Dynamic drawing and caching method for electronic whiteboard large-screen equipment

By setting up temporary drawing layers and cache layers in electronic whiteboard devices, and using convolutional neural network to analyze memory pressure and dynamically adjust memory allocation, the performance bottleneck problem of traditional electronic whiteboard devices when processing large-scale handwriting data is solved, and the handwriting is realized The timely processing and display of data improves the user experience.

CN118915992APending Publication Date: 2024-11-08GUANGZHOU LANGO ELECTRONICS TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411084324.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Traditional electronic whiteboard devices are prone to performance bottlenecks when processing large-scale handwriting data, resulting in insufficient smooth handwriting display and even obvious lag, affecting user experience and application effects.

Method used

By setting up temporary drawing layers and cache layers in memory, dynamically adjusting the memory space according to user needs, using temporary drawing layers to receive and render handwriting data in real time, clearing temporary drawing layer data based on user hand-raising status, and analyzing memory with convolutional neural network Pressure, dynamically adjust memory allocation.

Benefits of technology

It effectively reduces the delay in handwriting display, improves system stability and response speed, and ensures the smoothness and efficiency of the user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915992A_ABST
    Figure CN118915992A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of electronic whiteboards, in particular to a dynamic drawing and caching method for large-screen equipment of an electronic whiteboard, which comprises the following steps of: dividing a temporary drawing layer and a caching layer in a memory space according to the resolution ratio and the use condition of a writing area, monitoring the memory occupation in real time, and dynamically adjusting the memory space based on the occupation rate and the memory pressure. And the temporary drawing layer receives user handwriting data, feeds back the user handwriting data in real time, and performs real-time rendering and display after capturing the user handwriting data. And based on the hand raising state of the user, the cache layer receives and empties the data of the temporary drawing layer, so that memory allocation equalization is realized, and the space occupancy rate and the display delay are reduced. The handwriting data of the cache layer generates a data set, and the memory space is dynamically adjusted in combination with the memory pressure information, so that the response speed and the memory use efficiency are improved. By drawing scenes in a classified manner, performing task distinguishing based on a classification result, judging accidental demands and conventional demands, removing overpressure timestamps of the accidental demands, optimizing a memory adjustment strategy, and ensuring system stability and high efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of electronic whiteboards, and in particular to a dynamic drawing and caching method for a large-screen electronic whiteboard device. Background Art

[0002] Large-screen electronic whiteboard devices have been widely used in education, business, and conferences, providing an intuitive and efficient way of interaction. However, with the continuous improvement of user needs, electronic whiteboards face challenges in memory management and performance optimization when handling large-scale drawing tasks. Especially in the real-time drawing process, the dynamic allocation and efficient use of memory become key issues. In traditional electronic whiteboard systems, the user's handwriting data is usually captured through a capacitive screen and fed back to the display screen in real time. However, as the user's writing speed increases and the drawing content becomes more complicated, the system is prone to performance bottlenecks when processing large-scale data, resulting in the handwriting display not being smooth enough or even obvious lag. This lag not only affects the user's writing experience, but also reduces the application effect of electronic whiteboards in efficient teaching and business meetings.

[0003] Traditional electronic whiteboard devices usually use a memory allocation method that shares writing and read-write storage. They are unable to dynamically adjust the memory space according to actual usage, resulting in insufficient memory during use. During the capture, processing and display of handwriting data, due to the large amount of data and the high memory space occupancy rate, delays are prone to occur, resulting in a decrease in system processing speed, which in turn causes delays in the display feedback of the handwriting. Summary of the invention

[0004] To solve the above problems, the present invention provides a dynamic drawing and caching method for a large-screen electronic whiteboard device. By setting up a temporary drawing layer and a cache layer in the memory room and dynamically adjusting the memory according to user needs, the handwriting delay is effectively reduced and the system stability is enhanced.

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

[0006] A dynamic drawing and caching method for an electronic whiteboard large-screen device comprises the following steps:

[0007] Initialize memory allocation of temporary drawing layer and cache layer in the memory space of the electronic whiteboard;

[0008] Receive the user's handwriting data through the temporary drawing layer and provide real-time feedback, and receive the handwriting data of the temporary drawing layer through the cache layer based on the user's hand-raising state and clear the handwriting data of the temporary drawing layer;

[0009] The handwriting data of the storage cache layer generates a data set, and the memory space of the temporary drawing layer and the cache layer is dynamically adjusted according to the data set and memory pressure.

[0010] Further, the initializing of the temporary drawing layer and the cache layer in the memory space of the electronic whiteboard includes:

[0011] Get the resolution setting of the current electronic whiteboard writing area, and calculate the memory space required for the writing area based on the resolution;

[0012] Allocate memory space for temporary drawing layers and cache layers in the memory space required by the writing area;

[0013] Monitor the memory usage of temporary drawing layers and cache layers, obtain the occupancy rate and calculate the memory pressure.

[0014] Furthermore, the receiving of the user's handwriting data through the temporary drawing layer and providing real-time feedback includes:

[0015] The user's handwriting data is captured by the capacitive screen of the electronic whiteboard;

[0016] The user handwriting data is temporarily stored by using the memory occupation of the temporary drawing layer, and the user handwriting data is rendered in real time using the GPU of the temporary drawing layer and the signal is output to the electronic whiteboard for display.

[0017] Furthermore, the receiving of handwriting data of the temporary drawing layer through the cache layer based on the hand raising state of the user and clearing the handwriting data of the temporary drawing layer includes:

[0018] Determine the user's hand-raising state according to the pause time obtained from the user's handwriting data and the preset continuous time, wherein the user's hand-raising state includes the hand-raising state and the hand-not-raising state;

[0019] If the user's hand-raising state is "raised", the handwriting data of the temporary drawing layer is obtained through the cache layer and the handwriting data of the temporary drawing layer is cleared;

[0020] If the user's hand-raising state is not raised, the user's handwriting data continues to be received through the temporary drawing layer.

[0021] Furthermore, the generation of a data set from the handwriting data of the storage cache layer comprises the following steps:

[0022] Convert handwriting data and its timestamp into handwriting vector;

[0023] The handwriting vectors are sorted and stored according to timestamps to construct a data set.

[0024] Furthermore, dynamically adjusting the memory space of the temporary drawing layer and the cache layer according to the data set and the memory pressure includes the following steps:

[0025] Classify the drawing scenes and distinguish the tasks based on the classification results;

[0026] The timestamps when the memory pressure of the temporary drawing layer and the cache layer in a single task exceeds the preset pressure threshold are extracted and marked as overpressure timestamps, the total overpressure time is calculated according to the overpressure timestamps, and the first overpressure frequency is calculated according to the total overpressure time and the total time of the single task;

[0027] Triggering handwriting purpose determination according to the first overpressure frequency, the handwriting purpose including accidental needs and regular needs;

[0028] Remove the overpressure timestamps of handwritings for occasional needs and calculate the second overpressure frequency;

[0029] The memory space of the temporary drawing layer and the cache layer is adjusted according to the second overvoltage frequency.

[0030] Furthermore, the classifying of the drawing scenes and distinguishing the tasks based on the classification results includes:

[0031] Classifying the drawing scenes according to scene categories, wherein the scene categories include speech annotation, graphic drawing, and text input;

[0032] Based on the determined drawing scenarios, tasks are differentiated according to different documents.

[0033] Further, the step of acquiring the handwriting vector and the drawing scene corresponding to the overpressure timestamp in the data set according to the first overpressure frequency trigger includes:

[0034] If the first overpressure frequency is greater than the preset pressure frequency, the handwriting purpose determination is triggered;

[0035] If the first overpressure frequency is less than or equal to the preset pressure frequency, the handwriting purpose determination is not triggered.

[0036] Furthermore, the handwriting purpose determination includes obtaining the handwriting vector and drawing scene corresponding to the overpressure timestamp in the data set, and performing classification determination based on a convolutional neural network to determine the handwriting purpose.

[0037] Furthermore, the formula of the convolutional neural network is as follows:

[0038]

[0039] Among them, Output is the final classification result of the convolutional neural network, including accidental demand and regular demand; W i is the weight of the i-th convolution kernel; x i is the i-th element of the input data, including the handwriting vector, overpressure timestamp and drawing scene; ReLU(z) is the rectified linear unit activation function, defined as ReLU(z)=max(0,z); Softmax(z) is the activation function.

[0040] Further, adjusting the memory space of the temporary drawing layer and the cache layer according to the second overvoltage frequency includes:

[0041] If the second overpressure frequency is greater than the preset pressure frequency, the memory space of the temporary drawing layer and the cache layer is increased;

[0042] If the second overpressure frequency is less than or equal to the preset pressure frequency, the memory space of the temporary drawing layer and the cache layer is kept unchanged.

[0043] Furthermore, the formula for increasing the memory space of the temporary drawing layer and the cache layer is as follows:

[0044]

[0045] Among them, M new is the memory size after adjustment; M current is the current memory space size; frequency is the second overpressure frequency; threshold is the preset pressure frequency; G is the gradient coefficient.

[0046] The beneficial effects of the present invention are as follows: the present invention divides the temporary drawing layer and the cache layer in the memory space according to the resolution and usage of the current electronic whiteboard writing area, initializes the memory allocation of the temporary drawing layer and the cache layer, monitors the memory occupancy in real time, and dynamically adjusts the memory space according to the occupancy rate and memory pressure. The user's handwriting data is received through the temporary drawing layer and real-time feedback is provided to ensure timely processing and display of the handwriting data. After the user's handwriting data is captured, it is rendered in real time through the temporary drawing layer and the signal is output to the electronic whiteboard for display. Based on the user's hand-raising state, the handwriting data of the temporary drawing layer is received through the cache layer and the handwriting data of the temporary drawing layer is cleared, thereby achieving balanced memory allocation processing, effectively reducing the simultaneous occupancy rate of the space, and thus reducing the display feedback delay of the handwriting. The handwriting data of the cache layer is stored to generate a data set, and the memory space of the temporary drawing layer and the cache layer are dynamically adjusted in combination with the memory pressure information to improve the response speed of the system and the memory usage efficiency. By classifying the drawing scenes, distinguishing tasks based on the classification results, and analyzing the user's handwriting data through a convolutional neural network (CNN), it is determined whether it is an occasional demand or a regular demand, and the overpressure timestamps of occasional demands are removed to further optimize memory usage. CNN is introduced to classify and determine handwriting data, and the purpose of the handwriting data is determined by analyzing the handwriting data with overpressure timestamps, and the overpressure timestamps of occasional demands are removed. Memory allocation is recalculated to ensure the stability and efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a step flow chart of a dynamic drawing and caching method of an electronic whiteboard large-screen device in the present invention.

[0048] Figure 2 It is a flow chart of the steps of initializing a temporary drawing layer and a cache layer in the memory space of an electronic whiteboard in the present invention.

[0049] Figure 3 It is a flow chart of the steps of dynamically adjusting the memory space of the temporary drawing layer and the cache layer according to the data set and the memory pressure in the present invention. DETAILED DESCRIPTION

[0050] See also Figure 1-3 As shown, the present invention relates to a dynamic drawing and caching method for an electronic whiteboard large-screen device, comprising the following steps:

[0051] Initialize memory allocation of temporary drawing layer and cache layer in the memory space of the electronic whiteboard;

[0052] Receive the user's handwriting data through the temporary drawing layer and provide real-time feedback, and receive the handwriting data of the temporary drawing layer through the cache layer based on the user's hand-raising state and clear the handwriting data of the temporary drawing layer;

[0053] The handwriting data of the storage cache layer generates a data set, and the memory space of the temporary drawing layer and the cache layer is dynamically adjusted according to the data set and memory pressure.

[0054] In some embodiments, first, according to the writing area resolution of the current electronic whiteboard, the memory space required for the writing area is calculated. By analyzing the size and resolution of the writing area, the initial memory allocation of the temporary drawing layer and the cache layer is determined. In actual use, the user's handwriting data is received through the temporary drawing layer and real-time feedback is given. When the user writes on the electronic whiteboard, the capacitive screen captures the user's handwriting data and transmits it to the temporary drawing layer. The temporary drawing layer uses the GPU for fast rendering to display the handwriting data on the electronic whiteboard in real time. In this way, the user's writing experience is smooth and there is no obvious lag. Based on the user's hand-raising state, the handwriting data of the temporary drawing layer is received through the cache layer and the handwriting data of the temporary drawing layer is cleared. In order to achieve this function, the present invention designs a mechanism for detecting the user's hand-raising state. When the user stops writing, the pause time of the handwriting data and the preset continuous time are analyzed to determine whether the user raises his hand. If the user raises his hand, the handwriting data in the temporary drawing layer is transferred to the cache layer, and the data of the temporary drawing layer is cleared to prepare for receiving new handwriting data. The handwriting data of the cache layer is stored to generate a data set, and the memory space of the temporary drawing layer and the cache layer is dynamically adjusted according to the data set and memory pressure. The handwriting data stored in the cache layer will be regularly converted into handwriting vectors and sorted according to timestamps to build a data set. By analyzing the distribution of handwriting data in the data set and memory pressure information, the memory allocation strategy of the temporary drawing layer and the cache layer is dynamically adjusted. Specifically, if the memory pressure is high, part of the memory of the temporary drawing layer is released first to transfer more data to the cache layer; if the memory pressure is low, the memory space of the temporary drawing layer is increased to improve the real-time performance of drawing and feedback.

[0055] Further, the initializing of the temporary drawing layer and the cache layer in the memory space of the electronic whiteboard includes:

[0056] Get the resolution setting of the current electronic whiteboard writing area, and calculate the memory space required for the writing area based on the resolution;

[0057] Allocate memory space for temporary drawing layers and cache layers in the memory space required by the writing area;

[0058] Monitor the memory usage of temporary drawing layers and cache layers, obtain the occupancy rate and calculate the memory pressure.

[0059] It should be noted that, first, the resolution setting of the current electronic whiteboard writing area is obtained. The resolution setting of the electronic whiteboard determines the number of pixels to be processed per unit area, which is crucial for the allocation of memory space. By calling the system interface of the electronic whiteboard device, the resolution information of the current writing area is obtained, such as 1920x1080 pixels or a higher 4K resolution (3840x2160 pixels). The size of the resolution directly affects the memory space required for the temporary drawing layer and the cache layer. Next, memory space is allocated for the temporary drawing layer and the cache layer in the memory space required for the writing area. In the specific implementation process, first, according to the calculated total memory requirement, a part of the memory is reserved for the temporary drawing layer, and the other part is used for the cache layer. For example, in the above-mentioned 8MB total memory requirement, 5MB can be allocated to the temporary drawing layer and 3MB to the cache layer. Such an allocation ratio can be adjusted according to specific application requirements and user writing habits to achieve the best performance. Memory pressure can be calculated by analyzing the trend of memory occupancy over a period of time. If the memory occupancy rate continues to be higher than a certain threshold (for example, 80%), it is considered that the memory pressure is large and the memory space needs to be reallocated or optimized. The specific memory pressure calculation method can be combined with historical data and prediction models, and further optimized through machine learning algorithms to dynamically adjust the memory allocation of temporary drawing layers and cache layers to ensure that the system can maintain a smooth user experience under high load conditions.

[0060] Furthermore, the receiving of the user's handwriting data through the temporary drawing layer and providing real-time feedback includes:

[0061] The user's handwriting data is captured by the capacitive screen of the electronic whiteboard;

[0062] The user handwriting data is temporarily stored by using the memory occupation of the temporary drawing layer, and the user handwriting data is rendered in real time using the GPU of the temporary drawing layer and the signal is output to the electronic whiteboard for display.

[0063] In some embodiments, first, the user's handwriting data is captured through the capacitive screen of the electronic whiteboard. When the user writes on the electronic whiteboard, the capacitive screen can accurately capture the movement trajectory of the user's finger or stylus. These trajectory data include the starting point, end point and all coordinate points of the intermediate path of the handwriting. These data are converted into digital signals by the capacitive screen sensor and transmitted to the processor of the electronic whiteboard in real time. After the processor receives these handwriting data, the user's handwriting data is temporarily stored through the memory occupancy of the temporary drawing layer. The temporary drawing layer pre-allocates a specific memory area in the system memory for storing the handwriting data captured in real time. Every time the user writes, new handwriting data will be continuously stored in this memory area. The design of the memory area takes into account efficient data writing and reading to ensure that the system can quickly respond to the user's writing operation. Next, the GPU of the temporary drawing layer is used to render the user's handwriting data in real time. GPU, that is, graphics processing unit, is good at processing large-scale parallel computing tasks, especially the rendering of graphics and images. By calling the computing resources of the GPU, the system can quickly convert the handwriting data temporarily stored in the memory into a graphics signal. This process includes vectorizing the handwriting data, calculating the color value of each pixel, and generating the corresponding display signal. The generated graphic signal will be immediately output to the electronic whiteboard for display. After receiving the signal from the GPU, the display screen of the electronic whiteboard presents the user's handwriting in a high-resolution and high-precision manner. This display process is almost instantaneous, and users can see their handwriting appear on the electronic whiteboard while writing, providing a smooth and real-time writing experience. For example, when a user writes on an electronic whiteboard with a resolution of 1920x1080, the capacitive screen will capture the precise coordinates of each touch point (such as x=500, y=300), and these coordinates will be stored in the memory of the temporary drawing layer. The GPU will read these coordinate data and perform rendering processing to connect these points into continuous handwriting lines. Finally, the rendered graphic signal will be transmitted to the display screen, allowing users to see their writing in real time.

[0064] Furthermore, the receiving of handwriting data of the temporary drawing layer through the cache layer based on the hand raising state of the user and clearing the handwriting data of the temporary drawing layer includes:

[0065] Determine the user's hand-raising state according to the pause time obtained from the user's handwriting data and the preset continuous time, wherein the user's hand-raising state includes the hand-raising state and the hand-not-raising state;

[0066] If the user's hand-raising state is "raised", the handwriting data of the temporary drawing layer is obtained through the cache layer and the handwriting data of the temporary drawing layer is cleared;

[0067] If the user's hand-raising state is not raised, the user's handwriting data continues to be received through the temporary drawing layer.

[0068] In some embodiments, first, the user's hand-raising state is determined according to the pause time obtained by the user's handwriting data and the preset continuous time. The system records the timestamp of each handwriting data point by monitoring the user's writing behavior in real time. When the user stops writing, that is, when no new handwriting data is input, the system starts timing. If the handwriting pause time exceeds the preset continuous time (for example, 500 milliseconds), the system determines that the user has raised his hand. The user's hand-raising state can be divided into two situations: raised hand and not raised hand. When the user's hand-raising state is determined to be raised hand, the handwriting data of the temporary drawing layer is obtained through the cache layer and the handwriting data of the temporary drawing layer is cleared. Specifically, the system transfers all recorded handwriting data in the temporary drawing layer to the cache layer. This step ensures that the temporary drawing layer has enough space to receive new handwriting data, thereby avoiding data loss or processing delays caused by insufficient memory. At the same time, the memory area of ​​the temporary drawing layer is cleared to prepare for the next writing. For example, when the user finishes writing a word on the electronic whiteboard and raises his hand, the system transfers the handwriting data of the word to the cache layer and clears the temporary drawing layer. If the user's hand-raising state is not raised, the user's handwriting data will continue to be received through the temporary drawing layer. In this case, the system will maintain the current state of the temporary drawing layer and continue to receive and process new handwriting data. The user's writing data will be transmitted to the temporary drawing layer in real time and displayed in real time on the electronic whiteboard. In this way, the system can continuously and seamlessly process the user's writing movements to ensure the fluency and real-time nature of writing. For example, when the user is writing continuously, the system will continuously monitor the handwriting input data. If the user does not pause significantly during the writing process, the system will maintain the state of not raising the hand, and the temporary drawing layer will continue to receive and process the handwriting data. However, if the user pauses after writing a paragraph, the system detects that the pause time exceeds the preset threshold, then it will determine that the user has raised his hand, and transfer the handwriting data of the paragraph from the temporary drawing layer to the cache layer, and clear the temporary drawing layer at the same time.

[0069] Furthermore, the generation of a data set from the handwriting data of the storage cache layer comprises the following steps:

[0070] Convert handwriting data and its timestamp into handwriting vector;

[0071] The handwriting vectors are sorted and stored according to timestamps to construct a data set.

[0072] In some embodiments, first, the handwriting data and its timestamps are converted into handwriting vectors. During the user's writing process, each handwriting data point is recorded, and these data points usually include position coordinates (such as x, y coordinates) and timestamps. The timestamp records the specific time when each handwriting point is captured, in milliseconds. To facilitate subsequent data processing and analysis, these raw data points need to be converted into handwriting vectors. A handwriting vector is a multi-dimensional data structure, and each vector contains all the information of a handwriting point, including its position coordinates and the corresponding timestamp. Specifically, the system reads the handwriting data in the cache layer and processes each data point. Assuming the data of a handwriting point is (x1, y1, t1), where x1 and y1 are position coordinates and t1 is the timestamp, the corresponding handwriting vector is V1 = [x1, y1, t1]. In this way, the system converts all the handwriting data points into corresponding handwriting vectors, forming a set of handwriting vectors. Next, the handwriting vectors are sorted by timestamps and stored to build a data set. Since the user's writing process is a continuous time-series event, the handwriting vectors need to be sorted by timestamps to maintain the coherence and temporal information of the writing. The sorting process can use common sorting algorithms, such as quicksort or mergesort, to ensure that the handwriting vectors are arranged in ascending order of timestamps. After sorting, the system stores the sorted handwriting vectors in a data structure, usually a list or an array, to build a complete data set. This data set not only contains all the handwriting vectors but also retains the time order information during the writing process, facilitating subsequent data analysis and processing. For example, after the user finishes writing a text, the system processes the handwriting data in the cache layer. Assuming the following handwriting points are generated during the user's writing: [(x1, y1, t1), (x2, y2, t2), (x3, y3, t3)], where t1 < t2 < t3. The system converts these data points into handwriting vectors and sorts them by timestamps to obtain the sorted set of handwriting vectors: V = {[x1, y1, t1], [x2, y2, t2], [x3, y3, t3]}. Finally, these handwriting vectors will be stored in the data set to form a complete record.

[0073] Further, the dynamically adjusting the memory spaces of the temporary drawing layer and the cache layer according to the data set and the memory pressure includes the following steps:

[0074] Classify the drawing scenarios and distinguish tasks based on the classification results;

[0075] Extract the timestamps when the memory pressures of the temporary drawing layer and the cache layer exceed the preset pressure threshold in a single task, and mark them as overpressure timestamps. Calculate the total overpressure time through the overpressure timestamps, and calculate the first overpressure frequency through the total overpressure time and the total time of the single task;

[0076] Triggering handwriting purpose determination according to the first overpressure frequency, the handwriting purpose including accidental needs and regular needs;

[0077] Remove the overpressure timestamps of handwritings for occasional needs and calculate the second overpressure frequency;

[0078] The memory space of the temporary drawing layer and the cache layer is adjusted according to the second overvoltage frequency.

[0079] In some embodiments, first, the drawing scenes are classified, and tasks are distinguished based on the classification results. The system classifies according to different drawing scenes, such as speech annotation, graphic drawing, text input, etc. This classification process can be performed by machine learning algorithms, such as K-means clustering or decision tree classification. By classifying different scenes, the system can identify the specific memory requirements and performance requirements of each task, so as to make more accurate memory allocation. Next, the timestamps of the memory pressure of the temporary drawing layer and the cache layer exceeding the preset pressure threshold in a single task are extracted and marked as overpressure timestamps. During operation, the system monitors the memory usage of the temporary drawing layer and the cache layer in real time. Once it is detected that the memory pressure exceeds the preset threshold (for example, 80%), the system records the current timestamp and marks it as an overpressure timestamp. Memory pressure can be measured by memory occupancy. Through these overpressure timestamps, the system calculates the total overpressure time, that is, the accumulation of all overpressure times. Then, the first overpressure frequency is calculated by the ratio of the total overpressure time to the total time of a single task. The first overpressure frequency reflects the frequency of excessive memory pressure during task execution. According to the first overpressure frequency, the system triggers the handwriting purpose determination. The handwriting purpose includes two types: accidental demand and regular demand. The system uses convolutional neural network (CNN) to analyze the handwriting data in the data set to determine whether the current handwriting is an accidental demand. If it is determined to be an accidental demand, the system will remove the corresponding overpressure timestamp from the data set to ensure more efficient memory allocation. After removing the overpressure timestamp of the accidental demand, the system recalculates the second overpressure frequency. The second overpressure frequency reflects the memory pressure after removing the accidental demand. According to the second overpressure frequency, the system dynamically adjusts the memory space of the temporary drawing layer and the cache layer. If the second overpressure frequency is still high, the system will increase the memory allocation of the temporary drawing layer and the cache layer to ensure that the system can still maintain stability and efficiency under high load. For example, if the second overpressure frequency exceeds 80%, the system can increase the memory of the temporary drawing layer by 20% and the memory of the cache layer by 10% to relieve memory pressure.

[0080] Furthermore, the classifying of the drawing scenes and distinguishing the tasks based on the classification results includes:

[0081] Classifying the drawing scenes according to scene categories, wherein the scene categories include speech annotation, graphic drawing, and text input;

[0082] Based on the determined drawing scenarios, tasks are differentiated according to different documents.

[0083] In some embodiments, first, the drawing scenes are classified according to the scene categories. In actual applications, electronic whiteboards may involve a variety of different usage scenarios, such as speech annotation, graphic drawing, and text input. Each scenario has different requirements for system resources and performance requirements. In order to optimize memory usage and improve system response speed, these scenarios need to be accurately classified. Speech annotation: During a speech, users usually use an electronic whiteboard for real-time annotation, including drawing circles, drawing lines, and writing short notes. This type of operation requires the system to respond quickly and occupy relatively little memory. Graphic drawing: Graphic drawing involves complex graphic operations, such as drawing, inserting graphics, editing graphics, etc. This type of operation requires the system to provide higher computing power and memory space to process complex graphic data. Text input: When a user enters text on an electronic whiteboard, the system needs to be able to accurately recognize and display the user's writing. The handwriting data in the text input scenario is relatively dense, and the requirements for memory and processing speed are relatively high.

[0084] The system uses a pre-trained machine learning model, convolutional neural network (CNN) and support vector machine (SVM) algorithms to monitor and analyze the user's writing behavior in real time. The model input includes the morphological characteristics of the handwriting data, writing speed, handwriting trajectory, etc. Based on these characteristics, the system can accurately classify the user's current drawing scene.

[0085] Further, the step of acquiring the handwriting vector and the drawing scene corresponding to the overpressure timestamp in the data set according to the first overpressure frequency trigger includes:

[0086] If the first overpressure frequency is greater than the preset pressure frequency, the handwriting purpose determination is triggered;

[0087] If the first overpressure frequency is less than or equal to the preset pressure frequency, the handwriting purpose determination is not triggered.

[0088] In some embodiments, if the first overpressure frequency is greater than the preset pressure frequency: for example, the preset pressure frequency threshold is 80%, if the calculated first overpressure frequency exceeds 80%, the system believes that there is a large memory pressure in the current task. At this time, the system triggers the handwriting purpose determination. Specifically, the system extracts the handwriting vector and drawing scene corresponding to the overpressure timestamp from the data set, and uses the convolutional neural network (CNN) to analyze these data to determine whether the handwriting data is an accidental demand or a regular demand. Accidental demand is usually a user's concentrated writing or drawing in a short period of time, while regular demand is the user's continuous writing or drawing behavior. Through this judgment, the system can identify which handwriting data needs to be processed first and which can be postponed or optimized. If the first overpressure frequency is less than or equal to the preset pressure frequency: if the calculated first overpressure frequency does not exceed the preset 80%, the system believes that the memory pressure of the current task is within an acceptable range, and the handwriting purpose determination is not triggered at this time. The system continues to process the handwriting data in a conventional manner to ensure the efficiency and rationality of resource allocation.

[0089] Furthermore, the handwriting purpose determination includes obtaining the handwriting vector and drawing scene corresponding to the overpressure timestamp in the data set, and performing classification determination based on a convolutional neural network to determine the handwriting purpose.

[0090] Furthermore, the formula of the convolutional neural network is as follows:

[0091]

[0092] Among them, Output is the final classification result of the convolutional neural network, including accidental demand and regular demand; W i is the weight of the i-th convolution kernel; x i is the i-th element of the input data, including the handwriting vector, overpressure timestamp and drawing scene; ReLU(z) is the rectified linear unit activation function, defined as ReLU(z)=max(0,z); Softmax(z) is the activation function.

[0093] Specifically, the convolutional neural network performs a series of convolution operations on the input handwriting vector and the drawing scene to extract the local features of the handwriting data. Through multiple layers of convolution and pooling, CNN is able to extract higher-level feature representations layer by layer. The handwriting vector includes the coordinate points and their timestamps that record the user's handwriting, while the drawing scene describes the specific application scenario of the handwriting (such as speech annotation, graphic drawing, text input, etc.). This information constitutes the input data of the convolutional neural network. Multiple convolutional filters are used to perform convolution operations on the input data. The convolutional kernel slides on the input data and calculates the weighted sum of the local area to extract local features. The convolutional kernel generates a feature map that represents different features of the input data. The rectified linear unit activation function (ReLU) is applied to the output of the convolutional layer to introduce nonlinear factors, allowing the network to learn more complex features. The feature map is downsampled through pooling operations (such as maximum pooling or average pooling) to reduce the data dimension and computation while retaining important features. Finally, the high-level features extracted by the convolution and pooling layers are expanded into a one-dimensional vector and further processed by a fully connected neural network. The fully connected layer calculates the output through linear transformation and activation function. Finally, the output of the fully connected layer is converted into a probability distribution through the Softmax activation function for classification judgment. Through the above steps, the convolutional neural network can perform deep learning on the input handwriting vector and drawing scene, extract key features, and finally output the classification result of the handwriting purpose. For example, the output result may contain two probability values: the probability of accidental demand and the probability of regular demand. If the probability of accidental demand is high, the system will perform special processing on these handwriting data; if the probability of regular demand is high, the system will process these data according to the normal process.

[0094] Further, adjusting the memory space of the temporary drawing layer and the cache layer according to the second overvoltage frequency includes:

[0095] If the second overpressure frequency is greater than the preset pressure frequency, the memory space of the temporary drawing layer and the cache layer is increased;

[0096] If the second overpressure frequency is less than or equal to the preset pressure frequency, the memory space of the temporary drawing layer and the cache layer is kept unchanged.

[0097] Furthermore, the formula for increasing the memory space of the temporary drawing layer and the cache layer is as follows:

[0098]

[0099] Among them, M new is the memory size after adjustment; M current is the current memory space size; frequency is the second overpressure frequency; threshold is the preset pressure frequency; G is the gradient coefficient.

[0100] In some embodiments, during a high-load drawing task, the system detects that the memory pressure exceeds the preset 80% threshold multiple times and records the relevant overpressure timestamp. After calculation, it is found that the first overpressure frequency is 95%, which triggers the handwriting purpose determination. After analyzing the data through a convolutional neural network, the system recalculates the second overpressure frequency to be 85%. Since the second overpressure frequency is higher than the preset threshold of 80%, the system decides to use the memory space of the temporary drawing layer and the cache layer through the above formula. The gradient coefficient can be used to adjust the extent of the memory increase, which can be reasonably set according to the size of the original device memory space.

[0101] 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 dynamic drawing and caching method for an electronic whiteboard large screen device, characterized in that: The following steps are involved: Initialize memory allocation of temporary drawing layer and cache layer in the memory space of the electronic whiteboard; Receive the user's handwriting data through the temporary drawing layer and provide real-time feedback, and receive the handwriting data of the temporary drawing layer through the cache layer based on the user's hand-raising state and clear the handwriting data of the temporary drawing layer; The handwriting data of the storage cache layer generates a data set, and the memory space of the temporary drawing layer and the cache layer is dynamically adjusted according to the data set and memory pressure.

2. The method for dynamic rendering and caching of a large-screen electronic whiteboard device according to claim 1, characterized in that: Initializing the temporary drawing layer and the cache layer in the memory space of the electronic whiteboard includes: Get the resolution setting of the current electronic whiteboard writing area, and calculate the memory space required for the writing area based on the resolution; Allocate memory space for temporary drawing layers and cache layers in the memory space required by the writing area; Monitor the memory usage of temporary drawing layers and cache layers, obtain the occupancy rate and calculate the memory pressure.

3. The dynamic drawing and caching method of a large-screen electronic whiteboard device according to claim 2, characterized in that: The receiving of the user's handwriting data through the temporary drawing layer and providing real-time feedback includes: The user's handwriting data is captured by the capacitive screen of the electronic whiteboard; The user handwriting data is temporarily stored by using the memory occupation of the temporary drawing layer, and the user handwriting data is rendered in real time using the GPU of the temporary drawing layer and the signal is output to the electronic whiteboard for display.

4. The method for dynamic rendering and caching of an electronic whiteboard large-screen device according to claim 1, characterized in that: The receiving of handwriting data of the temporary drawing layer through the cache layer based on the user's hand raising state and clearing the handwriting data of the temporary drawing layer comprises: Determine the user's hand-raising state according to the pause time obtained from the user's handwriting data and the preset continuous time, wherein the user's hand-raising state includes the hand-raising state and the hand-not-raising state; If the user's hand-raising state is "raised", the handwriting data of the temporary drawing layer is obtained through the cache layer and the handwriting data of the temporary drawing layer is cleared; If the user's hand-raising state is not raised, the user's handwriting data continues to be received through the temporary drawing layer.

5. The dynamic drawing and caching method of a large-screen electronic whiteboard device according to claim 1, characterized in that: The handwriting data of the storage cache layer generates a data set, comprising the following steps: Convert handwriting data and its timestamp into handwriting vector; The handwriting vectors are sorted and stored according to timestamps to construct a data set.

6. The method for dynamic rendering and caching of a large-screen electronic whiteboard device according to claim 1, characterized in that: The method of dynamically adjusting the memory space of the temporary drawing layer and the cache layer according to the data set and the memory pressure comprises the following steps: Classify the drawing scenes and distinguish the tasks based on the classification results; The timestamps when the memory pressure of the temporary drawing layer and the cache layer in a single task exceeds the preset pressure threshold are extracted and marked as overpressure timestamps, the total overpressure time is calculated according to the overpressure timestamps, and the first overpressure frequency is calculated according to the total overpressure time and the total time of the single task; Triggering handwriting purpose determination according to the first overpressure frequency, the handwriting purpose including accidental needs and regular needs; Remove the overpressure timestamps of handwritings for occasional needs and calculate the second overpressure frequency; The memory space of the temporary drawing layer and the cache layer is adjusted according to the second overvoltage frequency.

7. The method for dynamic rendering and caching of a large-screen electronic whiteboard device according to claim 6, characterized in that: The classifying of the drawing scenes and distinguishing the tasks based on the classification results include: Classifying the drawing scenes according to scene categories, wherein the scene categories include speech annotation, graphic drawing, and text input; Based on the determined drawing scenarios, tasks are differentiated according to different documents.

8. The method for dynamic rendering and caching of a large-screen electronic whiteboard device according to claim 6, characterized in that: The step of triggering acquisition of the handwriting vector and the drawing scene corresponding to the overpressure timestamp in the data set according to the first overpressure frequency comprises: If the first overpressure frequency is greater than the preset pressure frequency, the handwriting purpose determination is triggered; If the first overpressure frequency is less than or equal to the preset pressure frequency, the handwriting purpose determination is not triggered.

9. The method for dynamic rendering and caching of a large-screen electronic whiteboard device according to claim 6, characterized in that: The handwriting purpose determination includes obtaining the handwriting vector and drawing scene corresponding to the overpressure timestamp in the data set, and performing classification determination based on a convolutional neural network to determine the handwriting purpose.

10. The method for dynamic rendering and caching of a large-screen electronic whiteboard device according to claim 9, characterized in that: The formula of the convolutional neural network is as follows: Among them, Output is the final classification result of the convolutional neural network, including accidental demand and regular demand; W i is the weight of the i-th convolution kernel; x i is the i-th element of the input data, including the handwriting vector, overpressure timestamp and drawing scene; ReLU(z) is the rectified linear unit activation function, defined as ReLU(z)=max(0,z); Softmax(z) is the activation function.

11. The method for dynamic rendering and caching of a large-screen electronic whiteboard device according to claim 6, characterized in that: The step of adjusting the memory space of the temporary drawing layer and the cache layer according to the second overvoltage frequency includes: If the second overpressure frequency is greater than the preset pressure frequency, the memory space of the temporary drawing layer and the cache layer is increased; If the second overpressure frequency is less than or equal to the preset pressure frequency, the memory space of the temporary drawing layer and the cache layer is kept unchanged.

12. The method for dynamic rendering and caching of an electronic whiteboard large screen device according to claim 11, characterized in that: The formula for increasing the memory space of the temporary drawing layer and the cache layer is as follows: Among them, M new is the memory size after adjustment; M current is the current memory space size; frequency is the second overpressure frequency; threshold is the preset pressure frequency; G is the gradient coefficient.