A touch-sensitive control method

Abstract

This invention discloses a touch sensing control method, relating to the field of touchscreen technology. The method includes: collecting the capacitance response values ​​of all sensing units on the touchscreen; calculating capacitance mutation based on the capacitance response values ​​and the reference capacitance values ​​of the sensing units; calculating a mutation threshold based on noise statistics in a non-touch state; marking sensing units with capacitance mutation exceeding the mutation threshold as candidate touch points; clustering the candidate touch points using a spatial clustering algorithm, grouping spatially adjacent candidate touch points into clusters; calculating the centroid coordinates of each cluster as the effective touch point position; calculating the capacitance response value offset gradient between the effective touch point and its surrounding neighboring sensing units; identifying false touch points based on the capacitance mutation and offset gradient of the effective touch points, combined with preset high mutation thresholds and high gradient thresholds, and using a compensation algorithm for signal purification; and generating continuous touch trajectories by constructing a cost matrix for trajectory association and matching.

Description

Technical Field

[0001] This invention belongs to the field of touch screen technology, and specifically relates to a touch sensing control method. Background Technology

[0002] As a core component of human-computer interaction, touchscreens face a series of interconnected systemic problems in existing technologies. Their inherent defects lie in insufficient signal purification capabilities, rigid trajectory association mechanisms, and a lack of environmental adaptability. This leads to a significant decrease in the system's anti-interference ability, trajectory continuity, and positioning accuracy under complex scenarios such as high electromagnetic interference, multi-user concurrent operation, screen surface wear, and temperature and humidity changes. This makes it difficult to simultaneously meet end-users' expectations for a smooth and precise operating experience, as well as manufacturers' requirements for system cost, stability, and long-term reliability. Real-world testing shows that existing solutions have significant shortcomings in key performance areas such as false touch detection, trajectory tracking robustness, multi-user differentiation accuracy, and positioning stability under environmental changes. They are unable to adapt to increasingly complex real-world application environments, hindering the further development and improvement of touch interaction technology and user experience. Summary of the Invention

[0003] The purpose of this invention is to provide a touch sensing control method that solves the problems in the prior art, such as touch signals being susceptible to noise interference, an increase in false touch points, inaccurate multi-touch point clustering, high trajectory breakage rate, difficulty in distinguishing multiple user identities, and positioning drift caused by screen surface deformation and temperature and humidity changes. It achieves the technical effects of robust touch detection and clustering, false touch point identification and purification, continuous trajectory tracking, multi-user invisible identification and collaborative conflict arbitration, as well as surface state adaptive compensation and health monitoring.

[0004] The touch-sensing control method proposed in this invention comprises the following steps:

[0005] (1): Collect the capacitance response values ​​of all sensing units of the touch screen, and calculate the capacitance change of the sensing power supply based on the capacitance response values ​​and the reference capacitance values ​​of the sensing units; calculate the change threshold according to the noise statistics in the no-touch state, and mark the sensing units whose capacitance change exceeds the change threshold as candidate touch points.

[0006] (2): Spatial clustering algorithm is used to cluster candidate touch points, and spatially adjacent candidate touch points are classified into clusters, with each cluster representing a touch event;

[0007] (3): Calculate the centroid coordinates of each cluster as the effective contact position; for each effective contact, calculate the capacitance response value offset gradient between it and the surrounding neighboring sensing units; based on the capacitance change amount and capacitance response value offset gradient of the sensing power supply described in step (1), identify false contacts by combining the preset high change amount threshold and high gradient threshold.

[0008] (4): For the false touch points identified in step (3), a compensation algorithm based on neighborhood consistency is used to clean up the signal; for the cleaned false touch points, a cost matrix is ​​constructed to perform trajectory association and matching to generate a continuous touch trajectory.

[0009] Furthermore, the formula for calculating the capacitance change of the sensing unit in step (1) is as follows:

[0010] ,

[0011] in, For at a certain point in time ,coordinate The sudden change in capacitance of the sensing unit at that location; For the sensing unit at time point Located at coordinates The capacitance value measured in real time by the sensing unit at the location; The reference capacitance value is the value at which the touchscreen is located when it is not touched or is stationary. The inherent capacitance value measured by the sensing unit at that location.

[0012] Furthermore, the formula for calculating the mutation threshold in step (1) is as follows:

[0013] ,

[0014] in, The mutation threshold, This represents the average capacitance change of all sensing units in a non-touch state. The standard deviation of capacitance variation for all sensing units in a non-touch state; The sensitivity coefficient is determined experimentally, with a value range of [2.5, 3.5].

[0015] Furthermore, the formula for calculating the capacitor response value offset gradient in step (3) is as follows:

[0016] ,

[0017] in, For time points Located at coordinates The effective contact offset gradient value at the location; It is a function of standard deviation. For time points A neighboring unit located around the central unit ( The real-time capacitance response value;

[0018] The neighborhood includes units in eight directions around the center point: up, down, left, right, upper left, lower left, upper right, and lower right.

[0019] Further, the identification of false contacts in step (3) includes: if the capacitance change of a valid contact is greater than or equal to a preset high change threshold and its offset gradient is greater than or equal to a preset high gradient threshold, then the valid contact is determined to be a false contact.

[0020] The high mutation threshold is set based on the noise standard deviation in the non-touch state, and the high gradient threshold is set based on the gradient distribution of historical real touch data.

[0021] Further, the signal purification in step (4) includes: finding a contact that is not marked as a false contact and is spatially closest within the cluster where the false contact is located as a reference contact; calculating the ratio of the capacitance change between the false contact and the reference contact as a compensation coefficient; and using the compensation coefficient to linearly correct the capacitance response value of the false contact to obtain the purified capacitance value.

[0022] Furthermore, the cost matrix mentioned in step (4) includes: constructing a cost matrix. Its dimensions are ,in It is the number of existing trajectory point sets. This represents the number of points in the set to be matched; the cost matrix is ​​calculated from weighted scores across multiple dimensions.

[0023] ,

[0024] in, For trajectory association cost function, The distance cost is the point to be matched. With trajectory The Euclidean distance between the predicted locations; To achieve speed consistency, the trajectory is calculated. The instantaneous velocity vectors of the two most recent points and the trajectory The last point points to the contact to be matched. The cosine of the angle between the vectors is obtained; The signal feature similarity cost is used to evaluate the points to be matched. With trajectory Similarity in signal characteristics; The weights, calibrated experimentally, and .

[0025] Furthermore, the trajectory association and matching in step (4) includes a trajectory retention mechanism: if an existing trajectory is not successfully matched in the current time frame, the trajectory is not terminated immediately, but is allowed to be temporarily retained for a preset number of frames; during the retention period, if a touch point can be successfully matched with it in a subsequent frame, the trajectory is restored.

[0026] Furthermore, when user identity concealment recognition, personalized feature fusion, and real-time conflict arbitration mechanisms are lacking in multi-user operation scenarios, the method also includes multi-user identification:

[0027] Pressure data from the touchscreen pressure-sensitive layer is collected at the same sampling frequency as the capacitive response value, organized into a pressure data cube, and after noise reduction by moving average filtering, it is aligned with the timestamp of the capacitive data to generate a multi-user enhanced spatiotemporal dataset.

[0028] Extract the pressure and gesture features of each valid touch point and combine them into a user feature vector;

[0029] By integrating physical layer electromagnetic spectrum data, environmental layer device status data, and behavioral layer operation timing data, a spectrum template is generated after device status dependency correction of the electromagnetic spectrum data, and user basic information is bound to it and stored in the archive database.

[0030] The similarity of the electromagnetic spectrum of the new contact point with the spectrum, the similarity of the operation timing, and the consistency index of the equipment status are calculated, and a comprehensive correlation score is obtained by weighted summation.

[0031] If the overall correlation score is greater than the matching threshold, the corresponding user ID is bound to the touchpoint; otherwise, it is marked as a new user and a profile is created, and the user spectrum template is updated using an exponentially weighted moving average.

[0032] Furthermore, the method lacks an adaptive compensation mechanism for capacitance baseline drift and trajectory positioning errors caused by changes in physical states such as microscopic deformation of the touchscreen surface and temperature and humidity gradients. Therefore, the method also includes adaptive surface state compensation.

[0033] Deformation depth data, temperature distribution data, and humidity distribution data of the touch screen surface are collected at the same sampling frequency as the capacitance response value, and deformation depth map, temperature distribution map, and humidity distribution map are generated respectively; the deformation gradient is calculated based on the deformation depth map, and the temperature gradient and humidity gradient are calculated based on the temperature distribution map and humidity map respectively;

[0034] Establish a coupling model between deformation depth, temperature gradient, humidity gradient and capacitance change, and calculate the total capacitance compensation for each sensing unit.

[0035] The position offset is calculated based on the linear transformation model between the total capacitance compensation and the position offset, and directional compensation is performed in combination with the gradient direction to generate a real-time compensation mapping table. After the contact point is detected, the corresponding position offset is obtained by querying the compensation mapping table according to the contact point coordinates, the original contact point coordinates are corrected, and the corrected contact point coordinates are output.

[0036] The beneficial effects of this invention are as follows:

[0037] By employing noise adaptive thresholding and spatial clustering, the detection accuracy and positioning precision of touch signals in interference environments are improved. Combined with gradient analysis and neighborhood compensation mechanisms, false touch points are effectively identified and eliminated, reducing false positives. Multi-dimensional trajectory association and dwell strategies enhance the continuity and stability of trajectories in complex scenarios such as intersections and discontinuities. Multi-source data integration enables user-invisible differentiation and intelligent conflict arbitration, supporting multi-user collaborative operation. Real-time sensing of surface deformation and temperature / humidity changes, along with a dynamic compensation model to eliminate positioning drift, improves long-term accuracy and system reliability, and provides self-monitoring capabilities for screen health status. Attached Figure Description

[0038] Figure 1 This is a flowchart of a touch-sensing control method according to an embodiment of the present invention. Detailed Implementation

[0039] To facilitate understanding of the present invention, a more complete description of the invention will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein; rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the invention.

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0041] Example 1: As Figure 1 As shown, a touch-sensing control method is described.

[0042] S1: Collect the capacitance response values ​​of all sensing units of the touch screen, and calculate the capacitance change based on the capacitance response values ​​and the reference capacitance values ​​of the sensing units; calculate the change threshold based on the noise statistics in the no-touch state, and mark the sensing units whose capacitance change exceeds the change threshold as candidate touch points;

[0043] Specifically, the capacitance response values ​​of all sensing units on the touchscreen are continuously acquired at a high sampling rate (100Hz) for a duration of one sliding window (0.5 seconds). The capacitance response value data for each timestamp is organized into a spatiotemporal data cube, with dimensions including spatial coordinates and timestamps. The data cube is stored in matrix form, where each element is the capacitance response value of that sensing unit at the corresponding moment. The capacitance response value is the specific numerical value measured for each sensing unit, indicating the change in capacitance relative to a reference state caused by touch behavior.

[0044] For each sensing unit in the spatiotemporal data cube, at each moment, the stable capacitance value of that sensing unit in a non-touch state is used as the reference capacitance value. Based on the capacitance response value acquired by that sensing unit at the current moment, its abrupt change is calculated:

[0045] ,

[0046] in, The capacitance change is at a specific time point. ,coordinate The change or offset of the capacitance value of the sensing unit at that location relative to its reference state; This is the current capacitor response value, which is the current time point. Located at coordinates The capacitance value measured in real time by the sensing unit at the location; The reference capacitance value is the value at which the touchscreen is located when it is not touched or is stationary. The inherent capacitance value measured by the sensing unit at that location.

[0047] Not all capacitance changes are valid touch signals; filtering can be performed by setting a dynamic abrupt change threshold.

[0048] ,

[0049] in, The mutation threshold, This represents the average capacitance change, which is the average level of capacitance change across all sensing units in a non-touch state. Ideally, the capacitance should not change when there is no touch. It should be very close to 0, but in reality, due to the influence of environmental factors, it will have a small positive or negative value. The standard deviation of capacitance variation represents the variation in capacitance across all sensing units around the mean when there is no touch input. The degree of fluctuation or dispersion; The higher the value, the more intense and unstable the environmental noise. The sensitivity coefficient is determined experimentally and its value ranges from 2.5 to 3.5.

[0050] Each sensing unit makes a judgment at each moment, if Then the unit will be in The time point is marked as a candidate contact.

[0051] S2: A spatial clustering algorithm is used to cluster candidate touch points, classifying spatially adjacent candidate touch points into clusters, with each cluster representing a touch event;

[0052] Specifically, after initial detection using a dynamic mutation threshold, a series of candidate touch points with capacitance mutations exceeding the threshold are obtained. However, a single finger touch typically covers multiple adjacent sensing units, resulting in multiple units within a region being simultaneously marked as candidate touch points. Even if the center of the touch point is clearly defined, its electric field effect will cause surrounding units to experience capacitance changes below the center point but above the threshold, forming a diffusion region. Therefore, these candidate touch points cannot be directly used as the final valid touch points. Thus, DBSCAN clustering (density-based noisy spatial clustering algorithm) is used to group spatially adjacent candidate touch points into the same cluster, with each cluster representing an independent touch event to address the aforementioned problem.

[0053] The neighborhood radius is set to the spacing of 2-3 sensing units, roughly corresponding to the range of sensing units that an adult fingertip can cover on a typical touchscreen. Setting it within this range ensures that multiple candidate touch points generated by the same finger and naturally clustered in space can be identified as neighboring areas by the algorithm and thus grouped into the same cluster. The minimum number of points is set to 2 to distinguish between real touch clusters and noise points; a truly meaningful touch, due to its area, will induce capacitance changes on multiple adjacent sensing units (i.e., generating at least two or more candidate touch points). An isolated point with only one unit exceeding the threshold is highly likely to be random noise or slight electromagnetic interference. Each cluster identified by the DBSCAN algorithm represents an independent touch event. The centroid of the coordinates of all candidate touch points within the cluster is calculated, and this centroid coordinate is taken as the valid touch point position at that moment. The abrupt change of this touch point is taken as the maximum or average of the abrupt changes of all units within the cluster.

[0054] S3: Calculate the centroid coordinates of each cluster as the effective contact point position; for each effective contact point, calculate the capacitance response value offset gradient between it and the surrounding neighboring sensing units; based on the capacitance change amount and offset gradient of the effective contact point, combined with the preset high change amount threshold and high gradient threshold, identify false contact points;

[0055] Specifically, to further distinguish between genuine touch and interference, the relationship between each touch point and its surrounding environment is analyzed. For each valid touch point, neighboring sensing units in eight directions are acquired, and the degree of difference in capacitive response values ​​between the touch point and all neighboring units is calculated. The standard deviation is used as a gradient metric.

[0056] ,

[0057] in, The offset gradient value represents the gradient at a specific time. Located at coordinates The degree of dispersion of the capacitance response values ​​between the central sensing unit and all its neighboring units. The larger the value, the more inconsistent the differences between the center point and the surrounding points, and the more abnormal it may be. It is a function of standard deviation. For a moment Located at coordinates The real-time capacitance response value of the central sensing unit. For a moment A neighboring unit located around the central unit ( The real-time capacitance response value, the neighborhood refers to the cells in eight directions around the center point: up, down, left, right, upper left, lower left, upper right, and lower right.

[0058] False touchpoints are identified based on abrupt changes and offset gradients. High abrupt change thresholds and high gradient thresholds are set: High abrupt change threshold... The system is used to capture abnormally strong signals; the high gradient threshold is set based on the gradient distribution of historical real touch data (95th percentile) to identify signal discontinuities. For each valid touch point, if the abrupt change at that point is greater than or equal to the high abrupt change threshold and the offset gradient is greater than or equal to the high gradient threshold, it is determined to be a false touch point.

[0059] S4: For the identified false touch points, a compensation algorithm based on neighborhood consistency is used to clean up the signal; for the cleaned touch points, a cost matrix is ​​constructed to perform trajectory association and matching to generate a continuous touch trajectory.

[0060] The signal purification includes: finding a reference contact that is not marked as a false contact and is spatially closest to it within the cluster where the false contact is located; calculating the ratio of the capacitance change between the false contact and the reference contact as a compensation coefficient; and using the compensation coefficient to linearly correct the capacitance response value of the false contact to obtain the purified capacitance value.

[0061] Specifically, for spurious touch points, a linear compensation algorithm based on neighborhood consistency is used for purification, rather than simple deletion, to preserve potentially genuine touch information. Within the DBSCAN cluster where the spurious touch point resides, the nearest unmarked touch point is selected as a reference touch point; the signal characteristics of this reference touch point are considered reliable. The abrupt change ratio between the spurious touch point and the reference touch point is calculated.

[0062] ,

[0063] in, For compensation coefficient, The capacitance change at the reference contact. This represents the capacitance change at the spurious contact.

[0064] Using compensation coefficient Correcting the capacitance response value of spurious contacts:

[0065] ,

[0066] in, The capacitance value after purification. The original capacitance value of the dummy contact. The reference capacitance value, To correct the weights, The smaller, The larger the value, the stronger the correction. After completing all purification operations, a list of purified contacts is generated.

[0067] After obtaining the cleaned contact list, in order to associate the cleaned, discrete time-series contacts into continuous and accurate touch trajectories, trajectory association and matching are performed. The first time frame... All touchpoints in the initial trajectory serve as the starting point, and each trajectory is assigned a unique ID. For each subsequent time frame ( ), and regard its touch points as the set of points to be matched, and take the previous time frame ( The last touchpoint of all active trajectories (i.e., trajectories not yet marked as terminated) in the dataset is considered as the set of existing trajectory points. A cost matrix is ​​constructed. Its dimensions are ,in It is the number of existing trajectory point sets. This represents the number of points in the set to be matched. Each element in the matrix... The representative will track Matching point The cost of linking them. The lower the cost, the greater the probability that the two belong to the same trajectory. Cost Calculated from weighted scores across multiple dimensions:

[0068] ,

[0069] in, For trajectory association cost function, The distance cost is the point to be matched. With trajectory The Euclidean distance between the predicted locations; The cost of speed consistency is used to evaluate the points to be matched. With trajectory The consistency of historical movement direction is calculated through trajectory calculation. The instantaneous velocity vectors of the two most recent points and the trajectory The last point points to the contact to be matched. The cosine value of the angle between vectors (cosine similarity) is obtained. The signal feature similarity cost is used to evaluate the points to be matched. With trajectory Similarity in signal characteristics. The corresponding weights were calibrated experimentally, and .

[0070] The formula for calculating the speed consistency cost is:

[0071] ,

[0072] in, For the sake of speed consistency, It is a trajectory The instantaneous velocity vectors of the two most recent points and the trajectory The last point points to the contact to be matched. The angle between vectors, when the two directions are exactly the same. but When the two directions are completely opposite, ,but Therefore The value of is in the range of [0, 2]. The smaller the angle, the closer the cosine value is to 1, and the lower the cost.

[0073] The formula for calculating the signal feature similarity cost is:

[0074] ,

[0075] in, The cost is the similarity of signal features. Contacts to be matched The sudden change in capacitance. For trajectory The average value of capacitance changes over a recent number of points (e.g., 3-5 points). The absolute difference between the mutation amounts of the two. To take the maximum of the two.

[0076] The identification of false contacts includes: if the capacitance change of a valid contact is greater than or equal to a preset high change threshold, and its offset gradient is greater than or equal to a preset high gradient threshold, then the valid contact is determined to be a false contact.

[0077] Specifically, the Hungarian algorithm is used to process the cost matrix. The algorithm solves for a set of matching relationships that minimizes the total association cost. Based on the matching results of the Hungarian algorithm, it manages the updating, creation, and termination of trajectories: for successfully matched trajectories... ,point Yes, place the point Add coordinates, timestamps, and other information to the track In the middle, update the trajectory The endpoint and velocity vector. Unmatched points to be matched are considered the starting point of a new trajectory and assigned a new trajectory ID. Existing unmatched trajectories do not terminate immediately; instead, a trajectory holding mechanism is initiated, allowing the trajectory to be temporarily held for several frames (e.g., 3 frames). If a point can be successfully matched with it in subsequent frames during this period, the trajectory is resumed. If no match is found after the holding frame period, the trajectory is officially terminated to handle the situation where the touch point temporarily disappears.

[0078] For each generated trajectory, a Kalman filter is applied for real-time smoothing to reduce jitter and predict future positions. Trajectory confidence levels are set.

[0079] ,

[0080] in, For trajectory confidence, As a continuous indicator, As a smoothness index, As a signal strength indicator, , and For the corresponding weight coefficients, and satisfying The weight values ​​need to be calibrated using experimental data.

[0081] The formula for calculating the continuity index is:

[0082] ,

[0083] in, As a continuous indicator, This represents the total number of pauses triggered within the lifecycle of this trajectory, i.e., the number of frames that were not successfully matched. This represents the total duration of the trajectory from its creation to its termination. The lower the percentage of frames retained, the closer the continuity index is to 1, indicating a more coherent trajectory.

[0084] The formula for calculating the smoothness index is:

[0085] ,

[0086] in, As a smoothness index, This is the standard deviation of the distance between consecutive pairs of points in the trajectory, and it is normalized using a reference distance. The smaller the value, the more uniform the movement speed, the smoother the trajectory, and the closer the index value is to 1. Conversely, a larger distance fluctuation indicates severe trajectory jitter and poor smoothness.

[0087] The formula for calculating the signal strength index is:

[0088] ,

[0089] in, As a signal strength indicator, This is the average value of the capacitance changes at all points on the trajectory. This is a high mutation threshold. The average signal strength is compared to a high threshold and normalized. The stronger the signal, the higher the index. When the average signal exceeds the high threshold, the index saturates to 1.0, indicating a very strong signal.

[0090] when If the confidence level is ≥0.7 (the confidence threshold), the trajectory is marked as a valid trajectory. Otherwise, the trajectory is considered a low-quality trajectory and may be selectively discarded or downgraded.

[0091] The optimized and filtered trajectories are converted into a system-usable format, generating a data packet for each valid trajectory, containing: trajectory ID, trajectory status (start, movement, end), current smoothed coordinates, and additional information. The final output is the optimized multi-touch trajectory.

[0092] The technical solutions in the above embodiments of the present invention have at least the following technical effects or advantages:

[0093] This invention utilizes a spatiotemporal data cube to organize capacitive response data, combined with a dynamic mutation threshold and the DBSCAN clustering algorithm, to achieve preliminary detection and clustering of touch signals. This effectively distinguishes between genuine touches and noise, improving the accuracy and robustness of touch detection. By introducing offset gradient analysis and a neighborhood consistency-based compensation mechanism, it can identify and eliminate false touch points, reducing misjudgments caused by electromagnetic interference or environmental noise, further enhancing the system's anti-interference capability. In addition, by employing multi-dimensional trajectory association matching and Kalman filtering smoothing, it achieves continuous and stable tracking of touch trajectories, maintaining trajectory consistency even when touch points temporarily disappear or signals fluctuate, thus improving user experience and system reliability.

[0094] Example 2: Example 1 achieved high-precision detection, clustering and purification, and continuous trajectory generation of single-point touch signals based on capacitance response values. However, it lacks user identity stealth recognition, personalized feature fusion, and real-time conflict collaborative arbitration mechanisms in multi-user operation scenarios. The following steps can be added.

[0095] The pressure data of the touch screen pressure-sensitive layer is collected at the same sampling frequency as the capacitance response value, organized into a pressure data cube, and after being filtered and denoised by moving average, it is aligned with the timestamp of the capacitance data to generate a multi-user enhanced spatiotemporal dataset.

[0096] Specifically, the pressure value of each touch point is collected through the touchscreen pressure-sensitive layer, also at a frequency of 100Hz, and organized into a pressure data cube with dimensions completely consistent with the capacitive data cube. Each element... This represents the pressure applied by the corresponding sensing unit at a specific moment. Each time frame within each sampling window is marked with a high-precision, monotonically increasing timestamp. A moving average filter is used to denoise the pressure data and to standardize the timestamp alignment.

[0097] After data collection and preprocessing, a multi-user augmented spatiotemporal dataset is generated, which includes capacitance data cubes, pressure data cubes, and high-precision timestamp sequences.

[0098] The pressure and gesture features of each valid touch point are extracted and combined into a user feature vector.

[0099] Specifically, for each valid touch point, pressure feature extraction and gesture feature extraction are performed. Pressure feature extraction includes pressure fluctuation coefficient, maximum pressure value, and pressure rise slope. The pressure fluctuation coefficient is the standard deviation of the pressure sequence divided by its mean, focusing on the relative fluctuation of pressure; the maximum pressure value is the maximum value of the pressure sequence over the entire duration of the touch point; the pressure rise slope is the linear regression slope of the pressure value over time within the first 100 milliseconds after the touch point begins. Gesture feature extraction includes trajectory curvature features, average velocity, and gesture type probability distribution. The trajectory curvature feature is the instantaneous curvature of a series of consecutive points (e.g., groups of three consecutive points) on the trajectory segment, and the mean and standard deviation of these instantaneous curvatures are taken as two features; the average velocity is the total displacement length of the trajectory segment divided by the total time; the gesture type probability distribution is obtained by using a pre-trained lightweight convolutional neural network (CNN) classifier, inputting the coordinate sequence of the trajectory segment into the classifier, and obtaining the probability vector of its belonging to various preset gesture types (e.g., click, swipe, pinch, rotate). The two highest probability values ​​are taken as features. The extracted features are combined into a unified user feature vector.

[0100] The electromagnetic spectrum data from the physical layer, the device status data from the environmental layer, and the operation timing data from the behavioral layer are integrated. After the electromagnetic spectrum data is corrected for device status dependence, a spectrum template is generated and bound to the user's basic information and stored in the archive database.

[0101] The similarity of the electromagnetic spectrum of the new contact point with the spectrum, the similarity of the operation timing, and the consistency index of the equipment status are calculated, and a comprehensive correlation score is obtained by weighted summation.

[0102] Specifically, after obtaining the user feature vector, multi-source data is simultaneously injected, including physical layer electromagnetic spectrum data, environmental layer device status data, and behavioral layer operation timing data. The physical layer electromagnetic spectrum data is generated by collecting high-frequency electromagnetic fluctuations (30-100Hz band) caused by user touches through the touchscreen's built-in spectrum sensor, producing a spectrum energy distribution vector. The environmental layer device status data includes battery level (%), CPU load rate (%), and device temperature (°C). The behavioral layer operation timing data is the time interval pattern of continuous user operations. Device status-dependent corrections are applied to the electromagnetic spectrum data. For example, when the battery level is below 20%, the electromagnetic signal may attenuate; a correction formula is applied as follows:

[0103] ,

[0104] in, The corrected spectral energy, The original spectral energy, This represents the battery percentage. The proportionality coefficient of the electrical quantity to the attenuation of the electromagnetic signal is obtained through experimental calibration, such as... For every 10% decrease in the representative quantity, the signal attenuation is at most 1%.

[0105] The average value within the time window of the corrected spectral data is taken to generate a spectral template:

[0106] ,

[0107] in, For spectral template, For the number of samples, For the first The corrected spectral vector is then used to bind the template with the user's basic information and initial operation timing pattern (such as average click interval), and stored in the archive database.

[0108] After each successful user identification, its template is updated using an Exponentially Weighted Moving Average (EWMA) to ensure it adapts to long-term changes.

[0109] ,

[0110] in, For the new spectrum template, The original spectrum template, This is the calibration spectrum for the current contact. is the forgetting factor, with a value range of [0.85, 0.95].

[0111] The strong correlation between electromagnetic spectrum data and user biometrics is utilized as the primary feature for user identification. This is achieved by calculating the matching degree between the electromagnetic spectrum of a new contact point and spectrum templates in historical user profiles.

[0112] ,

[0113] in, For spectral similarity scores, For the new contact spectral vector, For user profile spectrum vectors, For the current contact at frequency The electromagnetic energy value on For target user profiles at frequency The average electromagnetic energy on the surface, The sum of the squares of all components of the electromagnetic energy vector at the current contact point. This is the sum of the squares of all components of the electromagnetic energy vector in the user profile.

[0114] Calculate the overall association score between the new touchpoint and all user profiles:

[0115] ,

[0116] in, To achieve a comprehensive correlation score, For spectral similarity scores, For operational temporal similarity, This is an indicator of equipment status consistency. For the corresponding weight coefficients, and According to the experimental calibration.

[0117] The formula for calculating the temporal similarity of operations is:

[0118] ,

[0119] in, For operational temporal similarity, Accumulate distance to normalize the path and normalize it using reference distance. The smaller the value, the higher the similarity. The value range is [0,1].

[0120] When the battery level is >20% and the temperature is <40°C, the device status consistency index is 1; otherwise, it is 0.5.

[0121] If the overall correlation score is greater than the matching threshold, the corresponding user ID is bound to the touchpoint; otherwise, it is marked as a new user and a profile is created, and the user spectrum template is updated using an exponentially weighted moving average.

[0122] Specifically, if the highest association score is greater than the matching threshold (0.7), the corresponding user ID is bound to that touchpoint; otherwise, it is marked as a new user, and the profile creation process is triggered. For already bound users, their profiles are updated in real time.

[0123] ,

[0124] in, For user profile spectrum vectors, For the new contact spectral vector, The forgetting factor has a value range of [0.85, 0.95]. This is the original user profile spectrum vector.

[0125] By inputting user feature vectors and multi-source data, the system outputs a list of touchpoints with user IDs.

[0126] To address potential issues such as trajectory intersection and resource contention in multi-user operations, dynamic analysis of environmental load data and operation timing data is used to arbitrate and optimize the spatiotemporal intersection degree of conflicting trajectories. For each trajectory, the set of touchpoints in its current time frame is extracted, and spatial intersection calculations are performed with the touchpoints of other user trajectories.

[0127] ,

[0128] in, Let be the overlap between the trajectories of user i and user j. These are the sets of contact points for the two trajectories, Indicates the size of the set.

[0129] Quantify the impact of environmental load based on the real-time system status:

[0130] ,

[0131] in, For environmental load factor, for Occupancy rate For memory usage, Based on the average base load, The network latency is calculated and normalized using a reference latency.

[0132] Conflict tendencies are analyzed based on user operation history:

[0133] ,

[0134] in, For users and Timing competition factor, For the number of historical conflicts, This represents the total number of interactions. To alternate operating frequencies, and normalize using a reference frequency, The weighting coefficient is 0.1, determined experimentally.

[0135] The conflict index is calculated by combining trajectory overlap, environmental load, and temporal competition factors:

[0136] ,

[0137] in, As a conflict index, and For adjustment coefficients, , .

[0138] like If the value is less than 0.3, it is considered a low-risk collision scenario, and a collaborative mode is adopted: automatically allocating shared areas and allowing parallel trajectory processing. If 0.3 ≤ A score <0.7 indicates a medium risk of conflict, and a priority-based approach is adopted: priorities are assigned based on users' historical activity levels, with higher-priority users having their activity patterns kept continuous. If the value is ≥0.7, it is considered a high risk of conflict and a competition mode is adopted: the trajectory path is dynamically adjusted based on the temporal competition factor. The system will prioritize the continuity and optimal path of the trajectory of users with lower temporal competition factors, while applying a certain path offset or delay to the trajectory of users with higher competition factors, thereby maximizing the overall operational efficiency under limited resources.

[0139] When the system detects high temporal competition factors and high trajectory overlap among users, it triggers a phase analysis mechanism. An artificial phase offset is applied to the highly synchronized user group, achieving a misalignment of contact acquisition times by fine-tuning the sampling time window. The phase offset amount is dynamically adjusted based on the users' temporal competition factors. An interpolation algorithm is then used to reconstruct the offset contact sequence, generating temporally staggered but spatially continuous trajectory data.

[0140] The technical solutions in the above embodiments of the present invention have at least the following technical effects or advantages:

[0141] This invention constructs a multi-dimensional, high-precision user feature vector by introducing a pressure data cube, high-precision timestamp synchronization, and multi-source data fusion, achieving a refined description of touch behavior and high-accuracy user identification. Through a comprehensive correlation scoring mechanism based on electromagnetic spectrum similarity, operation timing matching, and device state consistency, it can distinguish different users in real time in multi-user collaborative operation scenarios. Through a conflict detection and arbitration mechanism, combined with multi-dimensional evaluation of trajectory overlap, environmental load factors, and timing competition factors, it achieves hierarchical classification and adaptive processing of multi-user operation conflicts.

[0142] Example 3: Example 2 implemented multi-user stealth recognition, personalized feature extraction and trajectory conflict collaborative arbitration based on multi-source data fusion. However, it lacks an adaptive compensation mechanism for capacitance baseline drift and trajectory positioning errors caused by changes in physical states such as micro-deformation of the touch screen surface and temperature and humidity gradients. The following steps can be added to address this issue.

[0143] The deformation depth data, temperature distribution data, and humidity distribution data of the touch screen surface are collected at the same sampling frequency as the capacitance response value, and deformation depth map, temperature distribution map, and humidity distribution map are generated respectively; the deformation gradient is calculated based on the deformation depth map, and the temperature gradient and humidity gradient are calculated based on the temperature distribution map and humidity distribution map respectively;

[0144] Establish a coupling model between deformation depth, temperature gradient, humidity gradient and capacitance change, and calculate the total capacitance compensation for each sensing unit.

[0145] Specifically, a macro optical sensor is used to scan the touchscreen surface to generate a deformation depth map. Deformation depth Defined as the deviation between the current surface and the ideal plane, in micrometers. Deformation data is acquired at a frequency of 100Hz to ensure synchronization with capacitance data. Based on the deformation depth map, the deformation gradient is calculated:

[0146] ,

[0147] in, For deformation gradient, For the deformation depth in Partial derivatives in direction, For the deformation depth in Partial derivatives in direction, It is the sum of the squares of the rates of change in the two directions.

[0148] An array of temperature and humidity sensors is deployed on the touchscreen surface to collect temperature data at a frequency of 100Hz. and humidity Calculate the temperature gradient based on the temperature and humidity distribution map. and humidity gradient For each sensing unit, a surface state vector is generated. .

[0149] Surface micro-deformation alters the effective distance between the sensing unit and the touched object, causing distortion in the electric field distribution and a shift in the capacitance baseline. Therefore, a linear relationship model is established between the deformation depth and the abrupt change in capacitance.

[0150] ,

[0151] in, This represents the sudden change in capacitance. For the depth of deformation, The deformation-capacitance coupling coefficient is determined experimentally and its value ranges from [0.015 to 0.025].

[0152] Changes in temperature and humidity affect the dielectric constant, thus altering the distribution of the capacitive induced field. Therefore, a linear combination model of temperature and humidity gradients and abrupt changes in capacitance is established:

[0153] ,

[0154] in, This represents the sudden change in capacitance. For temperature gradient, This is the temperature-capacitance coupling coefficient, calibrated experimentally, such as 0.04 pF / (°C / mm). For humidity gradient, The humidity-capacitance coupling coefficient is determined experimentally, such as 0.008pF / (% / mm).

[0155] For each sensing unit, the total capacitance compensation is the sum of the effects of deformation and temperature / humidity:

[0156] ,

[0157] in, This represents the total capacitance compensation.

[0158] The position offset is calculated based on the linear transformation model between the total capacitance compensation and the position offset, and directional compensation is performed in combination with the gradient direction to generate a real-time compensation mapping table. After the contact point is detected, the corresponding position offset is obtained by querying the compensation mapping table according to the contact point coordinates, the original contact point coordinates are corrected, and the corrected contact point coordinates are output.

[0159] Specifically, based on the mapping relationship between the total capacitance change and the position offset, a linear transformation model is established:

[0160] ,

[0161] in, This is the position offset. The position-capacitance conversion factor is determined experimentally, such as 0.08 mm / pF.

[0162] Positional offset is directional, therefore directional compensation is performed. The offset vector is decomposed based on the directions of the deformation gradient and the temperature / humidity gradient:

[0163] ,

[0164] in, for Directional position offset for Directional position offset The direction angle is determined by the direction of the deformation gradient vector to compensate for the angle.

[0165] Calculate position offset compensation parameters for each sensing unit This generates a real-time compensation mapping table. The table uses spatial coordinates... Use an index to store the corresponding offset vector. , When the average temperature and humidity gradient change exceeds 10% or a new wear area is detected in the deformation depth, a dynamic update mechanism is triggered, and the compensation mapping table is updated every 30 seconds to adapt to environmental changes.

[0166] By applying compensation parameters, real-time correction of contact point coordinates is achieved, eliminating surface effects. The original coordinates of the contacts are obtained from the capacitance data. Based on these coordinates, the offset at that position is retrieved from the compensation mapping table, and the corrected coordinates are calculated.

[0167] ,

[0168] in, For the corrected coordinates, Original coordinates This represents the position offset.

[0169] After obtaining the corrected coordinates, further calculations are performed using the directional position offset:

[0170] ,

[0171] in, For the final Direction coordinates For the final Direction coordinates For correction Direction coordinates For correction Direction coordinates for Directional position offset for Directional position offset and Together they form the final coordinates .

[0172] if Areas exceeding a threshold (e.g., 0.3 mm) are marked as high-error areas and trigger intensive monitoring. A surface condition recalibration process is then initiated using the historical average compensation amount.

[0173] After obtaining the final coordinate sequence, where each coordinate point contains spatial location, timestamp, and user identification, the touchpoints are grouped according to the user identification, and an independent trajectory container is created for each user; this includes creating a trajectory object for each newly appearing user identification and initializing the trajectory state.

[0174] Calculate the spatiotemporal overlap of different user trajectories:

[0175] ,

[0176] in, For the spatiotemporal overlap of the trajectory, For users and The time period of the trajectory Let be the length of the time intersection of the two trajectories. The length of the trajectory time period, The spatial decay factor, This represents the minimum spatial distance between the two trajectories, and is normalized using a reference distance. The larger the value, the closer the factor is to 0, and the spatial distance reduces the risk of conflict. After obtaining the spatiotemporal overlap of the trajectories, the conflict index is updated to replace the original trajectory overlap.

[0177] when When <0.3, a cooperative mode is adopted, allowing parallel trajectory processing; when 0.3≤ When the value is less than 0.7, a priority mode is adopted, and resources are allocated according to the user's historical activity level. When the value is ≥0.7, a competition mode is adopted, and path offsets or delays are applied to the trajectories of users with high conflicts.

[0178] After completing the multi-user trajectory generation and arbitration, the surface health status is inferred from the multi-user operation data, the compensation model parameters are dynamically adjusted, and potential faults are warned in advance.

[0179] Analyze the hotspots of user operations, and for each surface area (divided into 5mm×5mm grids), calculate its trajectory drift variance:

[0180] ,

[0181] in, For trajectory drift variance, Let i be the coordinates of the i-th touch point. Let i be the ideal coordinates of the i-th touch point. The average drift amount, This represents the number of sampling points.

[0182] After normalizing the drift variance and operation frequency, the deformation wear index is defined as follows:

[0183] ,

[0184] in, The deformation and wear index, For operation frequency, This represents the average deformation depth of the region. The weighting coefficient, calibrated experimentally, is set to a value of 0.1. At that time, it was marked as a high-risk wear area.

[0185] The system continuously monitors the trajectory correction residuals:

[0186] ,

[0187] in, To correct the residuals of the trajectory, For the final coordinates, The true coordinates are the moving average residuals of 10 consecutive sampling points. When the value is greater than 0.05 mm, it is determined that the current compensation model has a significant deviation, and the adaptive process is immediately triggered.

[0188] The deformation-capacitance coupling coefficient is updated using the gradient descent method:

[0189] ,

[0190] in, The learning rate is approximately calculated using historical data differencing and is set to 0.01. The original deformation-capacitance coupling coefficient, This is the new deformation-capacitance coupling coefficient.

[0191] Synchronously adjust the temperature and humidity coupling coefficient:

[0192] ,

[0193] in, For temperature gradient measurement error, This is the nominal value. The attenuation coefficient, calibrated experimentally, is set to 0.05. This is the original temperature-capacitance coupling coefficient. This is the new temperature-capacitance coupling coefficient. The adjustment logic for the humidity-capacitance coupling coefficient is the same as that for the temperature-capacitance coupling coefficient.

[0194] The optimized new parameter set ( , , The updated compensation model for that region takes effect immediately. The system continues to collect residual data from the next 10 sampling points and calculates the new moving average residual. .like If the value is ≤0.03mm, the adaptive process is considered successful, and the new parameters are permanently stored in the compensation mapping table for that region. If... If the error is still higher than 0.03mm but has improved, the process will start with the new parameters and proceed with the next round of iterative optimization (maximum of 3 iterations to avoid infinite loops). If the residual cannot be reduced after iterations, the system will determine that there is a hardware problem that cannot be compensated by software and generate a high-level maintenance alert.

[0195] The technical solutions in the above embodiments of the present invention have at least the following technical effects or advantages:

[0196] This invention employs a real-time surface deformation compensation mechanism based on multi-sensor fusion. By combining deformation depth, temperature, and humidity gradient data, a coupled model of environmental factors and capacitance changes is established, achieving high-precision touch positioning even under conditions of screen surface wear or changes in ambient temperature and humidity. Through dynamic updates to the compensation mapping table and adaptive parameter adjustments, the system can continuously optimize touch coordinate correction, improving the stability and reliability of touch response. By using multi-user operation data to infer the surface health status, intelligent monitoring and fault warning of screen wear areas are achieved, enhancing the system's self-maintenance and long-term availability.

[0197] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A touch-sensing control method, characterized in that, The specific steps are as follows: (1): Collect the capacitance response values ​​of all sensing units of the touch screen, and calculate the capacitance change of the sensing power supply based on the capacitance response values ​​and the reference capacitance values ​​of the sensing units; calculate the change threshold according to the noise statistics in the no-touch state, and mark the sensing units whose capacitance change exceeds the change threshold as candidate touch points. (2): Spatial clustering algorithm is used to cluster candidate touch points, and spatially adjacent candidate touch points are classified into clusters, with each cluster representing a touch event; (3): Calculate the centroid coordinates of each cluster as the effective contact point position; For each valid contact, calculate the capacitance response value offset gradient between it and the surrounding neighboring sensing units; based on the capacitance change amount and capacitance response value offset gradient of the sensing power supply described in step (1), identify false contacts by combining the preset high change amount threshold and high gradient threshold. (4): For the false touch points identified in step (3), a compensation algorithm based on neighborhood consistency is used for signal purification; For the cleaned-up fake touch points, a cost matrix is ​​constructed to perform trajectory association and matching, generating continuous touch trajectories.

2. The touch sensing control method according to claim 1, characterized in that, The formula for calculating the capacitance change of the sensing unit in step (1) is as follows: ， in, For at a certain point in time ,coordinate The sudden change in capacitance of the sensing unit at that location; For the sensing unit at time point Located at coordinates The capacitance value measured in real time by the sensing unit at the location; The reference capacitance value is the value at which the touchscreen is located when it is not touched or is stationary. The inherent capacitance value measured by the sensing unit at that location.

3. The touch sensing control method according to claim 1, characterized in that, The formula for calculating the mutation threshold in step (1) is: ， in, The mutation threshold, This represents the average capacitance change of all sensing units in a non-touch state. The standard deviation of capacitance variation for all sensing units in a non-touch state; The sensitivity coefficient is determined experimentally, with a value range of [2.5, 3.5].

4. The touch sensing control method according to claim 1, characterized in that, The formula for calculating the capacitor response value offset gradient in step (3) is as follows: ， in, For time points Located at coordinates The effective contact offset gradient value at the location; It is a function of standard deviation. For time points A neighboring unit located around the central unit ( The real-time capacitance response value; The neighborhood includes units in eight directions around the center point: up, down, left, right, upper left, lower left, upper right, and lower right.

5. The touch sensing control method according to claim 1, characterized in that, The high gradient threshold identification of false touch points in step (3) includes: if the capacitance change of a valid touch point is greater than or equal to a preset high change threshold, and its offset gradient is greater than or equal to a preset high gradient threshold, then the valid touch point is determined to be a false touch point. The high mutation threshold is set based on the noise standard deviation in the non-touch state, and the high gradient threshold is set based on the gradient distribution of historical real touch data.

6. The touch-sensing control method according to claim 1, characterized in that, The signal purification in step (4) includes: finding a contact that is not marked as a false contact and is spatially closest to it within the cluster where the false contact is located as a reference contact; calculating the ratio of the capacitance change between the false contact and the reference contact as a compensation coefficient; and using the compensation coefficient to linearly correct the capacitance response value of the false contact to obtain the purified capacitance value.

7. The touch-sensing control method according to claim 1, characterized in that, The cost matrix mentioned in step (4) includes: constructing a cost matrix. Its dimensions are ,in It is the number of existing trajectory points. This represents the number of points in the set to be matched; the cost matrix is ​​calculated from weighted scores across multiple dimensions. ， in, For trajectory association cost function, The distance cost is the point to be matched. With trajectory The Euclidean distance between the predicted locations; To achieve speed consistency, the trajectory is calculated. The instantaneous velocity vectors of the two most recent points and the trajectory The last point points to the contact to be matched. The cosine of the angle between the vectors is obtained; The signal feature similarity cost is used to evaluate the points to be matched. With trajectory Similarity in signal characteristics; The weights, calibrated experimentally, and .

8. The touch sensing control method according to claim 1, characterized in that, The trajectory association and matching described in step (4) includes a trajectory retention mechanism: if an existing trajectory is not successfully matched in the current time frame, the trajectory is not terminated immediately, but is allowed to be temporarily retained for a preset number of frames; during the retention period, if a touch point can be successfully matched with it in a subsequent frame, the trajectory is restored.

9. The touch sensing control method according to claim 1, characterized in that, When user identity concealment recognition, personalized feature fusion, and real-time conflict arbitration mechanisms are lacking in multi-user operation scenarios, the method further includes multi-user identification: Pressure data from the touchscreen pressure-sensitive layer is acquired at the same sampling frequency as the capacitance response value of the sensing unit, organized into a pressure data cube, and after noise reduction by moving average filtering, it is aligned with the timestamp of the capacitance data to generate a multi-user enhanced spatiotemporal dataset. Extract the pressure and gesture features of each valid touch point and combine them into a user feature vector; By integrating physical layer electromagnetic spectrum data, environmental layer device status data, and behavioral layer operation timing data, a spectrum template is generated after device status dependency correction of the electromagnetic spectrum data, and user basic information is bound to it and stored in the archive database. The similarity of the electromagnetic spectrum of the new contact point with the spectrum, the similarity of the operation timing, and the consistency index of the equipment status are calculated, and a comprehensive correlation score is obtained by weighted summation. If the overall correlation score is greater than the matching threshold, the corresponding user ID is bound to the touchpoint; otherwise, it is marked as a new user and a profile is created, and the user spectrum template is updated using an exponentially weighted moving average.

10. The touch sensing control method according to claim 1, characterized in that, The method lacks an adaptive compensation mechanism for capacitance baseline drift and trajectory positioning errors caused by changes in physical states such as microscopic deformation of the touchscreen surface and temperature and humidity gradients. Therefore, the method also includes adaptive surface state compensation. Deformation depth data, temperature distribution data, and humidity distribution data of the touch screen surface are collected at the same sampling frequency as the capacitance response value, and deformation depth map, temperature distribution map, and humidity distribution map are generated respectively; the deformation gradient is calculated based on the deformation depth map, and the temperature gradient and humidity gradient are calculated based on the temperature distribution map and humidity map respectively; Establish a coupling model between deformation depth, temperature gradient, humidity gradient and capacitance abrupt change, and calculate the total capacitance compensation for each sensing unit. The position offset is calculated based on the linear transformation model between the total capacitance compensation and the position offset, and directional compensation is performed in combination with the gradient direction to generate a real-time compensation mapping table. After the contact point is detected, the corresponding position offset is obtained by querying the compensation mapping table according to the contact point coordinates, the original contact point coordinates are corrected, and the corrected contact point coordinates are output.

Images