Capacitive touch screen multi-point touch response optimization method based on AI algorithm

Abstract

The invention relates to the technical field of multi-point touch, in particular to a capacitive touch screen multi-point touch response optimization method based on an AI (artificial intelligence) algorithm, which comprises the following steps: acquiring touch point instantaneous capacitance signals, converting the touch point instantaneous capacitance signals into a two-dimensional coordinate sequence, calculating track displacement deviation and local curvature, generating basic track characteristic data, and calculating the basic track characteristic data; the method comprises the following steps: calculating a curvature change rate, identifying abnormal points, extracting an average curvature and a direction parameter, generating track segment characteristic data, inputting the data into a support vector machine to judge an action mode, carrying out clustering analysis to judge stability, generating an action classification stable identifier, and calculating a response weight to generate a multi-point touch response optimization signal. By dynamically analyzing the curvature and the change rate of a touch track, describing touch characteristics in real time and recognizing an action mode, stability is evaluated by utilizing curvature and direction clustering, touch weight is adaptively adjusted, multi-point consistent response is kept, high-precision feedback is provided in rapid sliding and complex interaction, the operation accuracy is improved, and the user experience is improved. The problems of response delay and trajectory drift are solved.

Description

Technical Field

[0001] This invention relates to the field of multi-touch technology, and in particular to a method for optimizing the multi-touch response of capacitive touchscreens based on AI algorithms. Background Technology

[0002] Multi-touch technology encompasses a range of touch technologies capable of simultaneously recognizing multiple contact points, widely used in smartphones, tablets, touchscreens, and other devices. Core aspects of this field include capacitive touch technology, touch point recognition, and touchscreen response optimization. Capacitive touch technology detects the contact of a finger or other conductor through changes in the electric field, thus achieving input recognition. Multi-touch, on the other hand, can simultaneously recognize the position and actions of multiple contact points on the same touchscreen. This technology needs to handle various complex input information, such as finger swipes, clicks, and zooming, while ensuring high accuracy and smoothness. Furthermore, touchscreen response optimization is also a crucial research direction in this field, aiming to improve the sensitivity, stability, and accuracy of touch systems.

[0003] Traditional capacitive touchscreen multi-touch response optimization methods refer to optimizing the response of touch points in multi-touch screens through algorithms to improve the accuracy and smoothness of touch operations. Traditional methods rely on simple signal processing techniques, such as filtering and smoothing touch signals, or identifying and judging touch points based on fixed thresholds. This method mainly relies on traditional hardware circuits and algorithms to achieve basic touchscreen response optimization, but it cannot achieve intelligent adaptive adjustment in complex multi-touch scenarios, resulting in insufficient response accuracy and user experience.

[0004] Existing capacitive multi-touch response optimization methods, based on static thresholds and simple filtering algorithms, cannot effectively cope with dynamically changing touch environments. When touch signals are interfered with or finger movements are rapid, the system struggles to accurately capture subtle changes in the touch trajectory, leading to errors and delays in touch point recognition. In complex touch scenarios, such as multi-finger interaction or rapid swiping, traditional methods cannot distinguish the movement trajectories of different touch points, resulting in trajectory drift or action overlap. Furthermore, they lack adaptive adjustment capabilities and cannot dynamically optimize response strategies based on different touch behaviors, leading to unstable touch data classification and an inability to effectively handle complex touch patterns. Summary of the Invention

[0005] To address the limitations of existing capacitive multi-touch response optimization methods based on static thresholds and simple filtering algorithms, which cannot effectively handle dynamically changing touch environments, this invention provides an AI-based multi-touch response optimization method for capacitive touchscreens. When touch signals are interfered with or finger movements are rapid, the system struggles to accurately capture subtle changes in the touch trajectory, leading to errors and delays in touch point recognition. In complex touch scenarios, such as multi-finger interaction or rapid swiping, traditional methods cannot distinguish the movement trajectories of different touch points, resulting in trajectory drift or overlapping movements. Furthermore, the lack of adaptive adjustment capabilities prevents dynamic optimization of response strategies based on different touch behaviors, leading to unstable touch data classification and an inability to effectively handle complex touch patterns.

[0006] To achieve the above objectives, this invention employs an AI-based multi-touch response optimization method for capacitive touchscreens, comprising the following steps:

[0007] S1: The instantaneous capacitance signal of each touch point is obtained by sampling the touch signal and converted into a two-dimensional touch coordinate sequence. The trajectory displacement deviation is calculated based on the continuous adjacent coordinates, the local curvature values ​​at three consecutive time points are analyzed, and the basic trajectory feature data set is output.

[0008] S2: Based on the basic trajectory feature data set, perform differential operation to calculate the curvature change rate sequence, compare the curvature change rate with a preset curvature change rate threshold, identify the time points where the change rate exceeds the threshold, and extract the average curvature and curvature change direction parameters for the corresponding time period to generate trajectory segmentation feature data.

[0009] S3: Call the trajectory segmentation feature data, input it into the support vector machine classification model, judge the trajectory action features by the continuity of the curvature change direction, combine the average curvature interval to perform action pattern discrimination and classification, and output the action pattern recognition result.

[0010] S4: Based on the action pattern recognition results, filter the corresponding trajectory features, monitor the direction turning angle and curvature amplitude of the touch point, use density clustering algorithm to perform cluster analysis, determine the stability of the action pattern of the touch point, and generate an action classification stability identifier.

[0011] As a further embodiment of the present invention, the basic trajectory feature data set includes displacement deviation value, radius of curvature value and time series data; the trajectory segmentation feature data includes curvature change rate parameter, average curvature value and curvature direction change amount; the action pattern recognition result includes action category label, direction continuity index and classification confidence value; and the action classification stability identifier includes direction turning angle feature, curvature amplitude feature and clustering density parameter.

[0012] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0013] S101: Obtain the instantaneous capacitance signal data frame of each touch point through touch signal sampling, map the capacitance amplitude to the two-dimensional coordinate position according to the row and column scanning index, and serialize and sort the touch point coordinates of the continuous sampling frames according to the timestamp to generate a two-dimensional touch coordinate sequence set.

[0014] S102: Based on the two-dimensional touch coordinate sequence set, perform difference calculation on the x-axis and y-axis components of adjacent coordinate points, weight and accumulate the difference amplitude to form the displacement, construct the displacement deviation sequence according to the time index, and perform smoothing correction processing on the sequence to generate the trajectory displacement deviation sequence.

[0015] S103: Call the trajectory displacement deviation sequence, calculate the cosine of the included angle for the coordinate difference vector of three consecutive time points, and take the inverse cosine to obtain the curvature parameter. Call the touch point sequence identifier in the trajectory displacement deviation sequence for joint mapping, aggregate curvature and displacement information, and generate a basic trajectory feature data set.

[0016] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0017] S201: Based on the basic trajectory feature data set, perform differential operation on the curvature parameters of adjacent sampling points, calculate the rate of curvature change between adjacent time points and arrange them in time order, perform smoothing correction on abnormal mutation values ​​based on the neighborhood mean comparison, and generate curvature change rate sequence data.

[0018] S202: Based on the curvature change rate sequence data, read the change rate value corresponding to the time index in the sequence point by point, compare it with the preset curvature change rate threshold point by point, and when the rate value exceeds the threshold, record the time point index and perform time-series processing to establish a set of time point indexes exceeding the threshold.

[0019] S203: Call the basic trajectory feature data set, extract the corresponding curvature sequence according to the super-threshold time point index set, calculate the average curvature of the time period and determine the direction according to the sign of the rate of change, combine the average curvature and the change direction parameter to generate trajectory segment feature data.

[0020] As a further aspect of the present invention, the preset curvature change rate threshold is dynamically corrected based on the ratio of the standard deviation to the mean of the curvature change rate sequence data.

[0021] The step of extracting the curvature sequence from the curvature change rate sequence data and calculating the average curvature over a time period refers to using a sliding window integration method to weight and average the curvature data within a time period.

[0022] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0023] S301: Call the curvature sequence and time sequence in the trajectory segmentation feature data, calculate the curvature direction vector difference between adjacent sampling points, group adjacent points according to the continuity of direction change, record the index of continuously changing direction sequence, aggregate the direction continuous interval, and generate a curvature direction continuity matrix.

[0024] S302: Based on the curvature direction continuity matrix, extract the curvature change amplitude sequence within the continuous interval, calculate the weighted average of the sequence amplitude, divide the interval into levels according to the numerical distribution characteristics, and generate an average curvature interval mapping set;

[0025] S303: Based on the average curvature interval mapping set, input the support vector machine classification model, perform feature comparison on the directional continuity vector of the curvature interval index, calculate the confidence score, output the category based on the maximum confidence score, and obtain the action pattern recognition result.

[0026] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0027] S401: Based on the action pattern recognition result, retrieve the touch point trajectory sequence data frame, calculate the rate of change of the direction vector angle between adjacent coordinate points, perform differential calculation on the rate of change sequence and compare it with a preset turning threshold, mark the index position that exceeds the threshold, and generate a direction turning angle feature set.

[0028] S402: Call the direction turning angle feature set, calculate the curvature based on the local tangent slope change for the coordinate point data between adjacent turning points, calculate the derivative of curvature with respect to arc length, construct a curvature change matrix, compare the matrix unit with the preset curvature amplitude threshold and extract the difference as a positive index, and generate a curvature amplitude feature set.

[0029] S403: Based on the curvature amplitude feature set and the direction turning angle feature set, perform density clustering on the feature vector group, calculate the stability index according to the spatial density and contour factor of the cluster, extract the cluster identifiers whose cluster stability index is higher than the preset stability benchmark value, and generate action classification stability identifiers.

[0030] As a further aspect of the present invention, the preset turning threshold is determined by dynamic correction based on the statistical results of the standard deviation of the rate of change of the angle between the direction vectors in the pre-sampled touch point trajectory sample set.

[0031] The preset curvature amplitude threshold is determined by removing outliers from the ratio of the median to the range of the rate of change of the curvature derivative during the process of calculating the ratio of the curvature derivative to the arc length based on the local tangent slope.

[0032] The preset stability benchmark value is determined by weighting the average value and variance ratio of the stability index calculated from the spatial density of density clusters and the profile factor.

[0033] As a further aspect of the present invention, the method further includes step S5:

[0034] S5: Call the action classification stability identifier to make a consistency judgment on the action type of the touch point, divide the action group, filter the action cluster whose internal consistency index meets the preset threshold, calculate the touch point response weight coefficient based on different action categories, and generate a multi-touch response optimization signal.

[0035] The multi-touch response optimization signal includes action group identifier, action consistency coefficient, and touch response weight coefficient.

[0036] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0037] S501: Obtain the touch point trajectory dataset of the touch panel, call the action classification stability identifier, perform correlation analysis on the change rate of touch point speed and the change in contact area, aggregate the index of similar touch points according to the action type mapping rules, and generate an action group index matrix;

[0038] S502: Based on the action group index matrix, call the touch point trajectory dataset within the group, calculate the trajectory shape similarity and compare it with the preset action consistency threshold, filter the set of touch points with similarity values ​​higher than the threshold, and generate an action cluster filtering result set;

[0039] The preset action consistency threshold is based on the statistical distribution characteristics of the similarity of touch point trajectory shapes in the touch point trajectory dataset within the group;

[0040] S503: Call the set of touch points in the action cluster filtering result set, calculate the touch point response intensity parameters according to the action category, perform weighted fusion using the category weight parameters, and generate a multi-touch response optimization signal by combining the group dynamic state information;

[0041] The category weight parameter is set based on the priority of different gestures in user interaction.

[0042] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0043] This invention utilizes dynamic analysis of the curvature and rate of change of touch trajectories to capture the characteristics of touch actions in real time and identify different action patterns based on continuity. Cluster analysis of curvature and direction parameters is employed to assess the stability of touch action patterns, effectively filtering unstable or invalid action information. Based on the stability results, the response weights of touch points are adaptively adjusted to maintain response consistency across multiple touch points. This provides high-precision touch feedback during rapid swiping or complex interactions, improving the accuracy and smoothness of touch operations and resolving the response latency and trajectory drift problems in existing technologies. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the accompanying drawings without creative effort.

[0045] Figure 1 This is a schematic diagram of the steps of the present invention;

[0046] Figure 2 This is a detailed schematic diagram of S1 of the present invention;

[0047] Figure 3 This is a detailed schematic diagram of S2 of the present invention;

[0048] Figure 4 This is a detailed schematic diagram of S3 of the present invention;

[0049] Figure 5 This is a detailed schematic diagram of S4 of the present invention;

[0050] Figure 6 This is a detailed schematic diagram of S5 of the present invention. Detailed Implementation

[0051] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0052] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0053] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0054] In this embodiment of the invention, sometimes the subscript such as W1 is written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0055] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0056] Please see Figure 1 This invention provides a method for optimizing multi-touch response of capacitive touchscreens based on AI algorithms, including the following steps:

[0057] S1: The instantaneous capacitance signal of each touch point is obtained by sampling the touch signal and converted into a two-dimensional touch coordinate sequence. The trajectory displacement deviation is calculated based on the continuous adjacent coordinates, the local curvature values ​​at three consecutive time points are analyzed, and the basic trajectory feature data set is output.

[0058] S2: Based on the basic trajectory feature data set, perform differential operation to calculate the curvature change rate sequence, compare the curvature change rate with the preset curvature change rate threshold, identify the time points where the change rate exceeds the threshold, and extract the average curvature and curvature change direction parameters of the corresponding time period to generate trajectory segmentation feature data.

[0059] S3: Call the trajectory segment feature data, input it into the support vector machine classification model, judge the trajectory action features by the continuity of the curvature change direction, combine the average curvature interval to perform action pattern discrimination and classification, and output the action pattern recognition result.

[0060] S4: Based on the action pattern recognition results, filter the corresponding trajectory features, monitor the direction turning angle and curvature amplitude of the touch point, use density clustering algorithm to perform cluster analysis, judge the stability of the action pattern of the touch point, and generate action classification stability identifiers;

[0061] S5: Call the motion classification stability identifier to make a consistency judgment on the motion type of the touch point, divide the motion group, filter the motion clusters whose internal consistency index meets the preset threshold, calculate the touch point response weight coefficient based on different motion categories, and generate a multi-touch response optimization signal.

[0062] The basic trajectory feature dataset includes displacement deviation values, radius of curvature values, and time series data. The trajectory segmentation feature data includes curvature change rate parameters, average curvature values, and curvature direction change amounts. The action pattern recognition results include action category labels, direction continuity indicators, and classification confidence values. The action classification stability identifiers include direction turning angle features, curvature amplitude features, and clustering density parameters. The multi-touch response optimization signals include action group identifiers, action consistency coefficients, and touch response weight coefficients.

[0063] Please see Figure 2 The specific steps of S1 are as follows:

[0064] S101: Obtain the instantaneous capacitance signal data frame of each touch point through touch signal sampling, map the capacitance amplitude to the two-dimensional coordinate position according to the row and column scanning index, and serialize and sort the touch point coordinates of the continuous sampling frames according to the timestamp to generate a two-dimensional touch coordinate sequence set.

[0065] Based on the instantaneous capacitance signal data frames of each touch point obtained through touch signal sampling, the touch screen sensor array is first scanned. This array consists of, for example, lines and The columns of cross electrodes are configured to, in order to The frequency of Hertz is used, that is, every A complete frame of capacitance signal data is generated in milliseconds; this data frame is one... A matrix where each element represents the capacitance value at the intersection of the corresponding row and column, for example, at a timestamp. At that time, in the collected data frame, the index position The capacitance amplitude at that location is the background value. fF, and due to finger touch, index position The capacitance amplitude at that point increased significantly to Next, based on the row and column scan index, the capacitance value is mapped to a two-dimensional coordinate position. Specifically, a capacitance threshold is set to distinguish between effective touch and noise. This threshold is set based on the mean and standard deviation of the background capacitance. For example, if the mean capacitance when the screen is unloaded is fF... fF, standard deviation is If fF is the threshold, then the threshold can be set as the mean plus three times the standard deviation, i.e. fF, any capacitance value exceeding The entire fF region is identified as a touch area, and the center of the weighted average of capacitance values ​​exceeding a threshold within that region is calculated, thus obtaining a sub-pixel precision two-dimensional coordinate, such as a coordinate point. At the same time, append the current timestamp to the coordinate point. In consecutive sampling frames, for example in At that time, the system detected a new coordinate point. and a new touch point At this point, the touch point coordinates of the continuously sampled frames need to be serialized and sorted according to the timestamp. Specifically, the Euclidean distance between each point in the current frame and the point in the previous frame is calculated, and point pairs with a distance less than a preset tracking threshold are associated. This tracking threshold is set according to the screen resolution and the maximum expected scrolling speed. For example, for a single touch point... The screen resolution can be set to... Pixels, due to and The distance (approximately) (pixels) much smaller Therefore, Belonging to The touch sequence in which it is located, and Since there are no neighboring points in the previous frame, it is identified as a new touch sequence. Through this process, the original, discrete touch point data is organized into multiple temporally ordered sequences of two-dimensional coordinate points, distinguished by touch events (finger), for example, forming sequence one: and sequence two: This is the set of two-dimensional touch coordinate sequences.

[0066] S102: Based on the two-dimensional touch coordinate sequence set, the difference between the x-axis and y-axis components of adjacent coordinate points is calculated, the difference magnitude is weighted and accumulated to form the displacement, the displacement deviation sequence is constructed according to the time index, and the sequence is smoothed and corrected to generate the trajectory displacement deviation sequence.

[0067] The generated sequence one is called, which contains timestamps. to Five consecutive coordinate points, with specific data shown in Table 1. First, the difference between the x-axis and y-axis components of adjacent coordinate points is calculated to obtain the displacement vector within each time interval. For example, calculating... The displacement vector at time step needs to be called... Coordinates of time and coordinates of time The difference between the x-axis components is The difference in the y-axis components is ,get The displacement vector at time t is Similarly, calculate The displacement vector at time t is , Time for , Time for Next, these differences are weighted and summed to form the displacement, and a displacement deviation sequence is constructed based on the time index. This displacement deviation sequence is formed by arranging the displacement vectors obtained above in chronological order. Then, a smoothing correction is performed on the sequence, using a window size of [size missing]. The moving average filter corrects each vector in the sequence. displacement vector at time t The correction process involves taking the previous moment... vector Current moment vector And the next moment vector The corresponding components are added together and the average value is taken. The corrected x-component is: The corrected y-component is Therefore The displacement vector after time correction is For sequence endpoints, such as At any given time, an asymmetric window can be used, for example, by using only... and The vectors at time points are averaged, and the corrected x-components are: The y-component is The corrected displacement vector is obtained. By performing this smoothing operation on the vectors in the entire displacement deviation sequence, the final trajectory displacement deviation sequence is generated.

[0068] Table 1: Example of Touch Point Coordinate Sequence | Timestamp

[0069]

[0070] As shown in Table 1, this table lists the sequence of coordinate points in a continuous touch event, where the timestamp interval is... millisecond.

[0071] S103: Call the trajectory displacement deviation sequence, calculate the cosine of the included angle for the coordinate difference vector of three consecutive time points, and take the inverse cosine to obtain the curvature parameter. Call the touch point sequence identifier in the trajectory displacement deviation sequence for joint mapping, aggregate curvature and displacement information, and generate a basic trajectory feature data set.

[0072] Call a portion of the generated corrected sequence data, let The correction vector at time step is , Time for , Time for For the coordinate difference vectors at three consecutive time points, i.e., two consecutive corrected displacement vectors, the cosine of the included angle is calculated, and the inverse cosine is taken to obtain the curvature parameter, in order to calculate... Taking the curvature at time as an example, it is necessary to use vector of time Compared to the previous moment vector First, calculate the dot product of the two vectors. Then calculate the magnitudes of the two vectors, i.e., their Euclidean norms. The modulus length is ,vector The modulus length is Next, calculate the cosine of the included angle. Due to calculation errors, the cosine value is clamped to... Within the interval, take as Finally, taking the inverse cosine of this cosine value yields the curvature parameter, i.e., the included angle is 0 degrees. The curvature parameter of degrees is represented in At any given moment, the touch trajectory does not change direction and maintains a straight line motion. Subsequently, the touch point sequence identifier in the trajectory displacement deviation sequence is called, for example, the currently processed sequence identifier is "Sequence One", and a joint mapping is performed to calculate the curvature parameters. Degree and current time point This is associated with the touch point sequence identifier "Sequence One," and curvature and displacement information are aggregated, specifically, for time points. A touch event creates a data record containing a sequence identifier "Sequence 1" and a timestamp. Corrected displacement vector Displacement amplitude Pixel and curvature parameters The system combines these data records calculated at each time point to generate a basic trajectory feature data set.

[0073] Please see Figure 3 The specific steps of S2 are as follows:

[0074] S201: Based on the basic trajectory feature data set, perform differential operation on the curvature parameters of adjacent sampling points, calculate the rate of curvature change between adjacent time points and arrange them in time order, perform smoothing correction on abnormal abrupt values ​​based on the neighborhood mean comparison, and generate curvature change rate sequence data.

[0075] Calling the continuous timestamps to The curvature parameter sequence, which is First, a difference operation is performed on the curvature parameters of adjacent sampling points to calculate the rate of change of curvature between adjacent time points, where the sampling time interval is fixed at 1. Milliseconds, for example, calculation Rate of change at time, using Curvature of time Degree and Curvature of time The difference is degrees. Divide by the time interval Seconds, the rate of change of curvature is obtained. degrees / second; similarly, calculate the rate of change at subsequent time points to obtain an original rate of change sequence: degrees / second, the sequence is in chronological order. to The sequence is then arranged, and then smoothed by comparing the abnormal mutation values ​​in the sequence with the mean of their neighbors. A value is identified as an abnormal mutation based on whether its absolute value is greater than the mean of the absolute values ​​of its two nearest neighbors. Times, for example, for rate of change at time degrees / second, its neighborhood value Moment degrees / second and Moment degrees / second, neighborhood mean is degrees / second, with the multiplier set to [value]. The judgment criteria are degrees / second, due to Therefore, this value is identified as an anomalous abrupt change. The smoothing correction operation involves replacing this anomalous value with the arithmetic mean of its two nearest neighbors, i.e., using... Replace the original with degrees / second degrees / second, while for the remaining values ​​in the sequence, such as Moment degrees / second, its neighborhood mean is degrees / second, the judgment standard is degrees / second, due to This value is not considered abnormal and is retained. After performing this check and correction process on the entire sequence, the curvature change rate sequence data is finally generated.

[0076] S202: Based on the curvature change rate sequence data, read the change rate value corresponding to the time index in the sequence point by point, compare it with the preset curvature change rate threshold point by point, and record the time point index when the rate value exceeds the threshold and perform time-series processing to establish a set of time point indexes exceeding the threshold.

[0077] Based on the corrected sequence, for clarity, a new set of scene data is introduced here, as shown in Table 2. First, the rate of change value corresponding to the time index in the sequence is read point by point, for example, from... Rate value at time Starting at degrees per second, a point-by-point comparison is performed with a dynamically adjusted curvature change rate threshold. The process for setting the curvature change rate threshold is as follows: First, based on the entire curvature change rate sequence data, i.e., the "Curvature Change Rate" column in Table 2, its standard deviation and mean are calculated. This sequence is... Its mean for degrees per second, and its standard deviation for Degrees per second, the dynamic adjustment of the threshold is based on a base threshold and a correction term consisting of the ratio of the standard deviation to the mean. Based on experience, for example, for scenarios involving rapid gesture operations, it can be set to... degrees / second, correction factor Set as This coefficient is used to adjust the degree of influence of the standard deviation on the threshold, and the final threshold... The calculation method is to add the absolute value of the mean to the base threshold, and then add the product of the standard deviation and the correction factor, i.e. Substitute the values, degrees / second, compare this threshold with the absolute value of each rate value in the sequence, for example No record. Not recorded, but Then record its time point index. Similarly, Record its time point index The following and If all values ​​are less than the threshold, they are not recorded. Finally, the recorded time point indices are arranged in chronological order to create a set of time point indices exceeding the threshold.

[0078] Table 2: Examples of Touch Trajectory Feature Representation

[0079]

[0080] As shown in Table 2, this table displays the curvature calculated from consecutive sampling points and its rate of change after first-order difference in a simulated user touch operation drawing a sharp corner, with time intervals of [missing information]. millisecond.

[0081] S203: Call the basic trajectory feature data set, extract the corresponding curvature sequence according to the index set of time points exceeding the threshold, calculate the average curvature of the time period and determine the direction according to the sign of the rate of change, combine the average curvature and the change direction parameter to generate trajectory segmentation feature data;

[0082] For the established set of indexes of time points exceeding the threshold The time point serves as the dividing point of the trajectory, dividing the entire trajectory into multiple time periods. The first time period is from the start to the first threshold point. Previously, that is to The second time period is to Previously, only included The third time period is Until the next threshold point or the end of the sequence, i.e. to Next, the corresponding time period range is called, and the curvature sequence for each time period is extracted from the generated curvature change rate sequence data or the original data as shown in Table 2. For the first time period ( to ), and its corresponding curvature sequence is Degrees, and then calculate the average curvature over that time period, i.e. Degree, and based on the end of that time period, i.e., the dividing point. The sign of the rate of change of curvature determines the direction. At time t, the rate of curvature change is Degrees per second, with a positive sign, indicates that the curvature is increasing. This directional parameter is quantized as... Then, the calculated average curvature Degree and direction of change parameters By combining these data, segmented feature data of the first trajectory segment is formed, i.e. For the third time period ( to Its curvature sequence is degrees, mean curvature The starting point of this segment is... Its rate of change is negative ( (degrees / second), representing a decreasing trend in curvature after this point (relative to the previous drastic change). Here, we focus on the overall trend of this segment and observe... The subsequent velocity signs are all positive, therefore the direction parameter remains the same. The combined data yields the feature data of the third segment. The segmented feature data is collected together to finally generate trajectory segment feature data.

[0083] Please see Figure 4 The specific steps of S3 are as follows:

[0084] S301: Call the curvature sequence and time series in the trajectory segmentation feature data, calculate the curvature direction vector difference between adjacent sampling points, group adjacent points according to the continuity of direction change, record the index of continuously changing direction sequence, aggregate the continuous direction intervals, and generate a curvature direction continuity matrix.

[0085] The generated trajectory segmentation feature data is invoked, such as a sequence describing a user drawing a sinusoidal waveform gesture on the screen. This sequence contains seven consecutive segmented features, whose direction parameters (i.e., the sign of the rate of change of curvature) are as follows: This sequence is the curvature direction vector. First, the curvature direction vector difference between adjacent segments is calculated. Specifically, starting from the second segment, the direction parameter of the current segment is subtracted from the direction parameter of the previous segment. For the sequence... Calculate the difference sequence ,in Substituting the values, the difference sequence is: The calculation result is Next, adjacent points are grouped according to the degree of continuity of directional change. The grouping rule is that when the difference value is... When the difference value is not changed, it indicates that the direction has not changed, and the current segment and the previous segment belong to the same continuous interval in the same direction. When the direction is reversed, this indicates that the interval boundary has been reached. The sequence is then scanned accordingly. Therefore, the index and The segments belong to the same group. , indicating in the index The direction changes at the index, therefore the first interval is at the index. End of section ,index and Belongs to a new group. , indicating in the index The direction changes at point 1, and the second interval is at index 2. End of section and ,index , , Belonging to the third group, the subsequent records record the continuously changing direction sequence index, i.e., the first group contains the index. The second group contains indexes. The third group contains indexes. Then, aggregate continuous directional intervals, and associate the index intervals with the common directional symbols within those intervals. The directional symbols of the first group are... The second group is The third group is After integrating the information, a curvature direction continuity matrix is ​​generated.

[0086] S302: Based on the curvature direction continuity matrix, extract the curvature change amplitude sequence within a continuous interval, calculate the weighted average of the sequence amplitudes, divide the intervals hierarchically according to the numerical distribution characteristics, and generate an average curvature interval mapping set;

[0087] Based on the generated curvature direction continuity matrix, which records three directional continuity intervals, namely interval one (index) ,direction Interval 2 (index) ,direction ) and interval three (index) ,direction First, extract the curvature change amplitude sequence corresponding to each continuous interval. This requires calling the piecewise average curvature value. Assuming the average curvature sequence corresponding to the direction sequence is... If the degree is given, then the amplitude sequence of interval one is: The amplitude sequence of interval two is The amplitude sequence of interval three is as follows Next, a weighted average of the magnitudes of each sequence is calculated, with the weights increasing with the time index to give greater emphasis to the characteristics of the later stages of the trajectory. The weighting coefficients are... By location index The relative position in the sequence is determined, for a sequence of length . The sequence, the first elements ( from Weight of the start) Set as The weighted average is calculated by summing the products of each element and its corresponding weight, divided by the sum of the weights. For interval one, the weighted average is... For interval two, the weighted average is... For interval three, the weighted average is... The degree is then used to classify these intervals hierarchically based on their numerical distribution characteristics. The classification criteria are pre-set, with the average curvature value being lower than [a certain value]. The degree interval is defined as the "low curvature" level (coded as...). ), between Degree and The degree is defined as the "medium curvature" level (coded as...). ), higher than or equal to The degree is defined as the "high curvature" level (coded as...). According to this standard, Degree belongs to Hierarchy Degree belongs to Hierarchy Degree belongs to The hierarchy maps the intervals to their corresponding hierarchy codes, generating a set of average curvature interval mappings.

[0088] S303: Based on the average curvature interval mapping set, input the support vector machine classification model, perform feature comparison on the direction continuity vector of the curvature interval index, calculate the confidence score, output the category based on the maximum confidence score, and obtain the action pattern recognition result;

[0089] Based on the mean curvature interval mapping set, its content is interval one mapped to the level. Interval 2 is mapped to the hierarchy. Interval 3 is mapped to the hierarchy. At the same time, combined with the direction sign of the interval The information is combined into a fixed-length feature vector, which is then input into a pre-trained classification function. This function makes a judgment based on the distance between the input feature vector and a preset category template. The feature vector is constructed as follows, setting the gesture to contain a maximum of... A series of consecutive directional intervals, each interval consisting of... directional symbols and Each level of state (using one-hot encoding) for , for , for Composed of ) with a total dimension of For the three intervals in this example, their vector representations are as follows: , , Concatenate these three vectors sequentially, and then use the remaining two unused interval positions... Fill, the final form 3D feature vectors are Next, this feature vector is input into a classification function, which internally stores feature templates for standard action patterns such as "straight slide," "circular arc," and "Z-shaped slide." Feature comparison is then performed, which calculates the inner product of the input vector and each template vector to determine the confidence score for each category. The hypothesis function returns a set of scores after calculation: {"straight slide":} "Arc": "Z-shaped lines": "Click": Then, based on the maximum confidence level, the category is output. In this group of values, the maximum value is... The corresponding category is "Z-shaped lines". Therefore, "Z-shaped lines" are used as the final recognition result output to obtain the action pattern recognition result.

[0090] Please see Figure 5 The specific steps of S4 are as follows:

[0091] S401: Based on the action pattern recognition results, retrieve the touch point trajectory sequence data frame, calculate the rate of change of the direction vector angle between adjacent coordinate points, perform differential calculation on the rate of change sequence and compare it with the preset turning threshold, mark the index position that exceeds the threshold, and generate a direction turning angle feature set.

[0092] Based on the action pattern recognition results, i.e., the obtained "Z-shaped swipe" category, the original touch point trajectory sequence data frame corresponding to this gesture is first retrieved, for example, a segment containing A sequence of coordinate points This sequence represents the process of a user drawing a zigzag shape on the screen, followed by calculating the rate of change of the angle between the direction vectors of adjacent coordinate points. Specifically, for each point in the sequence... (in from arrive Construct two vectors: the forward vector and the backward vector. and backward vector Then calculate the angle between the two vectors, for example =(30, 10), , ,but , The included angle between Degree, if , , ,but (This is a stationary point, ignore it), or if , , ,but , included angle for degrees, and the included angle sequence is obtained through this method. Then, the difference function is performed on the included angle sequence to obtain the rate of change sequence, i.e. For example, if the included angle sequence is The degree, then the rate of change near the inflection point is value , , , The peak value of this rate of change sequence Valley value The sharp turns in the corresponding trajectories are then compared with the absolute values ​​of these rates of change to a dynamically adjusted preset turning threshold. The determination of this threshold involves first collecting data containing... A pre-sampled set of touch point trajectories for various gestures (straight lines, curves, corners, etc.) is used. The rate of change of the angle between the direction vectors of the points in the samples is calculated, forming a large dataset of rate of change. This dataset is then statistically analyzed, and its standard deviation is calculated. Assuming the obtained standard deviation is... degrees / frame, inflection threshold The setting is the sum of the baseline value and the standard deviation correction term, the baseline value Set as degrees / frame, correction factor Set as ,but degrees / frame, the absolute value of the calculated rate of change With this threshold Compare degrees / frames one by one, for example No marking Mark its corresponding index position (e.g.) ), Mark its corresponding index position (e.g.) By traversing the entire rate of change sequence, the index positions that exceed the threshold are finally obtained, and a set of directional turning angle features is generated.

[0093] S402: Call the direction turning angle feature set, calculate the curvature based on the local tangent slope change for the coordinate point data between adjacent turning points, calculate the derivative of curvature with respect to arc length, construct the curvature change matrix, compare the matrix unit with the preset curvature amplitude threshold and extract the difference as the positive index, and generate the curvature amplitude feature set.

[0094] Call the generated direction turning angle feature set, for example, containing two turning point indices. The original trajectory is divided into three sub-segments, and then the coordinate point data between adjacent turning points, for example, the index from... arrive In this section, the slope of the local tangent is calculated to obtain the ratio of the derivative of curvature to the arc length. This ratio can be approximated as the rate at which curvature changes with arc length. For each point within this segment First, calculate its approximate curvature. Point of use The reciprocal of the radius of the circumcircle of the triangle formed, for example in Point, coordinates are Its preceding and following points are and The curvature was calculated. Pixels -1 , then calculate Curvature of a point Pixels -1 The arc length between two points Approximate Euclidean distance pixels, then Pointed The value is approximately Pixels -2 For the points within this segment ( to Repeat this process to build a system composed of these... The sequence of values ​​is the curvature variation matrix, for example, the sequence is... Next, each value in this sequence is compared to a preset curvature amplitude threshold, which is set as follows: first, a value containing... A sample library of different trajectory segments is used to calculate the trajectory for each segment. For each sequence, outlier removal is performed, i.e., removing any elements that exceed the definition based on the interquartile range. The range of values ​​is then used to calculate the midpoint of each processed sequence. and range (Maximum value minus minimum value), and calculate the ratio. Assuming for Calculated from samples The average of the values ​​is Then the curvature amplitude threshold Set as this average value times, that is However, due to the calculation The value is very small, so we adjust the parameters here and set the average value to be... The threshold coefficient is ,but Pixels -2 Then, each value in the sequence is compared with the threshold, and the indices of values ​​with positive differences (i.e., values ​​greater than the threshold) are extracted. greater than There are two values, with indices of and respectively. and The indexes are collected to ultimately generate a curvature amplitude feature set.

[0095] S403: Based on the curvature amplitude feature set and the direction turning angle feature set, perform density clustering on the feature vector group, calculate the stability index based on the spatial density and contour factor of the cluster, extract the cluster identifiers whose cluster stability index is higher than the preset stability benchmark value, and generate action classification stability identifiers.

[0096] Based on the generated directional turning angle feature set and curvature amplitude feature set, a feature vector is first constructed for each gesture trajectory. For example, the feature vector of a gesture can be composed of the number of turning angles it contains, the size of the first turning angle, the size of the second turning angle, and the number of curvature amplitude points in the segment between the turning points. For the aforementioned "Z-shaped stroke" gesture, its feature vector is... ,in It is the number of turning points. and It's the change in angle at the turning point. This is the number of curvature amplitude points in the middle segment, assuming they have already been collected. This set of feature vectors for each gesture, then... Density clustering is performed using four-dimensional vectors. Specifically, the clustering neighborhood radius is set. for (This value is determined by analyzing the k-distance graph of the distances between vector pairs to find the "inflection point"), setting the minimum number of samples to form the core point. for Then, iterate through each vector point and calculate its... If the number of points in the neighborhood is greater than or equal to If a point is found to be a core point, it is marked as such, and the cluster is expanded from that point, incorporating its neighboring points into the same cluster. This process is repeated recursively with newly added points until the cluster can no longer be expanded. Finally, the cluster is... The points are divided into several clusters (e.g., (One cluster and one noise set), and then, based on the spatial density and contour factor of the clusters, a stability index is calculated for a given set. Clusters of points Its spatial density is calculated by dividing the number of points by the volume of the smallest boundary hyperrectangle formed by the points within the cluster, let the volume be... The density is Its contour factor is obtained by averaging the contour coefficients of each point within the cluster. Assuming the cluster... The average contour factor is Stability indicators Obtained through weighted combination Then, cluster identifiers with stability indices higher than a preset stability benchmark value are extracted. The determination process is as follows: first calculate The stability index of each cluster is assumed to be... Calculate its average value and variance Calculate the ratio of the two. Then, through a pre-trained linear regression model After correction, the following results were obtained. The stability index of each cluster is compared with this benchmark value. , , Therefore, cluster and It is considered stable, and its identifier is... Extract the data and generate stable identifiers for action classification.

[0097] Please see Figure 6 The specific steps of S5 are as follows:

[0098] S501: Obtain the touch point trajectory dataset of the touch panel, call the action classification stability identifier, perform correlation analysis on the change rate of touch point speed and the change in contact area, aggregate the index of similar touch points according to the action type mapping rules, and generate an action group index matrix.

[0099] Obtain the dataset of active touch point trajectories on the current touch panel, for example, at a given time point. At that time, there are five touch points on the panel simultaneously, and their indices are as follows: Each touch point is associated with a generated action classification stability identifier. Assuming that, based on previous calculations, the touch point... The identifier is empty (noise). and The identifier is (This represents a stable two-finger zoom action.) The identifier is (This represents a stable single-finger dragging action.) The identifier is also empty. Then, these action classification stability identifiers are called, and the rate of change of speed and the change of contact area of ​​touch points with the same non-empty identifier are compared. First, these two parameters are calculated for each point, as shown in Table 3. For example, it is in The speed at time is pixels per second, in Time for pixels / second, sampling interval is If the velocity changes in milliseconds, then its rate of change is... pixels / second 2 Its contact area is from Pixels 2 Become Pixels 2 The change is Pixels 2 Then, based on the action type mapping rule, the indexes of similar touch points are aggregated. This rule is a preset set of conditions, for example, rule one: "If the identifier is..." (Scaling) then finds two points with the same identifier whose rates of change of velocity are roughly opposite in direction (the dot product of the vectors is less than 1). ), and the absolute value of the change in contact area is less than Pixels 2 Rule 2: "If the identifier is..." (Drag), then that point becomes an independent group. According to rule one, check... and The identifiers are all Assuming The direction of the rate of change of velocity is Conversely, the change in the contact area between the two ( for , for The absolute values ​​are all less than If the condition is met, the indexes will be aggregated into a group. According to rule two, The identifier is Therefore, it forms its own group. , and Because the identifier was empty, it was categorized into an "unclassified" group. The groups are integrated to generate an action group index matrix.

[0100] Table 3: Examples of Touch Point Status Parameters

[0101]

[0102] As shown in Table 3, this table lists the key dynamic parameters of the five active touch points on the screen at a given moment and their pre-classified action identifiers.

[0103] S502: Based on the action group index matrix, call the touch point trajectory dataset within the group, calculate the trajectory shape similarity and compare it with the preset action consistency threshold, filter the set of touch points with similarity values ​​higher than the threshold, and generate an action cluster filtering result set;

[0104] The action consistency threshold is based on the statistical distribution characteristics of the similarity of touch point trajectory shapes in the group's touch point trajectory dataset;

[0105] Based on the generated action group index matrix, for example, calling the group representing two-finger zoom. First, call the touch points within the group. and In recent The dataset consists of a sequence of coordinate points within milliseconds. The morphological similarity between these two trajectories is then calculated. This calculation involves aligning the coordinate point sequences of the two trajectories and summing the Euclidean distances between the aligned point pairs. A smaller sum indicates a closer similarity in morphology. To convert the distances into... The similarity values ​​of the intervals are transformed using a conversion operation. ,in It is the sum of the calculated distances. It is the average length of the two trajectories, assuming and The trajectories represent typical two-finger outward expansion movements, and the calculated distances and... for Average trajectory length for Pixels, then similarity Next, the similarity value is compared with a preset action consistency threshold. This threshold is set based on the statistical distribution characteristics of the similarity of touch point trajectory patterns within a group. Specifically, it is assumed that a group contains three points. Then calculate the pairwise similarity of the three groups. Assuming the value is Calculate the mean of this similarity set. and standard deviation Action consistency threshold The correction factor is set to the product of the mean minus a correction factor and the standard deviation. Set as ,but For groups with only two points If a distribution cannot be formed, a global default threshold based on the statistics of the original samples is used. This default value is set to... Then compare them. ,show and The actions are consistent enough that the touch point set is appropriate. If the similarity is below a threshold, the set will be discarded as inconsistent actions. By performing this filtering process on action groups, the final action cluster filtering result set is generated.

[0106] S503: Call the set of touch points in the action cluster filtering result set, calculate the touch point response intensity parameters according to the action category, perform weighted fusion using the category weight parameters, and combine the group dynamic state information to generate a multi-touch response optimization signal;

[0107] The generated action cluster is used to filter the result set, for example, to include a confirmed and consistent scaling action cluster. and a drag-and-drop action cluster First, the touch point response intensity parameters are calculated based on the action category of each cluster. For scaling clusters... Its response intensity parameter is defined as the rate of change of the distance between two touch points, assuming that in The distance between the two points at time is Pixels Time becomes pixels, sampling interval is milliseconds, then intensity parameter pixels per second, for draggable clusters Its intensity parameter is defined as the moving speed of the touch point, assuming its speed is... pixels per second, i.e. Pixels per second. Next, the intensity parameters are weighted and fused using preset category weight parameters. These category weight parameters are set based on the priority of different gestures in user interaction. For example, in map applications, zooming should be more responsive than dragging, so a weight can be set accordingly. , Click operation weight These weights were adjusted based on extensive user testing feedback. The fusion process involves multiplying the strength of each cluster by its corresponding weight to obtain the weighted strength. , Then, it combines the group's dynamic state information, which describes the group's lifecycle stages, for example, using a dynamic factor. This indicates that the newly formed group (those that have existed for less than [time period]) The millisecond factor is Stable groups (existing) (milliseconds) The groups that are about to disappear are Assuming the scaling cluster has just formed, its dynamic factor is The dragged cluster is stable, with a factor of [value missing]. Finally, the weighted intensity is multiplied by the dynamic factor to obtain the final signal value. , The signal value is packaged with its corresponding action category and touch point index to generate a multi-touch response optimization signal.

[0108] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for optimizing multi-touch response of capacitive touchscreens based on AI algorithms, characterized in that, Includes the following steps: S1: The instantaneous capacitance signal of each touch point is obtained by sampling the touch signal and converted into a two-dimensional touch coordinate sequence. The trajectory displacement deviation is calculated based on the continuous adjacent coordinates, the local curvature values ​​at three consecutive time points are analyzed, and the basic trajectory feature data set is output. S2: Based on the basic trajectory feature data set, perform differential operation to calculate the curvature change rate sequence, compare the curvature change rate with a preset curvature change rate threshold, identify the time points where the change rate exceeds the threshold, and extract the average curvature and curvature change direction parameters for the corresponding time period to generate trajectory segmentation feature data. S3: Call the trajectory segmentation feature data, input it into the support vector machine classification model, judge the trajectory action features by the continuity of the curvature change direction, combine the average curvature interval to perform action pattern discrimination and classification, and output the action pattern recognition result. S4: Based on the action pattern recognition results, filter the corresponding trajectory features, monitor the direction turning angle and curvature amplitude of the touch point, use density clustering algorithm to perform cluster analysis, determine the stability of the action pattern of the touch point, and generate an action classification stability identifier.

2. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 1, characterized in that, The basic trajectory feature data set includes displacement deviation value, radius of curvature value and time series data. The trajectory segmentation feature data includes curvature change rate parameter, average curvature value and curvature direction change amount. The action pattern recognition result includes action category label, direction continuity index and classification confidence value. The action classification stability identifier includes direction turning angle feature, curvature amplitude feature and clustering density parameter.

3. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Obtain the instantaneous capacitance signal data frame of each touch point through touch signal sampling, map the capacitance amplitude to the two-dimensional coordinate position according to the row and column scanning index, and serialize and sort the touch point coordinates of the continuous sampling frames according to the timestamp to generate a two-dimensional touch coordinate sequence set. S102: Based on the two-dimensional touch coordinate sequence set, perform difference calculation on the x-axis and y-axis components of adjacent coordinate points, weight and accumulate the difference amplitude to form the displacement, construct the displacement deviation sequence according to the time index, and perform smoothing correction processing on the sequence to generate the trajectory displacement deviation sequence. S103: Call the trajectory displacement deviation sequence, calculate the cosine of the included angle for the coordinate difference vector of three consecutive time points, and take the inverse cosine to obtain the curvature parameter. Call the touch point sequence identifier in the trajectory displacement deviation sequence for joint mapping, aggregate curvature and displacement information, and generate a basic trajectory feature data set.

4. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 3, characterized in that, The specific steps of S2 are as follows: S201: Based on the basic trajectory feature data set, perform differential operation on the curvature parameters of adjacent sampling points, calculate the rate of curvature change between adjacent time points and arrange them in time order, perform smoothing correction on abnormal mutation values ​​based on the neighborhood mean comparison, and generate curvature change rate sequence data. S202: Based on the curvature change rate sequence data, read the change rate value corresponding to the time index in the sequence point by point, compare it with the preset curvature change rate threshold point by point, and when the rate value exceeds the threshold, record the time point index and perform time-series processing to establish a set of time point indexes exceeding the threshold. S203: Call the basic trajectory feature data set, extract the corresponding curvature sequence according to the super-threshold time point index set, calculate the average curvature of the time period and determine the direction according to the sign of the rate of change, combine the average curvature and the change direction parameter to generate trajectory segment feature data.

5. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 4, characterized in that, The preset curvature change rate threshold is dynamically corrected based on the ratio of the standard deviation to the mean of the curvature change rate sequence data; The step of extracting the curvature sequence from the curvature change rate sequence data and calculating the average curvature over a time period refers to using a sliding window integration method to weight and average the curvature data within a time period.

6. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Call the curvature sequence and time sequence in the trajectory segmentation feature data, calculate the curvature direction vector difference between adjacent sampling points, group adjacent points according to the continuity of direction change, record the index of continuously changing direction sequence, aggregate the direction continuous interval, and generate a curvature direction continuity matrix. S302: Based on the curvature direction continuity matrix, extract the curvature change amplitude sequence within the continuous interval, calculate the weighted average of the sequence amplitude, divide the interval into levels according to the numerical distribution characteristics, and generate an average curvature interval mapping set; S303: Based on the average curvature interval mapping set, input the support vector machine classification model, perform feature comparison on the directional continuity vector of the curvature interval index, calculate the confidence score, output the category based on the maximum confidence score, and obtain the action pattern recognition result.

7. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 6, characterized in that, The specific steps of S4 are as follows: S401: Based on the action pattern recognition result, retrieve the touch point trajectory sequence data frame, calculate the rate of change of the direction vector angle between adjacent coordinate points, perform differential calculation on the rate of change sequence and compare it with a preset turning threshold, mark the index position that exceeds the threshold, and generate a direction turning angle feature set. S402: Call the direction turning angle feature set, calculate the curvature based on the local tangent slope change for the coordinate point data between adjacent turning points, calculate the derivative of curvature with respect to arc length, construct a curvature change matrix, compare the matrix unit with the preset curvature amplitude threshold and extract the difference as a positive index, and generate a curvature amplitude feature set. S403: Based on the curvature amplitude feature set and the direction turning angle feature set, perform density clustering on the feature vector group, calculate the stability index according to the spatial density and contour factor of the cluster, extract the cluster identifiers whose cluster stability index is higher than the preset stability benchmark value, and generate action classification stability identifiers.

8. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 7, characterized in that, The preset turning threshold is determined by dynamic correction based on the statistical results of the standard deviation of the rate of change of the angle between the direction vectors in the pre-sampled touch point trajectory sample set. The preset curvature amplitude threshold is determined by removing outliers from the ratio of the median to the range of the rate of change of the curvature derivative during the process of calculating the ratio of the curvature derivative to the arc length based on the local tangent slope. The preset stability benchmark value is determined by weighting the average value and variance ratio of the stability index calculated from the spatial density of density clusters and the profile factor.

9. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 1, characterized in that, The method also includes step S5: S5: Call the action classification stability identifier to make a consistency judgment on the action type of the touch point, divide the action group, filter the action cluster whose internal consistency index meets the preset threshold, calculate the touch point response weight coefficient based on different action categories, and generate a multi-touch response optimization signal. The multi-touch response optimization signal includes action group identifier, action consistency coefficient, and touch response weight coefficient.

10. The method for optimizing multi-touch response of a capacitive touchscreen based on an AI algorithm according to claim 9, characterized in that, The specific steps of S5 are as follows: S501: Obtain the touch point trajectory dataset of the touch panel, call the action classification stability identifier, perform correlation analysis on the change rate of touch point speed and the change in contact area, aggregate the index of similar touch points according to the action type mapping rules, and generate an action group index matrix; S502: Based on the action group index matrix, call the touch point trajectory dataset within the group, calculate the trajectory shape similarity and compare it with the preset action consistency threshold, filter the set of touch points with similarity values ​​higher than the threshold, and generate an action cluster filtering result set; S503: Call the set of touch points in the action cluster filtering result set, calculate the touch point response intensity parameters according to the action category, perform weighted fusion using the category weight parameters, and generate a multi-touch response optimization signal by combining the group dynamic state information.

Images