Handwriting board handwriting identification method and system

Abstract

The invention relates to the technical field of handwriting identification, in particular to a handwriting board handwriting identification method and system, and the method comprises the following steps: obtaining handwriting data, screening pen falling coordinates, marking a stable section, adjusting the identification intensity, removing an invalid track, reconstructing a path, and outputting an identification record in a unified format. According to the method, the handwriting starting interval is determined through coordinate positioning of the pen falling event, the stable track section is identified and marked by combining jitter detection and continuous sudden change identification of the pressure difference value in the handwriting section, the identification intensity of the stable section is subjected to normalized limitation and identification weight adjustment, and low-pressure sensing signal interference is eliminated; according to the method, the tracks which do not reach the pressure sensitivity threshold value are removed, path reconstruction is carried out, the recognition process is focused on effective handwriting signals, unified time resampling and format mapping operation are matched, standardization and structural optimization of a handwriting data structure are achieved, the handwriting analysis precision, the anti-noise capacity and the recognition stability are improved, and the handwriting dynamic analysis capacity is enhanced.

Description

Technical Field

[0001] This invention relates to the field of handwriting recognition technology, and in particular to a handwriting recognition method and system for a handwriting tablet. Background Technology

[0002] The field of handwriting recognition technology involves the collection, processing, and recognition of handwriting information generated during human writing. This includes capturing handwriting trajectories, extracting and analyzing handwriting features, and determining the correspondence between handwriting data and characters. Typically, graphical input devices (touchscreens, handwriting tablets, etc.) are used to acquire dynamic parameters such as coordinates, speed, and pressure during the user's writing process. The handwriting trajectory is then used for character reconstruction and semantic analysis, and is widely applied in intelligent input, identity verification, and human-computer interaction systems. Among these, handwriting tablet handwriting recognition methods involve acquiring handwriting data from user input via graphical input devices and analyzing and recognizing the handwriting trajectory using template matching or neural network training-based character recognition methods. This extracts and interprets the unstructured handwriting information formed by the user on the handwriting tablet. Piezoelectric handwriting tablets are typically used for trajectory coordinate acquisition, combined with support vector machines or convolutional neural networks to complete character feature comparison and character classification.

[0003] In current handwriting recognition systems, the primary reliance is on basic acquisition of continuous handwriting trajectories and static feature recognition. However, the precise definition of the pen placement position can easily introduce interference from invalid trajectories in the initial stage. Furthermore, the processing of sampling pressure parameters fails to distinguish between valid signals and pressure fluctuations, leading to inaccurate judgment of trajectory stability. Additionally, the lack of a dynamic correction mechanism for recognition weights poses a risk of misrecognition of data segments with insufficient pressure sensitivity. Especially under conditions of complex writing actions or rapid handwriting input, problems such as blurred recognition, path distortion, and character restoration deviations can easily occur, affecting the response accuracy and character restoration reliability of the handwriting recognition system. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a handwriting recognition method for handwriting tablets.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a handwriting recognition method for a writing tablet, comprising the following steps:

[0006] S1: Acquire continuous handwriting data from the handwriting tablet, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, detect the coordinates of the sampling point where the first pressure value is not lower than the contact pressure threshold, use them as the coordinates of the pen stroke event, retain only the subsequent trajectory sequence, and generate the first segment of the handwriting trajectory interval.

[0007] S2: Read the pressure parameters of continuous sampling points in the first segment of the handwriting trajectory interval, determine whether the maximum and minimum pressure difference in the handwriting segment is lower than the pressure jitter tolerance, perform continuous sudden change judgment to mark stable segments, and generate a set of stable handwriting segment markers.

[0008] S3: Based on the stable marker set of the track handwriting segment, the ratio limit is applied to the original path recognition weight of the stable segment, the intensity data is limited to the normalized pressure coefficient, the original recognition weight of the handwriting segment is replaced with the updated intensity coefficient, and a set of handwriting segment recognition adjustment parameters is generated.

[0009] S4: Based on the handwriting segment recognition adjustment parameter set, remove handwriting segments whose intensity coefficient does not reach the effective pressure sensitivity threshold, and reconstruct the handwriting to generate pressure sensitivity focusing trajectory data;

[0010] S5: Based on the coordinate sequence, time sequence and recognition weight in the pressure-sensitive focusing trajectory data, perform feature format conversion, time resampling and intensity mapping operations respectively, and output the handwriting recognition record after unified encapsulation.

[0011] As a further embodiment of the present invention, the first segment of the handwriting trajectory interval includes an initial coordinate point, an activation start time, and an effective trajectory length; the handwriting segment stability marker set includes pressure stability identifier records, trajectory segment sequence indexes, and handwriting consistency labels; the handwriting segment recognition adjustment parameter set includes normalized intensity coefficients, recognition weight correction factors, and path confidence scalars; the pressure-sensitive focusing trajectory data includes reconstructed paths, filtered handwriting segments, and pressure-sensitive mapping results; and the handwriting recognition record on the writing tablet includes a coordinate sequence format, a unified time step, and an intensity weight encapsulation structure.

[0012] As a further aspect of the present invention, the step of obtaining the first segment of handwriting trajectory interval specifically includes:

[0013] S111: Obtain the continuous handwriting data sequence collected by the handwriting tablet terminal device, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, count all sampling points whose pressure parameters are not lower than the contact pressure threshold, and generate the contact pressure screening interval.

[0014] S112: Based on the contact pressure screening range, detect the coordinates of the sampling point where the pressure parameter is not lower than the contact pressure threshold for the first time, record the horizontal and vertical coordinates of the corresponding sampling point, and generate a pen drop event coordinate record.

[0015] S113: Based on the coordinate record of the pen stroke event, filter the corresponding sampling point and all subsequent sampling points in the continuous handwriting data sequence, count the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to all sampling points in the interval, obtain the overall trajectory change information, and generate the first segment of the handwriting trajectory interval.

[0016] As a further aspect of the present invention, the step of obtaining the trajectory segment stable marker set specifically includes:

[0017] S211: Based on the first segment of handwriting trajectory interval, obtain the pressure parameters of continuous sampling points in the interval, extract the pressure value corresponding to each sampling point in sequence, record the timestamp and horizontal and vertical coordinate information corresponding to each pressure value in the set, and generate a pressure parameter set.

[0018] S212: Based on the set of pressure parameters, perform pressure difference comparison point by point. For all pressure values ​​in each handwriting segment, extract the maximum pressure value and the minimum pressure value respectively, calculate the difference and obtain the pressure difference judgment value. If it is less than or equal to the pressure jitter tolerance, mark the corresponding handwriting segment as a stable segment and generate a stable segment marking result.

[0019] S213: Based on the stable segment marking results and combined with the handwriting continuity detection standard, determine the trajectory continuity between adjacent stable segments in turn. If the segments are determined to be continuous, merge the adjacent stable segments and organize them into overall stable marking information to obtain a set of stable handwriting segment markings.

[0020] As a further aspect of the present invention, the step of obtaining the handwriting segment recognition adjustment parameter set specifically includes:

[0021] S311: Based on the track handwriting segment stable marker set, the original path identification weights of each stable segment are read and grouped, and the pressure fluctuation amplitude between adjacent sampling points is compared to see if it is within the minimum pressure resolution range of the sampling device. If it is greater than the minimum pressure resolution value, the corresponding weight is marked as a weight to be restricted, forming a weight ratio before restriction.

[0022] S312: Based on the weight ratio before the restriction, compare the normalized pressure coefficient threshold range, compare each ratio with the pressure coefficient interval 0 to 1, if the ratio is less than 0, set it to zero, if the ratio is greater than 1, truncate it to 1, and calculate to obtain the normalized weight adjustment value.

[0023] S313: Based on the normalized weight adjustment value, the original path recognition weight of the stable segment is replaced in a covering manner, the path recognition weight of each sampling point in the stable segment is updated segment by segment, and the recognition number, normalized pressure coefficient and corresponding updated intensity coefficient of each sampling point are re-recorded to establish a set of handwriting segment recognition adjustment parameters.

[0024] As a further aspect of the present invention, the step of acquiring the pressure-sensitive focusing trajectory data specifically includes:

[0025] S411: Based on the handwriting segment recognition adjustment parameter set, read the intensity coefficient and effective pressure sensitivity threshold in each handwriting segment and compare them item by item. By comparing each intensity coefficient with the threshold point by point, organize the corresponding substandard intensity data according to the handwriting segment number and generate a substandard intensity list.

[0026] S412: Based on the list of substandard intensity, filter the corresponding coordinate point data and timestamp information, compare the sampling point data of substandard handwriting segments and complete trajectory segments, determine the continuity of the coordinate sequence, distinguish and process discontinuous point sequences, remove handwriting segments with substandard intensity thresholds, and obtain an effective pressure-sensitive filtering sequence.

[0027] S413: Based on the effective pressure-sensitive filtering sequence, combined with adjacent coordinate points and timestamp data, the trajectory path interpolation is performed on the continuous points in sequence. When performing interpolation, the coordinate distance and time interval of adjacent points are evenly divided to fill in the missing trajectory sampling points and generate pressure-sensitive focusing trajectory data.

[0028] As a further aspect of the present invention, the step of acquiring the handwriting recognition record on the writing tablet specifically includes:

[0029] S511: Based on the coordinate sequence in the pressure-sensitive focusing trajectory data, collect the coordinate value and point number of each trajectory point, combine the coordinate values ​​of adjacent points to calculate the horizontal and vertical increments, determine whether the increment is within the reference pixel range, traverse all coordinate point pairs in turn to filter, organize and classify to generate coordinate format conversion values;

[0030] S512: Based on the coordinate format conversion value, read the time series in the trajectory dataset, perform interval difference calculation on the timestamps of adjacent sampling points in sequence, determine whether the time interval deviates from the standard interval, and if there is a deviation between adjacent time intervals, perform interpolation processing on the time series to form a uniform distribution, and obtain the time resampling sequence.

[0031] S513: For the time resampling sequence, call the corresponding pressure sensitivity value and handwriting recognition weight for each sampling point, compare the two values ​​in turn, perform difference calculation and combination, combine the time point and coordinate point information to organize the values, and encapsulate the organized value sequence and the number of each trajectory point to generate a handwriting recognition record for the writing tablet.

[0032] A handwriting recognition system for a writing tablet includes:

[0033] The pen-drop detection module is used to execute S1: acquire continuous handwriting data from the handwriting tablet, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, detect the coordinates of the sampling point whose first pressure value is not lower than the contact pressure threshold, use them as the coordinates of the pen-drop event, retain only the subsequent trajectory sequence, and generate the first segment of the handwriting trajectory interval.

[0034] The pressure analysis module is used to execute S2: read the pressure parameters of continuous sampling points in the first segment of the handwriting trajectory interval, determine whether the maximum and minimum pressure difference in the handwriting segment is lower than the pressure jitter tolerance, perform continuous change judgment to mark stable segments, and generate a set of stable handwriting segment markers.

[0035] The weight adjustment module is used to execute S3: based on the track handwriting segment stability marker set, the ratio limit is applied to the original path recognition weight of the stable segment, the intensity data is limited to the normalized pressure coefficient, the updated intensity coefficient is used to replace the corresponding recognition weight of the original handwriting segment, and a handwriting segment recognition adjustment parameter set is generated.

[0036] The trajectory reconstruction module is used to perform S4: based on the handwriting segment recognition adjustment parameter set, remove handwriting segments whose intensity coefficient does not reach the effective pressure sensitivity threshold, and perform handwriting reconstruction to generate pressure-sensitive focusing trajectory data;

[0037] The recognition output module is used to execute S5: based on the coordinate sequence, time sequence and recognition weight in the pressure-sensitive focusing trajectory data, perform feature format conversion, time resampling and intensity mapping operations respectively, and output the handwriting recognition record after unified encapsulation.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0039] In this invention, the starting interval of the handwriting is determined by locating the coordinates of the pen stroke event. By combining the jitter detection of pressure difference within the handwriting segment with the recognition of continuous abrupt changes, stable trajectory segments can be identified and marked. Then, the recognition intensity of the stable segment is normalized and the recognition weight is adjusted to eliminate interference from low pressure-sensitive signals. By eliminating trajectories that do not reach the pressure sensitivity threshold and reconstructing the path, the recognition process focuses on effective handwriting signals. With the cooperation of unified time resampling and format mapping operations, the standardization and structural optimization of the handwriting data structure are achieved, improving the accuracy, noise resistance and recognition stability of handwriting analysis, and significantly enhancing the dynamic analysis capability and reconstruction reliability of unstructured handwriting. Attached Figure Description

[0040] Figure 1 This is a flowchart of the main steps of the present invention;

[0041] Figure 2 This is a flowchart of the process for obtaining the first segment of the handwriting trajectory range in this invention;

[0042] Figure 3 This is a flowchart of the process for obtaining the stable marker set of handwriting road segments according to the present invention;

[0043] Figure 4 This is a flowchart of the process for obtaining the set of adjustment parameters for handwriting segment recognition in this invention;

[0044] Figure 5 This is a flowchart of the pressure-sensitive focusing trajectory data acquisition process of the present invention;

[0045] Figure 6 This is a flowchart of the handwriting recognition and record acquisition process of the present invention. Detailed Implementation

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0047] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0048] Please see Figure 1 A handwriting recognition method for a writing tablet includes the following steps:

[0049] S1: Acquire the continuous handwriting data sequence collected by the handwriting tablet terminal device, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, detect the first sampling point coordinates where the pressure value is not lower than the contact pressure threshold (complies with the 0.1N stylus activation pressure in the USBHID standard), take the sampling point where the corresponding coordinates are located as the pen drop event coordinates (refer to the pen tip contact determination in the WacomFeel technical specification), retain only the subsequent trajectory sequence, and generate the first segment of handwriting trajectory interval;

[0050] S2: Read the pressure parameters of continuous sampling points in the first segment of the handwriting trajectory interval, perform difference comparison point by point, determine whether the difference between the maximum and minimum pressure parameters in each handwriting segment is lower than the pressure jitter tolerance (based on the ±5% pressure fluctuation range set by the Microsoft handwriting recognition API), and perform continuous change judgment (using the handwriting continuity detection standard of the "Pen Interaction Interface"). If the condition is met, mark the corresponding handwriting segment as a stable segment and generate a set of stable handwriting segment markers.

[0051] S3: Based on the stable marker set of the handwriting segment, the original path recognition weight of the stable segment is subjected to ratio restriction processing, the intensity data is limited to the normalized pressure coefficient (0-1 normalization processing of Unicode TR10 handwriting recognition standard is performed), and the corresponding recognition weight of the original handwriting segment is replaced with the updated intensity coefficient to generate a set of handwriting segment recognition adjustment parameters.

[0052] S4: Based on the handwriting segment recognition adjustment parameter set, perform intensity threshold filtering, and perform a rejection operation on handwriting segments whose intensity coefficient does not reach the effective pressure sensitivity threshold (≥2048 levels of pressure sensitivity according to the Wacom ProPen2 technical specification). Combined with coordinate point and timestamp information, perform handwriting reconstruction (execute the path interpolation algorithm of "Handwriting Animation Representation") to generate pressure sensitivity focusing trajectory data.

[0053] S5: Based on the coordinate sequence, time sequence and recognition weight in the pressure-sensitive focusing trajectory dataset, perform feature format conversion, time resampling (using a standard 50ms interval) and intensity mapping operations respectively, and then organize and encapsulate them in a unified data structure form to output the handwriting recognition record of the writing tablet.

[0054] The first segment of the handwriting trajectory range includes the initial coordinate point, activation start time, and effective trajectory length. The handwriting segment stability marker set includes pressure stability marker records, trajectory segment sequence index, and handwriting consistency labels. The handwriting segment recognition adjustment parameter set includes normalized intensity coefficient, recognition weight correction factor, and path confidence scalar. The pressure-sensitive focusing trajectory data includes reconstructed path, filtered handwriting segments, and pressure-sensitive mapping results. The handwriting recognition record of the handwriting tablet includes coordinate sequence format, unified time step, and intensity weight encapsulation structure.

[0055] Please see Figure 2 Step S1 is as follows:

[0056] S111: Obtain the continuous handwriting data sequence collected by the handwriting tablet terminal device, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, count all sampling points whose pressure parameters are not lower than the contact pressure threshold, and generate the contact pressure screening interval.

[0057] The process involves acquiring a continuous handwriting data sequence from a handwriting tablet terminal device. First, the connected handwriting tablet terminal is invoked, and its internal pressure and position sensing modules synchronously collect data in real time. The data includes timestamps, x-axis, y-axis, and pressure values. Timestamps are in milliseconds, x and y axes are in millimeters, and pressure values ​​are in Newtons (N). Following the WacomFeel technical specifications, the data acquisition frequency is set to 200 times per second to ensure continuous and stable data. After acquiring the data, the pressure values ​​at each data point are filtered. The filtering criterion is that the pressure value is not lower than the contact pressure threshold, which is 0.1N according to the stylus activation pressure in the USBHID standard. During the filtering process, the pressure value of each data point is compared with 0.1N. If the pressure value of a data point is ≥0.1N, the data point is considered valid. If the pressure is 1N, the data point is retained; otherwise, it is discarded. The filtering process is as follows: Using sample data, assume the pressure values ​​of five consecutively acquired data points are 0.05N, 0.12N, 0.15N, 0.09N, and 0.13N. The comparison process is as follows: the first data point (0.05N < 0.1N) is discarded; the second data point (0.12N ≥ 0.1N) is retained; the third data point (0.15N ≥ 0.1N) is retained; the fourth data point (0.09N < 0.1N) is discarded; and the fifth data point (0.13N ≥ 0.1N) is retained. The final filtering result is the 2nd, 3rd, and 5th data points. The filtered data points form a continuous data interval, defined as the contact pressure filtering interval. The data points in this interval are the basis for subsequent pen placement determination and trajectory extraction. Data examples are shown in the table below.

[0058] Table 1. Contact Pressure Screening Range Data Table

[0059] Timestamp (ms) x-axis (mm) Vertical axis (mm) Pressure value (N) 102 15.2 30.1 0.12 107 15.5 30.4 0.15 115 15.9 30.8 0.13

[0060] As shown in Table 1, the final selected data range is the contact pressure screening range.

[0061] S112: Based on the contact pressure screening range, detect the coordinates of the sampling point where the pressure parameter is not lower than the contact pressure threshold for the first time, record the corresponding horizontal and vertical coordinates of the sampling point, and generate the pen drop event coordinate record.

[0062] Based on the contact pressure screening interval, the coordinates of the sampling point where the pressure parameter is not lower than the contact pressure threshold of 0.1N are detected for the first time. First, the data of the contact pressure screening interval in Table 1 is called, and each data point is traversed in turn to compare whether its pressure value meets the standard of not lower than 0.1N. Since the interval itself has already screened out points less than 0.1N, the first data point is directly taken as the sampling point corresponding to the pen stroke event. The horizontal and vertical coordinates of the sampling point are extracted. In the example, the timestamp of the first data point in Table 1 is 102ms, the horizontal coordinate is 15.2mm, the vertical coordinate is 30.1mm, and the pressure value is 0.12N. This point meets the standard of not lower than 0.1N. Therefore, the horizontal and vertical coordinates of this point are extracted as the coordinate values ​​of the pen stroke event. In the subsequent interface display or data record, this coordinate is the coordinate of the starting point of the handwriting. The final result is the pen stroke event coordinate record, and its value is (15.2mm, 30.1mm).

[0063] S113: Based on the coordinate record of the pen stroke event, filter the corresponding sampling point and all subsequent sampling points in the continuous handwriting data sequence, statistically analyze the timestamp, horizontal and vertical coordinates and pressure parameters of all sampling points in the interval, obtain the overall trajectory change information, and generate the first segment of the handwriting trajectory interval.

[0064] Based on the coordinate records of the pen stroke event, the sampling point and all subsequent sampling points in the continuous handwriting data sequence are filtered. First, the position of the data point corresponding to the pen stroke event in the complete handwriting data sequence is located. Let the sequence number of this data point be n. Starting from this data point, all data points with numbers ≥ n in the sequence are extracted to form a continuous handwriting trajectory data set. The timestamp, x-coordinate, y-coordinate and pressure parameter of each data point in the set are called, and the trajectory change information is statistically analyzed point by point. Combining examples, assuming that the pen stroke event corresponds to the 50th data point in the complete data sequence, and a total of 150 data points are collected, the data of the 50th to 200th data points are extracted. During the statistical process, for each data point, the changes in x-coordinate and y-coordinate Δx and Δy, as well as the pressure change value ΔP are calculated. Δx = x-coordinate of the current point - x-coordinate of the previous point, Δy = y-coordinate of the current point - y-coordinate of the previous point, ΔP = pressure value of the current point - pressure value of the previous point. Through the above statistics, the dynamic changes in the overall trajectory interval are gradually obtained, and the final result is the first segment of the handwriting trajectory interval.

[0065] Please see Figure 3 Step S2 is as follows:

[0066] S211: Based on the first segment of the handwriting trajectory interval, obtain the pressure parameters of continuous sampling points in the interval, extract the pressure value corresponding to each sampling point in sequence, record the timestamp and horizontal and vertical coordinate information corresponding to each pressure value in the set, and generate a pressure parameter set.

[0067] Based on the initial handwriting trajectory interval, the pressure parameters of continuous sampling points within this interval are first obtained. Specifically, the pressure information of each sampling point in the trajectory interval is called sequentially, with the sampling point numbers ranging from 1 to n. The pressure value of each point is read, in Newtons. For example, assuming there are 10 sampling points in the initial handwriting trajectory interval, with pressure values ​​of 0.12N, 0.15N, 0.13N, 0.14N, 0.16N, 0.15N, 0.13N, 0.17N, 0.14N, and 0.15N respectively, the timestamp and horizontal and vertical coordinates corresponding to each pressure value are recorded. The timestamp is in milliseconds, and the horizontal and vertical coordinates are in millimeters. All pressure values ​​are organized into a pressure parameter set, and corresponding time and location data sets are also established to form a data table. An example of the data structure is shown below.

[0068] Table 2 Example of Pressure Parameter Set

[0069] Sampling point number Timestamp (ms) x-axis (mm) Vertical axis (mm) Pressure value (N) 1 101 15.2 30.1 0.12 2 105 15.5 30.4 0.15 3 108 15.7 30.6 0.13 4 112 16.0 30.9 0.14 5 116 16.3 31.2 0.16 6 119 16.5 31.4 0.15 7 123 16.8 31.7 0.13 8 127 17.1 32.0 0.17 9 131 17.4 32.3 0.14 10 134 17.6 32.5 0.15

[0070] As shown in Table 2, the pressure parameter set is finally obtained by extracting and organizing the parameters point by point.

[0071] S212: Based on the pressure parameter set, perform pressure difference comparison point by point. For all pressure values ​​within each handwriting segment, extract the maximum and minimum pressure values, calculate the difference, and apply the formula:

[0072]

[0073] The pressure difference determination value is obtained through calculation. If it is less than or equal to the pressure fluctuation tolerance, the corresponding handwriting segment is marked as a stable segment, and a stable segment marking result is generated; where D p This represents the pressure difference determination value, in Newtons (P). max This represents the maximum pressure value for the corresponding handwriting segment, in Newtons (P). min This represents the minimum pressure value for the corresponding handwriting segment, in Newtons (P). i This represents the pressure value at the i-th sampling point of the corresponding handwriting segment, in Newtons. This represents the average pressure value of all sampling points within the corresponding handwriting segment, in Newtons, and m represents the number of sampling points within the corresponding handwriting segment.

[0074] Based on the pressure parameter set, pressure difference comparisons are performed point by point. Specifically, for each handwriting segment, the maximum and minimum pressure values ​​are first calculated. Referring to the data in Table 3, the maximum pressure value among the 10 pressure values ​​is 0.17N, and the minimum is 0.12N. The difference between the two is calculated as 0.17N - 0.12N = 0.05N. Then, all pressure values ​​are extracted, and the sum of the absolute deviations between each pressure value and the average pressure value is calculated. The average pressure value is calculated as follows:

[0075]

[0076] Furthermore, the absolute differences between each pressure value and the average pressure value are calculated, as shown in Table 3:

[0077] Table 3. Record of Pressure Values ​​and Absolute Differences

[0078] Pressure value (N) Absolute difference (N) 0.12 0.024 0.15 0.006 0.13 0.014 0.14 0.004 0.16 0.016 0.15 0.006 0.13 0.014 0.17 0.026 0.14 0.004 0.15 0.006

[0079] The sum of absolute deviations is 0.124N. Substituting the above data into the formula:

[0080]

[0081] Based on the pressure jitter tolerance of ±5% set by the Microsoft handwriting recognition API, and considering the actual range of the pressure parameter device (0-1N), the 5% pressure jitter tolerance value is 0.05N. Comparing the calculated result (0.0624N > 0.05N), it is determined that the handwriting segment does not meet the pressure stability standard and is not marked as a stable segment. If the subsequent calculation result is less than or equal to 0.05N, the handwriting segment will be marked, generating a stable segment marking result.

[0082] The pressure difference judgment value reflects a comprehensive numerical value of the overall pressure change within a handwriting segment. This value simultaneously considers the difference between the maximum and minimum pressure within the handwriting segment, as well as the deviation of the pressure values ​​at all sampling points from the overall average pressure value. The difference between the maximum and minimum pressure reflects whether there is a significant instantaneous pressure change in the local area of ​​the handwriting segment, while the average deviation of the pressure at each sampling point reflects the uniformity of the pressure distribution during the overall continuous writing process. With the combined effect of these two parts, the pressure difference judgment value can comprehensively describe the pressure fluctuation amplitude and dispersion within the handwriting segment. The smaller the value, the more stable and gradual the pressure change within the handwriting segment, and the more uniform the pressure distribution during the overall writing process, which is closer to the ideal stable writing state. The larger the value, the more abrupt the pressure changes within the handwriting segment or the significant overall fluctuation, which fails to meet the stability requirements. Therefore, the pressure difference judgment value is an important reference for measuring whether a handwriting segment has pressure stability characteristics.

[0083] The formula first uses the difference between the maximum and minimum pressure values, |P max -P minThe first part directly reflects the range of pressure changes within the handwriting segment, enabling rapid assessment of the overall fluctuation. The second part calculates the average deviation of pressure within the handwriting segment by summing the absolute differences between the pressure values ​​at each sampling point and the average pressure value, and then dividing by the total number of sampling points, *m*. This part measures the dispersion of the overall pressure distribution. The combined structure reflects a dual assessment of handwriting pressure fluctuations, considering both extreme pressure changes and overall stability distribution. The former highlights instantaneous fluctuations, while the latter reflects overall equilibrium. Through this combined logic, the formula avoids misjudgments caused by relying solely on a single extreme value and accurately reflects instability even when the overall pressure change is small but there are short-term, drastic fluctuations. All parts of the entire structure use Newton units, ensuring a unified physical dimension for the addition operation. The final output pressure difference judgment value comprehensively quantifies the degree of pressure fluctuation within the handwriting segment, demonstrating logical scientific rationality.

[0084] S213: Based on the stable segment marking results and combined with the handwriting continuity detection standard, determine the trajectory continuity between adjacent stable segments in turn. If the continuity is determined, merge the adjacent stable segments, organize them into overall stable marking information, and obtain the handwriting segment stable marking set.

[0085] Based on the stable segment marking results and combined with the handwriting continuity detection standard of the "Pen-based Interactive Interface", the trajectory continuity between adjacent stable segments is determined sequentially. Specifically, for all handwriting segments marked as stable segments, the changes in horizontal and vertical coordinates and the time interval between the end sampling point and the start sampling point of two adjacent segments are compared. If the changes in both horizontal and vertical coordinates do not exceed 0.5mm and the time interval does not exceed 10ms, the trajectory is determined to be continuous; otherwise, it is discontinuous. For example, suppose the end point of the 3rd stable segment has an horizontal coordinate of 17.4mm, a vertical coordinate of 32.3mm, and a timestamp of 131ms, and the start point of the 4th segment has an horizontal coordinate of 17.5mm, a vertical coordinate of 32.4mm, and a timestamp of 140ms. The changes in horizontal and vertical coordinates are 0.1mm and 0.1mm respectively, with a time interval of 9ms. These meet the above criteria, so the trajectory is determined to be continuous. The two segments are merged into a whole stable segment. This operation is repeated to form complete stability information and obtain a set of stable handwriting segment markings.

[0086] Please see Figure 4 Step S3 is as follows:

[0087] S311: Based on the stable marker set of the track handwriting segment, the original path identification weights of each stable segment are read and grouped. The pressure fluctuation amplitude between adjacent sampling points is compared to see if it is within the minimum pressure resolution range of the sampling device. If it is greater than the minimum pressure resolution value, the corresponding weight is marked as the weight to be restricted, forming the weight ratio before restriction.

[0088] Based on the stable marker set of track handwriting segments, when reading and grouping the original path recognition weights of each stable segment, starting from the first stable segment, the initial pressure value and original path recognition weight of each sampling point are retrieved. Assuming the trajectory sampling points are numbered from 1 to 10, the pressure values ​​are collected sequentially as follows: point 1: 0.12; point 2: 0.14; point 3: 0.11; point 4: 0.15; point 5: 0.16; point 6: 0.18; point 7: 0.15; point 8: 0.14; point 9: 0.13; point 10: 0.15. All units are Newtons. The corresponding original path recognition weights are 0.35, 0.40, 0.33, 0.38, 0.45, and 0.50, respectively. The pressure values ​​are 0.42, 0.39, 0.37, and 0.40, with dimensionless units. Each pressure value is compared with the pressure fluctuation amplitude of adjacent points. When comparing, the pressure difference between point n and point n+1 is called to determine whether it exceeds the minimum pressure resolution of the equipment, which is 0.02 Newtons. If the pressure difference between point 1 and point 2 is 0.02 Newtons, it is determined to be under boundary conditions and needs to be included in the limit. If the pressure difference between point 2 and point 3 is 0.03 Newtons, it exceeds the resolution, and point 3 is marked as a limit. This process is repeated until point 10 to form a complete sequence of weights to be limited. Then, each marked weight is grouped to form a set of weights that need to be processed by the limit ratio within each stable segment. The ratio set is divided for subsequent calculations. Table 4 shows the stable segment pressure and original weight table.

[0089]

[0090]

[0091] As shown in Table 4, by calling the pressure value of the sampling point and the weight and comparing it with the minimum resolution, the set of weight ratios before the constraint is obtained.

[0092] S312: Based on the pre-weighted ratio, compare it with the normalized pressure coefficient threshold range, and compare each ratio with the pressure coefficient interval of 0 to 1. If the ratio is less than 0, it is set to zero; if the ratio is greater than 1, it is truncated to 1, and the formula is applied:

[0093]

[0094] The normalized weight adjustment value is obtained by calculation, where W adj This represents the path recognition weight adjustment value. This represents the normalized original path recognition weight (0~1) for the i-th sampling point. The normalized pressure coefficient (0-1) for the i-th sampling point is represented by n, where n represents the total number of sampling points to be weighted.

[0095] Based on the pre-constraint weight ratio set, the normalized pressure coefficient threshold range of 0 to 1 for each point is used as the comparison benchmark. The original value of the weight to be constrained is compared with the corresponding normalized pressure coefficient point by point. If the ratio of a single point is less than 0, it is set to 0; if it is greater than 1, it is set to 1. After normalization processing is performed on all sampling points that meet the conditions after comparison, the weight is adjusted using a modified formula. Specifically, the original path identification weight of each point is first normalized. For example, n = 4 sampling points are selected, which are: After normalization comparison confirmed that all samples were within the interval, they were substituted into the formula, and the calculations were performed item by item according to the sampling point order:

[0096] Item 1:

[0097]

[0098] Item 2:

[0099]

[0100] Item 3:

[0101]

[0102] Item 4:

[0103]

[0104] Sum all terms:

[0105]

[0106] The obtained normalized weight adjustment value is 0.4755. Compared with the normalized interval of 0 to 1, it meets the restriction conditions and can be used as the basis for subsequent ratio replacement to obtain the normalized weight restriction value.

[0107] The normalized weight adjustment value refers to the weight correction result obtained by averaging the original path recognition weights and pressure sampling values ​​at the same location after normalization of the original path recognition weights and the pressure sampling values ​​at the same location during the path recognition process. This value measures the degree of direct deviation between the path weights and the pressure coefficients through absolute value calculation, and then describes the mutual matching relationship between the two through the square root of the product. The combination of the two reflects the contribution intensity of each sampling point to the overall handwriting segment recognition stability. After averaging the weighted results of all sampling points, the normalized weight adjustment value can be used to constrain the ratio of the original recognition weights that exceed the limit range or fluctuate greatly, ensuring that all updated weights are within the reasonable range of 0 to 1 after normalization. This allows the recognition weights based on a unified scale to match the pressure stability judgment requirements of adjacent segments during subsequent handwriting segment merging or trajectory continuity analysis. The normalized weight adjustment value is essentially a balance measure of the relationship between path weights and pressure features, and it has the feasibility of being repeatedly called, compared, and replaced.

[0108] The formula uses a two-part summation structure to jointly measure the difference and coupling between the original path identification weights and the normalized pressure coefficients at each sampling point, where the absolute value term... This represents a direct quantification of the degree of deviation at each sampling point, used to measure the direct distance between the weight and the pressure coefficient after normalization, ensuring that the deviation is non-negative, and the product square root term. This is used to reflect the degree of interaction and coupling between the two. When the weight and pressure values ​​are both large or both small, the square root term of the product smoothly corrects the overall result, avoiding the unidirectional pull of extreme differences on the final adjustment value. The two are added together to form a single-point correction value, and then summed over all sampling points. The summation result is divided by n to achieve equalization of the entire data segment, making the output W... adj Maintaining comparability consistent with the normalized interval ensures that the overall dimensions are consistent when used for ratio replacement, and that the calculation logic can reflect both local deviations and the nonlinear distribution of the overall weights, thus forming a reasonable limit value for the path identification weights.

[0109] S313: Based on the normalized weight adjustment value, the original path recognition weight of the stable section is replaced in an overlay manner, the path recognition weight of each sampling point in the stable section is updated segment by segment, and the recognition number, normalized pressure coefficient and corresponding updated intensity coefficient of each sampling point are re-recorded to establish a set of handwriting segment recognition adjustment parameters.

[0110] When replacing the original path recognition weights based on the normalized weight limit value, the normalized weight limit value of 0.25 obtained from the previous segment is used as the ratio replacement benchmark. The original path recognition weights of points 1 to 10 are adjusted according to the ratio. If the original weight of point 1 is 0.35, it will be updated to 0.35 × 0.25 = 0.0875. If the original weight of point 2 is 0.40, it will be updated to 0.10. All sampling points are updated in sequence to form the updated intensity coefficient sequence of each stable segment. The corresponding sampling point number, normalized pressure coefficient, and updated intensity coefficient are recorded again. The results of all stable segments are summarized and combined by segment to establish a set of handwriting segment recognition adjustment parameters.

[0111] Please see Figure 5 Step S4 is as follows:

[0112] S411: Based on the handwriting segment recognition adjustment parameter set, read the intensity coefficient and effective pressure sensitivity threshold in each handwriting segment and compare them item by item. By comparing each intensity coefficient with the threshold point by point, organize the corresponding non-compliant intensity data according to the handwriting segment number and generate a non-compliant intensity list.

[0113] Based on the handwriting segment recognition parameter set, when sequentially calling the intensity coefficient value of each sampling point in each handwriting segment and converting it into the corresponding pressure sensitivity level value, the example sampling point number and intensity coefficient value are first organized and then numerically converted. For example, the pressure sensitivity value of sampling point A with a coefficient value of 0.85 is converted to 1800, the pressure sensitivity value of sampling point B with a coefficient value of 0.95 is converted to 2100, and the pressure sensitivity value of sampling point C with a coefficient value of 0.80 is converted to 1750. Then, each point is compared with the pressure sensitivity threshold of 2048 of the Wacom ProPen2 device. The pressure sensitivity difference of sampling point A is calculated as ΔA = 1800 - 2048 = -248, which is considered less than zero and thus not meeting the standard. The calculation for sampling point B is ΔB = 2100 - 2048. If 48 = 52 is greater than zero, it is considered compliant. If the sampling point C's calculation ΔC = 1750 - 2048 = -298 is still less than zero, it is considered non-compliant. Continue performing difference calculations on other sampling points in the same section and organize them. Then, the sampling points determined to be non-compliant are judged for continuity. If the difference is 1, they are considered continuous; otherwise, they are split. For example, the differences in the number sequence 1, 2, 3, 4, 6 are 2-1 = 1, 3-2 = 1, 4-3 = 1, and 6-4 = 2, respectively. Numbers 1 to 4 form a continuous segment, and number 6 forms a separate segment. Record the pressure difference and the continuous segment number. After organizing the non-compliant intensity sampling points and their corresponding continuous segments, a summary data example is shown in Table 5. Table 5: Summary Table of Non-compliant Intensity Sampling Points.

[0114]

[0115]

[0116] As shown in Table 5, the list of substandard intensity is obtained after clarifying the pressure sensitivity values ​​and numbering relationships of each substandard sampling point.

[0117] S412: Based on the list of substandard intensity, filter the corresponding coordinate point data and timestamp information, compare the sampling point data of substandard handwriting segments and complete trajectory segments, determine the continuity of the coordinate sequence, distinguish and process discontinuous point sequences, remove handwriting segments with substandard intensity thresholds, and obtain effective pressure-sensitive filtering sequences.

[0118] Based on the list of substandard intensity levels, when filtering the coordinate values ​​and timestamp information of each sampling point in the substandard segments, the coordinates and timestamps of sample point 1 are first retrieved as X1 = 35.5, Y1 = 42.2, t1 = 0.02s, and those of sample point 2 are X2 = 36.0, Y2 = 42.5, t2 = 0.03s. The coordinate increment and time increment of adjacent sampling points are then calculated as ΔX. 1-2 =X2-X1=0.5, ΔY 1-2 =Y2-Y1=0.3, Δt 1-2 =t2-t1=0.01s, determine ΔX 1-2 with ΔY 1-2 If all values ​​are within the range of 0.5 to 2.0 pixels, then the data is considered continuous. If the values ​​of adjacent sampling points 3 are X3 = 36.2, Y3 = 42.9, and t3 = 0.04s, then ΔX... 2-3 =X3-X2=0.2, ΔY 2-3 =Y3-Y2=0.4, at this time ΔX 2-3 If a point is outside the range of 0.5 to 2.0, it is considered a breakpoint and removed. This process is repeated for all points within the non-compliant segments, performing coordinate increment and time series analysis. If fewer than two sampling points remain after removal, the entire segment is deleted. If two or more points remain, the coordinate and timestamp sequences of that continuous segment are retained for trajectory reconstruction. If timestamps are in reverse order, the process is based on the judgment t. j+1 <t j The sampling point order is swapped before and after the event, and the final result after rejection is compared with the original trajectory segment by number. After the filtering is completed, the retained sequence is called to obtain the effective pressure-sensitive filtering sequence.

[0119] S413: Based on the effective pressure-sensitive filtering sequence, combined with adjacent coordinate points and timestamp data, the trajectory path interpolation is performed on continuous points in sequence. When performing interpolation, the coordinate distance and time interval of adjacent points are evenly divided to fill in the missing trajectory sampling points and generate pressure-sensitive focusing trajectory data.

[0120] Based on the effective pressure-sensitive filtering sequence, when performing path interpolation on adjacent sampling points within a continuous segment, first select sample values ​​X for adjacent sampling points A and B, respectively. A =35.5, Y A =42.2, tA =0.02s and X B =36.0, Y B =42.5, t B =0.03s, adjacent increments are calculated as ΔX A-B =X B -X A =0.5, ΔY A-B =Y B -Y A =0.3, Δt A-B =t B -t A =0.01s, if this increment needs to be divided into 2 equal parts for interpolation, then the single increment is The coordinates of the generated interpolation point 1 are X I1 =X A +δX=35.75、Y I1 =Y A +δY=42.35、t I1 =t A +δt=0.025s, the coordinates of interpolation point 2 are X I2 =X I1 +δX=36.0、Y I2 =Y I1 +δY=42.5、t I2 =t I1 +δt=0.03s, repeat the interpolation increment allocation and interpolation point generation process for the remaining adjacent sampling points in the same continuous segment. If the distance between adjacent sampling points in a certain segment is greater than 2.0 pixels, the increment should be divided into segments of no less than 3 parts as needed to ensure that the increment is within the range of 0.5 to 2.0 and that the timestamp increment is synchronously and evenly distributed without reversal. Finally, the interpolation point and the original sampling point number are merged to form a complete trajectory point sequence, and the pressure-sensitive focusing trajectory data is obtained.

[0121] Please see Figure 6 The S5 steps are as follows:

[0122] S511: Based on the coordinate sequence in the pressure-sensitive focusing trajectory data, the coordinate value and point number of each trajectory point are collected, and the horizontal and vertical increments are calculated by combining the coordinate values ​​of adjacent points. It is determined whether the increment is within the reference pixel range, and all coordinate point pairs are traversed in turn for filtering, sorting and classifying to generate coordinate format conversion values.

[0123] Based on the coordinate sequence in the pressure-sensitive focusing trajectory dataset, obtain the X-axis of each trajectory point. i With Y i The coordinate values ​​and corresponding point numbers are determined by detecting changes in the initial coordinates of a single trajectory point and the coordinate values ​​of adjacent points. For each pair of adjacent points, the horizontal and vertical increments are calculated, with the horizontal increment calculated using ΔX.i =X i+1 -X i The vertical increment is obtained by calculating ΔY. i =Y i+1 -Y i To ensure the increments are within a reasonable range, all increments need to be compared with the set range of 0.5 to 2.0 pixels. This comparison process requires a comparison operation. For example, when sampling points X1 = 12.5 and X2 = 13.0, ΔX1 = 0.5, which meets the 0.5 to 2.0 range standard. If ΔX... i <0.5 or ΔX i If the value is greater than 2.0, the point pair needs to be removed from the sequence. For point pairs whose adjacent point increments meet the interval requirements in both the horizontal and vertical directions, their numbers should be recorded and they should be classified as qualified coordinate increment points. At the same time, abnormal increment points should be numbered and marked. The rationality should be further verified using actual handwriting sampling examples. For example, taking 5 trajectory points numbered 1 to 5, the corresponding coordinates are as follows:

[0124] X=[12.5, 13.0, 15.0, 15.7, 17.9]

[0125] Y=[10.2, 10.8, 12.0, 12.5, 14.0]

[0126] The horizontal increments are 0.5 for the first pair, 2.0 for the second, 0.7 for the third, and 2.2 for the fourth. The fourth pair exceeding 2.0 needs to be marked as an outlier and removed. In actual use, the physical unit range of the handwriting trajectory must correspond to the actual resolution of the handwriting tablet. Sampling data must be consistent in mm or pixel units and converted according to hardware resolution to ensure comparability between different devices. Furthermore, the interval thresholds must be experimentally verified. For example, using the same pressure conditions to sample 100 pairs of points, the applicability of the threshold is determined when the horizontal increment coverage reaches 95% in the 0.5 to 2.0 pixel range. The table below lists example trajectory points and increment judgment results.

[0127] Table 6 Example of coordinate increment judgment

[0128] Point number X coordinate Y coordinate Horizontal increment Vertical increment Judgment Result 1-2 12.5 10.2 0.5 0.6 qualified 2-3 13.0 10.8 2.0 1.2 qualified 3-4 15.0 12.0 0.7 0.5 qualified 4-5 15.7 12.5 2.2 1.5 Eliminate

[0129] As shown in Table 6, the incremental judgment is based on the consistency between coordinate changes and threshold intervals. Finally, the coordinate format conversion value is obtained by sorting out the retained point numbers and incremental sequences after elimination.

[0130] S512: Based on the coordinate format conversion value, read the time series in the trajectory dataset, perform interval difference calculation on the timestamps of adjacent sampling points in turn, determine whether the time interval deviates from the standard interval, and if there is a deviation between adjacent time intervals, perform interpolation processing on the time series to form a uniform distribution, and obtain the time resampling sequence.

[0131] Based on the coordinate format conversion value, the time series data in the trajectory dataset is retrieved, and the timestamps of adjacent sampling points are detected. Time resampling is performed by comparing the time difference between adjacent sampling points with the standard 50-millisecond interval. First, the difference is calculated for each pair of adjacent time points. If the difference deviates by more than 5 milliseconds, interpolation is performed to generate a new time point to maintain the consistency of the time series. For example, if the sampling point timestamps are t1 = 0.050s and t2 = 0.120s, the time interval is 70 milliseconds, exceeding the set threshold of 50 milliseconds ± 5 milliseconds. Therefore, a new timestamp t needs to be inserted. new =0.100s to homogenize the time series. If there are multiple consecutive large intervals in the whole sampling, multiple interpolations are required. After interpolation, the continuity of the new time series needs to be verified again and the time index updated. In the example, if the total number of trajectory points is 100 and 20 point pairs exceed the standard time interval, the actual number of interpolations needs to reach 20 to ensure data integrity. The comparison before and after time resampling shows that the sampling density has increased from the original 14Hz to 20Hz, which meets the standard handwriting tablet data sampling rate range. Finally, the time series after interpolation adjustment forms the time resampling sequence.

[0132] S513: For the time resampling sequence, call the corresponding pressure sensitivity value and handwriting recognition weight for each sampling point, compare the two values ​​in turn, perform difference calculation and combination, combine the time point and coordinate point information to organize the values, and encapsulate the organized value sequence and the number of each trajectory point to generate a handwriting recognition record for the handwriting tablet.

[0133] For time-resampled sequences, the pressure sensitivity value and recognition weight corresponding to each sampling point are called, and the numerical difference is compared before being combined and mapped. In actual operation, the original pressure sensitivity value P of each point needs to be called first. i With corresponding weight W i The difference between the two is calculated as |P i -W i | Operations with square root terms and timestamp value T i With coordinate value C i Synchronously imported data is processed, and intensity mapping values ​​are generated point by point. These values ​​are then aggregated from all sampling points, combined with a time index and trajectory number, and encapsulated into a unified structure. The resulting summary of combined values ​​constitutes the handwriting recognition record for the tablet, where P... iW samples data in real-time using a handwriting tablet and normalizes it to the device's pressure sensitivity threshold of 2048 levels. i After validation on the training dataset, T can be fixed in the range of 0.85 to 1.0. i and C i This is directly derived from the time resampling sequence and coordinate format conversion value. In the example, if point number 10 is P 10 =1700, then after normalization it becomes 0.83, W 10 =0.9, then the difference is 0.07, and the square root of the product term is approximately 0.86. This result is accumulated with the results of the other 99 sampling points to form a complete dataset and output as a recognition record, thus obtaining the handwriting recognition record of the writing tablet.

[0134] A handwriting recognition system for a writing tablet includes:

[0135] The pen-drop detection module is used to execute S1: acquire continuous handwriting data from the handwriting tablet, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, detect the coordinates of the sampling point whose first pressure value is not lower than the contact pressure threshold, use them as the coordinates of the pen-drop event, retain only the subsequent trajectory sequence, and generate the first segment of the handwriting trajectory interval.

[0136] The pressure analysis module is used to execute S2: read the pressure parameters of continuous sampling points in the first segment of the handwriting trajectory interval, determine whether the maximum and minimum pressure difference in the handwriting segment is lower than the pressure jitter tolerance, perform continuous change judgment to mark stable segments, and generate a set of stable handwriting segment markers.

[0137] The weight adjustment module is used to execute S3: Based on the stable marker set of the track handwriting segment, the original path recognition weight of the stable segment is restricted by a ratio, the intensity data is limited to the normalized pressure coefficient, the original recognition weight of the handwriting segment is replaced with the updated intensity coefficient, and the handwriting segment recognition adjustment parameter set is generated.

[0138] The trajectory reconstruction module is used to execute S4: based on the handwriting segment recognition, adjust the parameter set, remove handwriting segments whose intensity coefficients do not reach the effective pressure sensitivity threshold, and reconstruct the handwriting to generate pressure-sensitive focusing trajectory data;

[0139] The recognition output module is used to execute S5: based on the coordinate sequence, time sequence and recognition weight in the pressure-sensitive focusing trajectory data, it performs feature format conversion, time resampling and intensity mapping operations respectively, and outputs the handwriting recognition record after unified encapsulation.

[0140] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A handwriting recognition method for a writing tablet, characterized in that, Includes the following steps: S1: Acquire continuous handwriting data from the handwriting tablet, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, detect the coordinates of the sampling point whose first pressure value is not lower than the contact pressure threshold, use them as the coordinates of the pen stroke event, retain only the subsequent trajectory sequence, and generate the first segment of the handwriting trajectory interval. S2: Read the pressure parameters of continuous sampling points in the first segment of the handwriting trajectory interval, determine whether the maximum and minimum pressure difference in the handwriting segment is lower than the pressure jitter tolerance, perform continuous sudden change judgment to mark stable segments, and generate a set of stable handwriting segment markers. S3: Based on the stable marker set of the track handwriting segment, the ratio limit is applied to the original path recognition weight of the stable segment, the intensity data is limited to the normalized pressure coefficient, the original recognition weight of the handwriting segment is replaced with the updated intensity coefficient, and a set of handwriting segment recognition adjustment parameters is generated. S4: Based on the handwriting segment recognition adjustment parameter set, remove handwriting segments whose intensity coefficient does not reach the effective pressure sensitivity threshold, and reconstruct the handwriting to generate pressure sensitivity focusing trajectory data; S5: Based on the coordinate sequence, time sequence and recognition weight in the pressure-sensitive focusing trajectory data, perform feature format conversion, time resampling and intensity mapping operations respectively, and output the handwriting recognition record after unified encapsulation.

2. The handwriting recognition method for a writing tablet according to claim 1, characterized in that, The first segment of the handwriting trajectory range includes the initial coordinate point, activation start time, and effective trajectory length. The handwriting segment stability marker set includes pressure stability identifier records, trajectory segment sequence index, and handwriting consistency labels. The handwriting segment recognition adjustment parameter set includes normalized intensity coefficient, recognition weight correction factor, and path confidence scalar. The pressure-sensitive focusing trajectory data includes reconstructed path, filtered handwriting segments, and pressure-sensitive mapping results. The handwriting recognition record on the writing tablet includes coordinate sequence format, unified time step, and intensity weight encapsulation structure.

3. The handwriting recognition method for a writing tablet according to claim 1, characterized in that, The specific steps for obtaining the first segment of handwriting trajectory range are as follows: S111: Obtain the continuous handwriting data sequence collected by the handwriting tablet terminal device, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, count all sampling points whose pressure parameters are not lower than the contact pressure threshold, and generate the contact pressure screening interval. S112: Based on the contact pressure screening range, detect the coordinates of the sampling point where the pressure parameter is not lower than the contact pressure threshold for the first time, record the horizontal and vertical coordinates of the corresponding sampling point, and generate a pen drop event coordinate record. S113: Based on the coordinate record of the pen stroke event, filter the corresponding sampling point and all subsequent sampling points in the continuous handwriting data sequence, count the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to all sampling points in the interval, obtain the overall trajectory change information, and generate the first segment of the handwriting trajectory interval.

4. The handwriting recognition method for a writing tablet according to claim 1, characterized in that, The specific steps for obtaining the stable marker set for the trajectory segment are as follows: S211: Based on the first segment of handwriting trajectory interval, obtain the pressure parameters of continuous sampling points in the interval, extract the pressure value corresponding to each sampling point in sequence, record the timestamp and horizontal and vertical coordinate information corresponding to each pressure value in the set, and generate a pressure parameter set. S212: Based on the set of pressure parameters, perform pressure difference comparison point by point. For all pressure values ​​in each handwriting segment, extract the maximum pressure value and the minimum pressure value respectively, calculate the difference and obtain the pressure difference judgment value. If it is less than or equal to the pressure jitter tolerance, mark the corresponding handwriting segment as a stable segment and generate a stable segment marking result. S213: Based on the stable segment marking results and combined with the handwriting continuity detection standard, determine the trajectory continuity between adjacent stable segments in turn. If the segments are determined to be continuous, merge the adjacent stable segments and organize them into overall stable marking information to obtain a set of stable handwriting segment markings.

5. The handwriting recognition method for a writing tablet according to claim 1, characterized in that, The specific steps for obtaining the set of handwriting segment recognition adjustment parameters are as follows: S311: Based on the track handwriting segment stable marker set, the original path identification weights of each stable segment are read and grouped, and the pressure fluctuation amplitude between adjacent sampling points is compared to see if it is within the minimum pressure resolution range of the sampling device. If it is greater than the minimum pressure resolution value, the corresponding weight is marked as a weight to be restricted, forming a weight ratio before restriction. S312: Based on the weight ratio before the restriction, compare the normalized pressure coefficient threshold range, compare each ratio with the pressure coefficient interval 0 to 1, if the ratio is less than 0, set it to zero, if the ratio is greater than 1, truncate it to 1, and calculate to obtain the normalized weight adjustment value. S313: Based on the normalized weight adjustment value, the original path recognition weight of the stable segment is replaced in a covering manner, the path recognition weight of each sampling point in the stable segment is updated segment by segment, and the recognition number, normalized pressure coefficient and corresponding updated intensity coefficient of each sampling point are re-recorded to establish a set of handwriting segment recognition adjustment parameters.

6. The handwriting recognition method for a writing tablet according to claim 1, characterized in that, The specific steps for acquiring the pressure-sensitive focusing trajectory data are as follows: S411: Based on the handwriting segment recognition adjustment parameter set, read the intensity coefficient and effective pressure sensitivity threshold in each handwriting segment and compare them item by item. By comparing each intensity coefficient with the threshold point by point, organize the corresponding substandard intensity data according to the handwriting segment number and generate a substandard intensity list. S412: Based on the list of substandard intensity, filter the corresponding coordinate point data and timestamp information, compare the sampling point data of substandard handwriting segments and complete trajectory segments, determine the continuity of the coordinate sequence, distinguish and process discontinuous point sequences, remove handwriting segments with substandard intensity thresholds, and obtain an effective pressure-sensitive filtering sequence. S413: Based on the effective pressure-sensitive filtering sequence, combined with adjacent coordinate points and timestamp data, the trajectory path interpolation is performed on the continuous points in sequence. When performing interpolation, the coordinate distance and time interval of adjacent points are evenly divided to fill in the missing trajectory sampling points and generate pressure-sensitive focusing trajectory data.

7. The handwriting recognition method for a writing tablet according to claim 1, characterized in that, The specific steps for obtaining the handwriting recognition records are as follows: S511: Based on the coordinate sequence in the pressure-sensitive focusing trajectory data, collect the coordinate value and point number of each trajectory point, combine the coordinate values ​​of adjacent points to calculate the horizontal and vertical increments, determine whether the increment is within the reference pixel range, traverse all coordinate point pairs in turn to filter, organize and classify to generate coordinate format conversion values; S512: Based on the coordinate format conversion value, read the time series in the trajectory dataset, perform interval difference calculation on the timestamps of adjacent sampling points in sequence, determine whether the time interval deviates from the standard interval, and if there is a deviation between adjacent time intervals, perform interpolation processing on the time series to form a uniform distribution, and obtain the time resampling sequence. S513: For the time resampling sequence, call the corresponding pressure sensitivity value and handwriting recognition weight for each sampling point, compare the two values ​​in turn, perform difference calculation and combination, combine the time point and coordinate point information to organize the values, and encapsulate the organized value sequence and the number of each trajectory point to generate a handwriting recognition record for the writing tablet.

8. A handwriting recognition system for a writing tablet, characterized in that, The system is used to implement the handwriting recognition method for a writing tablet as described in any one of claims 1-7, comprising: The pen-drop detection module is used to execute S1: acquire continuous handwriting data from the handwriting tablet, record the timestamp, horizontal and vertical coordinates and pressure parameters corresponding to each sampling point, detect the coordinates of the sampling point whose first pressure value is not lower than the contact pressure threshold, use them as the coordinates of the pen-drop event, retain only the subsequent trajectory sequence, and generate the first segment of the handwriting trajectory interval. The pressure analysis module is used to execute S2: read the pressure parameters of continuous sampling points in the first segment of the handwriting trajectory interval, determine whether the maximum and minimum pressure difference in the handwriting segment is lower than the pressure jitter tolerance, perform continuous change judgment to mark stable segments, and generate a set of stable handwriting segment markers. The weight adjustment module is used to execute S3: based on the track handwriting segment stability marker set, the ratio limit is applied to the original path recognition weight of the stable segment, the intensity data is limited to the normalized pressure coefficient, the updated intensity coefficient is used to replace the corresponding recognition weight of the original handwriting segment, and a handwriting segment recognition adjustment parameter set is generated. The trajectory reconstruction module is used to perform S4: based on the handwriting segment recognition adjustment parameter set, remove handwriting segments whose intensity coefficient does not reach the effective pressure sensitivity threshold, and perform handwriting reconstruction to generate pressure-sensitive focusing trajectory data; The recognition output module is used to execute S5: based on the coordinate sequence, time sequence and recognition weight in the pressure-sensitive focusing trajectory data, perform feature format conversion, time resampling and intensity mapping operations respectively, and output the handwriting recognition record after unified encapsulation.

Images