. Use optimization algorithms such as stochastic gradient descent to minimize the loss function: where θ is the parameter of the Q network, s and a are the current state and action, r is the current reward, γ is the discount factor, s′ and a′ are the next state and action, and θ - are the parameters of the target network.

[0087] In order to improve the stability of training, the parameters of the current Q network are copied to a target network at regular intervals to calculate the target Q value. In the actual writing process, according to the current state, the trained deep Q network is used to select the optimal action, that is, the weight adjustment amount, and then the weights of the fuzzy logic algorithm and the color matching algorithm based on machine learning are updated according to the selected action. The state changes are continuously monitored, and the action is selected again according to the new state for weight adjustment, so as to realize dynamic weight adjustment and real-time update of handwriting color.

[0088] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A multi-dimensional touch data color matching system for an electronic whiteboard, characterized in that: include: The data acquisition module, data analysis module, dynamic adjustment module and application display module are connected in communication, wherein: The data acquisition module is used to obtain data on writing speed, angle, pressure and width through sensors; The data analysis module is used to determine the color value through comprehensive weights based on the real-time acquired data and the color matching model; The dynamic adjustment module is used to dynamically calculate the current comprehensive weight value based on the real-time acquired data; The application display module is used to apply the calculated color value to the handwriting display.

2. The multi-dimensional touch data color matching system of the electronic whiteboard according to claim 1, characterized in that: The method of obtaining data of writing speed, angle, pressure and width through a sensor comprises the following steps: A1: Calibrate the sensors that obtain writing speed, angle, pressure, width and environmental data one by one, check the connection between the sensors and the electronic whiteboard system, and initialize the software configuration; A2: When the user writes or touches the electronic whiteboard, the sensor monitors the digital and analog events in real time and transmits and obtains the touch data at the same time; A3: Filter the collected data to remove noise and normalize the data.

3. The multi-dimensional touch data color matching system of the electronic whiteboard according to claim 2, characterized in that: The acquisition of touch data includes: obtaining writing speed by calculating the change of touch point position per unit time; obtaining writing angle by the tangent direction of the touch track; obtaining touch pressure by the pressure value output by the sensor; and obtaining touch width by the touch width value output by the sensor.

4. The multi-dimensional touch data color matching system of the electronic whiteboard according to claim 3, characterized in that: The obtaining of the writing speed and the writing angle includes: detecting the touch position by electromagnetic induction technology, and optimizing the coordinate accuracy by multiple sampling and averaging methods; The writing speed is calculated based on the position change and the time interval, and the obtained speed is smoothed by the moving average method; The coordinates of the touch points are continuously recorded to form a touch trajectory. The touch trajectory is expressed as a function according to the curve fitting algorithm. The tangent direction of each point on the touch trajectory is calculated by numerical differentiation. The writing angle is then calculated by the inverse tangent function, and the angle is smoothed using a low-pass filter.

5. The multi-dimensional touch data color matching system of the electronic whiteboard according to claim 1, characterized in that: The color matching model includes: a color matching model formed by combining a fuzzy logic algorithm and a color matching algorithm based on machine learning through a distributed framework; the color value is specifically a GRB color value.

6. The multi-dimensional touch data color matching system of the electronic whiteboard according to claim 5, characterized in that: Determining the color value by comprehensive weight according to the data acquired in real time and the color matching model comprises the following steps: B1: Build a distributed framework based on Spark core components and Hadoop distributed file system, use HDFS for data storage, and monitor the running status of Spark jobs; B2: Fuzzify the touch data of writing speed, angle, pressure and width through fuzzy logic algorithm, use the touch data as input variable of fuzzy logic algorithm, and color value as output variable. Determine fuzzy rules and make inferences based on experience and knowledge domain, and convert the fuzzy set of output variables into specific color value through maximum membership method. B3: According to the data processing of the color matching algorithm based on machine learning, the fuzzy logic algorithm and the color matching algorithm based on machine learning are assigned comprehensive weights according to the specific requirements of the application scenario, and the color values ​​of the two algorithms are merged by weighted average according to the comprehensive weights; B4: Continuously monitor touch data and environmental changes, regularly evaluate and optimize color matching models, analyze algorithm performance and effects, and identify potential problems and areas for improvement.

7. The multi-dimensional touch data color matching system of the electronic whiteboard according to claim 6, characterized in that: The processing of data according to the color matching algorithm based on machine learning comprises the following steps: B31: Collect touch data samples and corresponding expected color values ​​to form a training data set, and divide the data set into a training set, a validation set, and a test set; B32: Extract statistical features from the touch data set, perform feature dimensionality reduction on the statistical features using the principal component analysis method, train the support vector machine model using the training set, and adjust the model parameters using the cross-validation method to minimize the prediction error; B33: Evaluate the trained model through the validation set. When new touch data is input, the trained model is used to predict and obtain the corresponding color value.

8. The multi-dimensional touch data color matching system of an electronic whiteboard according to claim 1, characterized in that: The method dynamically calculates the current comprehensive weight value based on the data obtained in real time, including: taking the weight adjustment of the color matching model as a reinforcement learning problem according to the deep Q network reinforcement learning algorithm, constructing and training the deep Q network by defining the state space, action space and reward function, thereby realizing online weight adjustment.

9. The multi-dimensional touch data color matching system of the electronic whiteboard according to claim 8, characterized in that: The state space includes: touch data of the current writing speed, angle, pressure and width, environmental data of the light intensity of the electronic whiteboard, and the weight value at the previous moment; the action space is specifically the adjustment amount of the weight; the reward function is specifically a reward function used to evaluate the effect of taking an action in the current state, which is designed based on the accuracy, stability and user satisfaction of color matching.

10. The multi-dimensional touch data color matching system of an electronic whiteboard according to claim 8, characterized in that: The method of constructing and training a deep Q network to achieve online weight adjustment includes the following steps: C1: Based on the PyTorch deep learning framework, a deep neural network is constructed to approximate and represent the Q function. Specifically, the input of the network is the state vector, and the output is the Q value corresponding to each action. In the network structure, the number of neurons in the input layer is the same as the dimension of the state space, and the number of neurons in the output layer is the same as the size of the action space. C2: The electronic whiteboard randomly takes actions in different writing scenarios, and records the state, action reward and next state to form a training data set. The collected experience data is stored in a playback buffer, and small batches of data are randomly extracted from the buffer for training. Then, the stochastic gradient descent optimization algorithm is used to minimize the loss function, and the parameters of the current Q network are copied to a target network at regular intervals. C3: Determine the optimal weight adjustment amount based on the current state and the trained deep Q network, update the weight of the color matching model based on the weight adjustment amount, continuously monitor the state changes, and adjust the optimal weight adjustment amount again based on the new state.

Images