A noise suppression method, system, device and medium for high signal-to-noise ratio touch screen touch signals

Abstract

The application relates to a noise suppression method, system, device and medium for a high signal-to-noise ratio touch screen touch signal. The method comprises the following steps: obtaining original touch data frames and performing isolated singular point pre-repairing based on a time sequence aligned display refresh state indication signal; extracting display noise spatial form estimation and common mode interference time domain characteristic parameters to obtain noise level data; constructing a dynamic confidence field fused with multiple factors; obtaining a high signal-to-noise ratio touch signal reconstruction frame through anisotropic diffusion and local surface fitting hybrid reconstruction; and extracting a net capacitance change matrix and solving a touch position coordinate. The method can balance background noise suppression and weak touch transient response sensitivity, and improve touch accuracy and operation fluency under complex electromagnetic working conditions.

Description

Technical Field

[0001] This invention belongs to the field of electronic equipment manufacturing technology, and in particular relates to a method, system, device and medium for noise suppression of touch signals of a high signal-to-noise ratio touch screen. Background Technology

[0002] With the widespread adoption of capacitive touchscreens in mobile terminals, automotive cockpits, and industrial control, the ability of touch systems to detect minute changes in capacitance has become a core indicator for evaluating device interaction experience. Currently, touchscreen technology is evolving towards higher refresh rates, higher resolutions, and ultra-thin stacked structures. This leads to a continuous reduction in the physical distance between the touch electrodes and the display cathode layer, resulting in a significant increase in the parasitic coupling effect of the display driving signal to the touch sensing channel. Simultaneously, the widespread use of portable devices in charging scenarios introduces common-mode power supply interference with complex ripple characteristics. These two noise sources exhibit a non-stationary, deeply superimposed state at the touch front end, constituting a major obstacle to improving the touch signal-to-noise ratio.

[0003] To suppress the adverse effects of the aforementioned noise on the touch signal, three types of processing methods are typically adopted: First, a hardware filter with fixed parameters is set at the analog front end to physically attenuate the interference components in a specific frequency band; second, temporal smoothing filtering is performed on the digitized touch data stream, and random glitches are eliminated through moving average or weighted sliding window algorithms; third, spatial template cancellation is performed using the display synchronization signal to directly subtract the display noise floor in the non-touch state from the sampling frame.

[0004] However, the above-mentioned solutions all have significant limitations when facing mixed non-stationary noise environments. For example, the hardware filter parameters are fixed and it is difficult to adaptively track the interference sources with frequency drift; the time-domain smoothing algorithm inevitably introduces touch response delay while suppressing noise, resulting in trajectory lag and disconnection in high-speed sliding or handwriting input scenarios; the template cancellation method based on synchronization signals is only effective when the display noise has a stable periodic distribution. Once the common-mode interference of the charger and the display noise are deeply superimposed, and the noise pattern undergoes nonlinear distortion with touch action, it is impossible to effectively identify large-area interference artifacts and weak signal changes generated by real touch. Especially when the amount of touch signal change is even lower than the fluctuation amplitude of the background noise, the algorithm lacks the ability to dynamically adjust between eliminating noise and preserving the edge details of weak signals, which restricts the interaction accuracy and reliability of the touch screen in harsh electromagnetic environments. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, system, device, and medium for noise suppression of touch signals of a high signal-to-noise ratio touch screen, which can suppress background noise while maintaining instantaneous response sensitivity to weak touch events, thereby improving touch accuracy and operation smoothness under complex electromagnetic conditions.

[0006] In a first aspect, this application provides a noise suppression method for high signal-to-noise ratio touchscreen touch signals, comprising:

[0007] The original touch data frame obtained by the touch screen sensing array in the current scan frame is acquired. Based on the display refresh status indication signal that is aligned with the timing of the current scan frame, the original touch data frame is pre-repaired for isolated singular points based on neighborhood median statistics to obtain the pre-repaired touch data frame.

[0008] Based on the pre-repaired touch data frame and the display refresh status indication signal, the spatial morphology estimation of display noise and the temporal characteristic parameters of common-mode interference in the current frame are extracted to obtain noise level data. The noise level data includes the display noise estimation map, the common-mode interference intensity estimation value, and the noise floor level estimation value.

[0009] A dynamic confidence field is constructed based on pre-repaired touch data frames and noise level data, which integrates spatial consistency factors, temporal continuity factors, and morphological constraint factors. The dynamic confidence field is a two-dimensional weight matrix of the same size as the touch screen sensing array, and each weight value represents the degree of confidence of the corresponding sensing node data in reflecting the real touch state.

[0010] Based on the dynamic confidence field, the pre-repaired touch data frame is subjected to a hybrid reconstruction process combining anisotropic diffusion and local surface fitting to obtain a high signal-to-noise ratio touch signal reconstruction frame.

[0011] Based on the reconstructed frame of the high signal-to-noise ratio touch signal and the pre-established silent baseline reference value, the net capacitance change matrix caused by touch is extracted, and local extremum search and sub-pixel coordinate calculation are performed on the net capacitance change matrix to obtain the touch position coordinate information of the current frame.

[0012] In one embodiment, the original touch data frame obtained by the touchscreen sensing array in the current scan frame is acquired. Based on a display refresh status indication signal aligned with the timing of the current scan frame, the original touch data frame is pre-repaired for isolated singularities based on neighborhood median statistics to obtain a pre-repaired touch data frame, including:

[0013] S21. Obtain the original touch data frame. For each sensing node in the original touch data frame, extract the values ​​of all nodes in the spatial neighborhood window with the current sensing node as the center, and calculate the median and median absolute deviation of the values ​​in the spatial neighborhood window.

[0014] S22. Calculate the absolute value of the deviation between the original value of the current sensing node and the median. When the absolute value of the deviation exceeds the dynamic elastic boundary determined based on the median of the absolute deviation, the current sensing node is determined to be an anomaly.

[0015] S23. For the current sensing node that is determined to be an anomaly, obtain the value of the nearest non-anomaly node in the same row or column direction as the current sensing node, perform distance-weighted interpolation based on the spatial geometric distance between the nearest non-anomaly node and the current sensing node, and generate a replacement value to fill in the position of the current sensing node.

[0016] S24. Traverse all sensing nodes in the original touch data frame until all sensing nodes have executed steps S21 to S23, and obtain the pre-repaired touch data frame.

[0017] In one embodiment, based on the pre-repair touch data frame and the display refresh status indication signal, the spatial morphology estimation of the display noise and the temporal characteristic parameters of the common-mode interference of the current frame are extracted to obtain noise level data, including:

[0018] Based on the display refresh status indicator signal, the display row scans the currently active area. For the sensing row data in the pre-repair touch data frame that corresponds to the display refresh activation status, the histogram statistics of the node values ​​in each column of the row are calculated row by row. The representative value corresponding to the highest frequency interval of the histogram statistics is used as the candidate value of the row display noise baseline.

[0019] The baseline candidate values ​​of row display noise for several consecutive rows within the same half-cycle of display refresh are smoothly fitted to generate the spatial envelope curve of display noise. The sensing rows corresponding to the display refresh blanking period are filled by extrapolation of the envelope curve of adjacent active periods or by the statistical value of historical blanking periods to obtain the display noise estimation map.

[0020] From the pre-repaired touch data frame, select the sensing row whose row display noise baseline candidate value is consistent with the mode of the row and whose data variance of the row is lower than the preset condition as the silent reference row, calculate the average value of all node values ​​in the silent reference row, and obtain the common mode interference intensity estimate.

[0021] The common-mode interference intensity estimate of the current frame is combined with the common-mode interference intensity estimates of several previous historical frames to form a time series. The first difference and short-term fluctuation variance of the time series are calculated to obtain the common-mode interference time-domain activity index.

[0022] The root mean square value is calculated based on the residual sequence of each node value in the silent reference row after removing the in-row trend, and the noise floor level is estimated.

[0023] In one embodiment, a dynamic confidence field is constructed based on pre-repaired touch data frames and noise level data, including:

[0024] For each sensing node in the pre-repaired touch data frame, the normalized local entropy is calculated in the local spatial neighborhood centered on the sensing node, and the absolute value of the difference between the sensing node value and the corresponding position value in the display noise estimation map is calculated as the local de-display residual. The normalized local entropy and the normalized ratio of the local de-display residual to the estimated value of the background noise level are combined to obtain the spatial consistency factor.

[0025] For each sensing node in the pre-repaired touch data frame, the historical data values ​​of the sensing node after pre-repair processing in several consecutive historical frames are obtained to form a time series sample. The time series sample is used to predict the theoretical expected value of the sensing node in the current frame through an autoregressive moving average model. The recent fluctuation standard deviation of the time series sample is calculated. The time continuity factor is obtained based on the degree of deviation between the absolute value of the deviation between the actual observed value and the theoretical expected value of the current frame and the recent fluctuation standard deviation.

[0026] A preliminary reconstruction based on spatial consistency factor and temporal continuity factor is performed on the pre-repaired touch data frame to obtain a coarsely denoised intermediate image. Connectivity analysis is performed on the coarsely denoised intermediate image to extract candidate touch connected regions. The similarity between the geometric morphological parameters of each candidate touch connected region and the reasonable morphological parameter range defined by the finger touch physical model is calculated. The morphological likelihood function value of the candidate touch connected region to which each sensing node belongs is used as the morphological constraint factor.

[0027] The spatial consistency factor, temporal continuity factor, and morphological constraint factor are synthesized and normalized through point-by-point multiplication at each sensing node to obtain the dynamic confidence field.

[0028] In one embodiment, a hybrid reconstruction process combining anisotropic diffusion and local surface fitting is performed on the pre-repaired touch data frame based on a dynamic confidence field to obtain a high signal-to-noise ratio touch signal reconstruction frame, including:

[0029] S61. Use the pre-repaired touch data frame as the current reconstruction frame;

[0030] S62. For each sensing node in the current reconstructed frame, calculate the local gradient vector in the spatial neighborhood. Based on the magnitude of the local gradient vector and the corresponding sensing node confidence value extracted from the dynamic confidence field, construct a joint diffusion control function. The joint diffusion control function is composed of the edge stopping function multiplied by the confidence modulation factor.

[0031] S63. Perform an anisotropic diffusion filter on the current reconstructed frame based on the joint diffusion control function to obtain the diffusion intermediate frame;

[0032] S64. In the diffusion intermediate frame, for the continuous region in the dynamic confidence field where the confidence value is lower than the dynamic reference, search for the reliable nodes with a confidence value higher than the dynamic reference. Based on the diffusion intermediate frame values ​​and spatial coordinates of the reliable nodes, perform moving least squares local polynomial surface fitting. Use the fitted surface to perform interpolation estimation on the nodes in the low confidence region to obtain the surface fitting repair frame.

[0033] S65. Use the surface fitting and repair frame as the updated current reconstruction frame, and repeat S62 to S64 until the overall change of the reconstruction frame between two consecutive iterations is lower than the convergence tolerance or the preset maximum number of iterations is reached. Then, determine the current reconstruction frame when the iteration stops as the high signal-to-noise ratio touch signal reconstruction frame.

[0034] Secondly, this application also provides a noise suppression system for high signal-to-noise ratio touchscreen signals, used to implement the noise suppression method for high signal-to-noise ratio touchscreen signals as provided in the first aspect, comprising:

[0035] The pre-repair module is used to acquire the original touch data frame obtained by the touch screen sensing array in the current scan frame, and perform isolated singularity pre-repair on the original touch data frame based on neighborhood median statistics based on the display refresh status indication signal aligned with the timing of the current scan frame to obtain the pre-repaired touch data frame.

[0036] The noise analysis module is used to extract the spatial morphology estimation of display noise and the temporal characteristic parameters of common-mode interference of the current frame based on the pre-repaired touch data frame and the display refresh status indication signal, so as to obtain noise level data. The noise level data includes display noise estimation map, common-mode interference intensity estimation value and noise floor level estimation value.

[0037] The confidence field construction module is used to construct a dynamic confidence field that integrates spatial consistency factor, temporal continuity factor and morphological constraint factor based on pre-repaired touch data frames and noise level data. The dynamic confidence field is a two-dimensional weight matrix of the same size as the touch screen sensing array, and each weight value represents the degree of confidence of the corresponding sensing node data in reflecting the real touch state.

[0038] The hybrid reconstruction module is used to perform hybrid reconstruction processing on the pre-repaired touch data frame by combining anisotropic diffusion and local surface fitting based on the dynamic confidence field, so as to obtain a high signal-to-noise ratio touch signal reconstruction frame.

[0039] The touch positioning module is used to reconstruct the frame based on the high signal-to-noise ratio touch signal and the pre-established silent baseline reference value, extract the net capacitance change matrix caused by the touch, and perform local extremum search and sub-pixel coordinate calculation on the net capacitance change matrix to obtain the touch position coordinate information of the current frame.

[0040] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a noise suppression method for a high signal-to-noise ratio touchscreen signal as provided in the first aspect.

[0041] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a noise suppression method for a high signal-to-noise ratio touchscreen signal as provided in the first aspect.

[0042] The aforementioned method, system, computer device, and storage medium for noise suppression of high signal-to-noise ratio touchscreen signals achieves this by using time-series alignment based on display refresh status indication signals and neighborhood median statistics to pre-repair isolated singular points in the original touch data frames, eliminating interference from random glitches and laying a clean data foundation for subsequent processing. It extracts the spatial morphology of display noise and the temporal characteristics of common-mode interference, accurately quantifying the noise levels of these two core types of interference, thus solving the problem of accurately representing mixed non-stationary noise in the background. Furthermore, it constructs a dynamic confidence field that integrates spatial consistency, temporal continuity, and morphological constraints, dynamically defining the reliability of data from each sensing node using a two-dimensional weight matrix. This method effectively distinguishes between large-area background interference and weak touch signals, solving the problems of traditional solutions being unable to dynamically adapt to noise changes and easily confusing interference with weak signals. Through hybrid reconstruction of anisotropic diffusion and local surface fitting, residual noise is further suppressed while preserving the edge details of the touch signal. This avoids the response delay caused by temporal smoothing filtering and overcomes the insufficient adaptability of static template cancellation to noise distortion. Based on the extreme value search and sub-pixel calculation of net capacitance change, accurate touch positioning is achieved on the basis of high signal-to-noise ratio signals. The above methods can achieve synergistic optimization of noise suppression, weak signal detection and low-latency response in complex electromagnetic environments, comprehensively improving the signal-to-noise ratio and interaction reliability of the touch system. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 A flowchart illustrating a noise suppression method for high signal-to-noise ratio touch signals of a touchscreen provided in one embodiment of this application. Figure 1 ;

[0045] Figure 2A flowchart illustrating a noise suppression method for high signal-to-noise ratio touch signals of a touchscreen provided in one embodiment of this application. Figure 2 ;

[0046] Figure 3 This is a schematic diagram of a noise suppression system for a high signal-to-noise ratio touchscreen signal provided in one embodiment of this application. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] First, a brief introduction to the terms used in the embodiments of this application will be given.

[0049] In capacitive touchscreen systems, the touch signal refers to the change in mutual capacitance between the sensing electrodes caused by a user's finger approaching or touching the screen surface. This change is typically extremely weak, and its amplitude depends on the coupling area and distance between the finger and the electrode, as well as the dielectric properties of the panel. The signal-to-noise ratio (SNR) is a dimensionless indicator that measures the relative strength of the useful signal to the background noise. A high SNR means that the capacitance change caused by touch can be clearly identified from chaotic circuit fluctuations and external interference, which is a fundamental physical prerequisite for ensuring touch accuracy, sensitivity, and multi-touch separation capabilities. In portable devices, due to the dense coexistence of modules such as the display, charger, and RF antenna, the touch SNR is easily degraded, leading to problems such as jitter and broken lines.

[0050] Display noise specifically refers to a type of periodic electromagnetic interference generated by the display driver circuit during operation and coupled to the touchscreen's sensing electrodes through parasitic capacitance. In modern high-resolution, high-refresh-rate display panels, the display driver chip scans and refreshes pixel voltages line by line at extremely high frequencies. The large amount of AC components generated during this process is injected into the touch receiver circuit through the tiny capacitive gap between the display and the touchscreen. Spatially, display noise typically exhibits a banded or gradient distribution aligned with the display scanning direction; temporally, it is strictly synchronized with the display refresh cycle, making it one of the most significant sources of interference for mobile terminal touchscreens.

[0051] Common-mode interference refers to interference signal components with essentially the same amplitude and phase that simultaneously act on all or most sensing nodes in a touchscreen's sensing array. Its sources include power ripple from the charger adapter output, fluctuations in the device's grounding level, and power frequency electromagnetic fields in the external environment. Unlike display noise, which has a specific spatial distribution, common-mode interference manifests as an overall rise or fall in the entire touch data frame. Since the actual touch signal typically only affects a local area, common-mode components can theoretically be suppressed through spatial differential methods. However, when the amplitude of common-mode interference far exceeds that of the touch signal, the limited dynamic range of the receiving circuit can suppress or even truncate the useful signal, constituting a major cause of performance degradation in touch systems during charging.

[0052] Based on the above definitions, the implementation environment of the noise suppression method for high signal-to-noise ratio touchscreen signals provided in this application embodiment will be described. Indicatively, the implementation environment includes: a terminal, sensors, and a processor. The terminal is any electronic device integrating a capacitive touchscreen, including but not limited to smartphones, tablets, laptops, in-vehicle infotainment terminals, industrial control touchscreen all-in-one machines, and wearable smart devices. The sensors include, but are not limited to, capacitive touchscreen sensing arrays and display synchronization signal interfaces. The capacitive touchscreen sensing array consists of a matrix of mutual capacitance sensing nodes formed by alternating row and column electrodes, used to convert the spatial electric field changes caused by user finger touches into raw capacitance change data corresponding to each node. The display synchronization signal interface is used to obtain a display refresh status indication signal aligned with the current touch scan frame timing from the display driving module. This indication signal at least includes vertical synchronization phase information and a display row scan activation area identifier. The processor includes, but is not limited to, a central processing unit, a multi-core processor, or an artificial intelligence chip, etc., and is not limited here.

[0053] Based on the above explanations of terms and implementation environments, the application scenarios of the embodiments of this application will be described. The noise suppression method for high signal-to-noise ratio touchscreen signals provided in the embodiments of this application can be applied to scenarios including but not limited to the following:

[0054] In mobile terminal charging scenarios, the non-stationary common-mode power frequency interference introduced by the charger adapter and the periodic coupled noise generated by the display driver superimpose each other, causing common problems such as point jitter, trajectory breakage, and even false triggering of the touch screen during charging operations. The method provided in this application separates and extracts the spatial morphology of display noise and the intensity of common-mode interference, and constructs a dynamic confidence field to guide signal reconstruction. It can effectively suppress dual noise components from the charger and the display screen during charging, while maintaining the response sensitivity during light touch and rapid swipe, significantly improving the smoothness and accuracy of user interaction behaviors such as text input and game operation during charging.

[0055] In outdoor portable device applications, touchscreens may be simultaneously affected by multiple factors, including baseline drift caused by changes in ambient temperature and humidity, transmission interference from the radio frequency communication module, and increased coupling noise due to enhanced high brightness of the display under direct sunlight. The method provided in this application dynamically tracks noise characteristics and adaptively updates the confidence field, enabling it to cope with real-time changes in the noise environment without relying on fixed filtering parameters. This allows the device to maintain stable touch response capabilities even under complex outdoor conditions, effectively reducing the adverse effects of environmental factors on touch performance and improving the user experience in outdoor navigation, photography, and other scenarios.

[0056] In in-vehicle infotainment systems, touchscreens need to operate stably in complex electromagnetic environments, including broadband electromagnetic interference generated by the vehicle's electrical system, engine ignition pulses, and crosstalk from multiple displays. Because in-vehicle touchscreens are large and operators often wear gloves, the capacitance changes caused by touch are relatively weak, making the signal-to-noise ratio requirements even more stringent. The method provided in this application utilizes a multi-dimensional fusion confidence assessment mechanism based on spatial local entropy, temporal continuity, and morphological constraints. This mechanism can accurately identify the weak edge features of real touch signals in strong background noise, ensuring the accuracy of point reporting and the robustness against interference for the in-vehicle touch system during driving, and preventing false triggering or touch malfunctions caused by noise.

[0057] As an illustration, the noise suppression method for high signal-to-noise ratio touch screen touch signals provided in this application embodiment can also be applied to other application scenarios. This is only an example and does not limit the specific application scenarios.

[0058] In one exemplary embodiment, such as Figure 1 As shown, a noise suppression method for high signal-to-noise ratio touchscreen signals is provided. This embodiment illustrates the application of this method to a terminal in the aforementioned implementation environment. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps 101 to 105:

[0059] Step 101: Obtain the original touch data frame obtained by the touch screen sensing array in the current scan frame. Based on the display refresh status indication signal that is aligned with the timing of the current scan frame, perform isolated singularity pre-repair on the original touch data frame based on neighborhood median statistics to obtain the pre-repaired touch data frame.

[0060] Specifically, this method can acquire the original touch data frame of the current scan frame by scanning the touch screen sensing array row by row and column by column. The original touch data frame is a two-dimensional data matrix of capacitance change corresponding to each sensing node in the sensing array. Its acquisition process is synchronized with the refresh timing of the display screen to ensure that each frame of original data can match the corresponding display working state.

[0061] Furthermore, to achieve precise timing alignment, the method receives a display refresh status indication signal output by the display driver chip. This signal includes key timing identifiers such as the display frame start pulse and the line scan enable signal. By locking the frame start time of the touch scan with the frame start time of the display refresh, each sampling period of the current scan frame corresponds one-to-one with a specific stage of the display refresh. Based on the aforementioned timing alignment results, this method performs isolated singularity pre-repair on the original touch data frame based on neighborhood median statistics. This utilizes the spatial continuity of the real touch signal to determine whether the signal value of each target sensing node deviates significantly from the signal values ​​of its neighboring nodes within a preset neighborhood. For example, four neighboring nodes (top, bottom, left, and right) or eight neighboring nodes within a 3×3 neighborhood are selected to form a statistical neighborhood. The median value is then taken after sorting the signal values ​​of all nodes within the neighborhood. If the difference between the target node's signal value and the median value exceeds a preset reasonable fluctuation range, the node is determined to be an isolated singularity, and the median value is used to replace and repair it. Through this process, discrete noise points caused by random electromagnetic pulses, instantaneous circuit jitter, etc., can be quickly eliminated, while preserving the spatial distribution characteristics of the original touch signal to the greatest extent, resulting in a pre-repaired touch data frame with preliminary noise suppression.

[0062] Step 102: Based on the pre-repaired touch data frame and the display refresh status indication signal, extract the display noise spatial morphology estimation and common-mode interference temporal characteristic parameters of the current frame to obtain noise level data. The noise level data includes the display noise estimation map, the common-mode interference intensity estimation value, and the noise floor level estimation value.

[0063] Specifically, after obtaining the pre-repaired touch data frame, the method combines the display refresh status indication signal that has achieved time alignment to extract the display noise spatial morphology estimation and common-mode interference temporal feature parameters of the current frame, and then integrates them to form noise level data that includes display noise estimation map, common-mode interference intensity estimation value and noise floor level estimation value. In the process of estimating the spatial morphology of display noise, this method can use the display refresh status indicator signal to determine the overlapping period between the current touch scan frame and the display row scan and column drive. Based on the distribution differences of the signals of each node in the pre-repaired touch data frame within this period, the spatial distribution law of the noise formed by the coupling of the display drive signal is inferred on the sensing array through spatial correlation analysis. For example, by comparing the signal change trends of the touch nodes under different display row drive periods, a display noise estimation map of the same size as the sensing array is constructed. Each element in the map represents the amplitude estimate of the corresponding node affected by the display noise. In the extraction of common-mode interference time-domain feature parameters, this method continuously collects the signal mean of all nodes in multiple pre-repaired touch data frames and analyzes the fluctuation characteristics of the mean through a time-domain sliding window. For example, the fundamental frequency and amplitude peak value of the fluctuation are extracted by using a moving average filter to obtain the common-mode interference intensity estimate. At the same time, by statistically analyzing the signal variance of multiple frames of data in the no-touch state, the system noise floor level estimate is determined, thereby achieving accurate quantification of the two types of core interference.

[0064] Step 103: Construct a dynamic confidence field that integrates spatial consistency factor, temporal continuity factor and morphological constraint factor based on the pre-repaired touch data frame and noise level data; wherein, the dynamic confidence field is a two-dimensional weight matrix of the same size as the touch screen sensing array, and each weight value represents the degree of confidence of the corresponding sensing node data in reflecting the real touch state.

[0065] Specifically, this method constructs a dynamic confidence field that integrates spatial consistency factors, temporal continuity factors, and morphological constraint factors based on pre-repaired touch data frames and the aforementioned noise level data. This dynamic confidence field is a two-dimensional weight matrix of the same size as the touch screen sensing array. Each weight value in the matrix quantitatively represents the degree of credibility of the corresponding sensing node data in reflecting the real touch state. The spatial consistency factor is calculated based on the pre-repaired touch data frame. By analyzing the signal similarity between the target node and its neighboring nodes, and combining the noise amplitude of the node in the display noise estimation map, a higher spatial consistency factor is assigned if the target node signal and its neighboring signals have the same trend and the noise amplitude is low. The temporal continuity factor is calculated by comparing the current frame's pre-repaired data with the previous frame's effective data after noise suppression. The temporal change rate of the target node signal is calculated, and the common-mode interference intensity estimate is used to determine whether the change conforms to the temporal evolution law of real touch. If the change rate is within a reasonable range and is unrelated to common-mode interference fluctuations, the temporal continuity factor is assigned a higher value. The morphological constraint factor is based on the inherent spatial morphological characteristics of the capacitance change signal generated by real touch (such as a Gaussian distribution trend with strong center and weak edge). The spatial distribution morphology of the target node signal is analyzed for matching degree. If the signal morphology has a high degree of fit with the preset touch signal morphology template, the morphological constraint factor is assigned a higher value. For example, this method can integrate the above three factors into a final weight value by means of weighted summation, etc. The weight value ranges from 0 to 1. The closer the weight value is to 1, the more reliable the corresponding node data is. Conversely, the more likely it is to be affected by noise, thereby realizing the dynamic definition of the boundary between signal and noise.

[0066] Step 104: Based on the dynamic confidence field, perform hybrid reconstruction processing on the pre-repaired touch data frame by combining anisotropic diffusion and local surface fitting to obtain a high signal-to-noise ratio touch signal reconstruction frame.

[0067] Specifically, in the anisotropic diffusion processing stage, this method uses the weight values ​​of the dynamic confidence field as adjustment factors for the diffusion coefficient. For node regions with high weight values ​​(i.e., reliable data), a lower diffusion intensity is used to preserve signal details, while for node regions with low weight values ​​(i.e., potentially affected by noise), a higher diffusion intensity is used to suppress noise. Simultaneously, the diffusion direction is adjusted according to the local gradient direction of the signal to ensure that diffusion occurs only along directions parallel to the edge in the signal edge region, avoiding edge blurring. In the local surface fitting processing stage, for the signal after anisotropic diffusion, a weighted surface fitting is performed on the neighborhood data of each target node using the dynamic confidence field as weights. For example, a quadratic surface model is constructed using the least squares method. Neighboring nodes with higher weight values ​​contribute more to the fitting result. This fitting can compensate for the distortion of local weak signals, improving the continuity and integrity of the signal. Through the synergistic effect of these two processing methods, both deep suppression of residual noise and complete preservation of the edge details and subtle variation characteristics of the real touch signal are achieved, effectively improving the signal-to-noise ratio of the reconstructed signal.

[0068] Step 105: Based on the high signal-to-noise ratio touch signal reconstructed frame and the pre-established silent baseline reference value, extract the net capacitance change matrix caused by touch, and perform local extremum search and sub-pixel coordinate calculation on the net capacitance change matrix to obtain the touch position coordinate information of the current frame.

[0069] Specifically, the silent baseline reference value is the reference value of each sensing node obtained by continuously collecting multiple frames of raw data and performing statistical averaging when the touch screen is in a silent state without any touch operation. This reference value already includes fixed interference components such as the inherent background noise of the system and static offset of the circuit. By subtracting the signal value of each node in the touch signal reconstruction frame from the corresponding silent baseline reference value, the fixed interference components can be eliminated, and the net capacitance change matrix caused only by touch operation can be obtained.

[0070] Optionally, this method locates the core region with the largest capacitance change by performing a local extremum search on the net capacitance change matrix. This region is the approximate location range of the touch operation. Then, based on the signal values ​​of multiple nodes within this core region, an interpolation algorithm or a surface fitting algorithm is used to calculate the sub-pixel coordinates. For example, by performing Gaussian surface fitting on the signals in the core region, the peak position of the fitted surface is used as the sub-pixel coordinates of the touch point to obtain the accurate touch position coordinate information of the current frame.

[0071] The noise suppression method for high signal-to-noise ratio touchscreen signals described above eliminates interference from random glitches by performing isolated singularity pre-repair on the original touch data frames through time-series alignment based on the display refresh status indicator signal and neighborhood median statistics, thus laying a clean data foundation for subsequent processing. It extracts the spatial morphology of display noise and the temporal features of common-mode interference, accurately quantifying the noise levels of these two core types of interference, solving the problem of accurately representing mixed non-stationary noise in the background. Furthermore, it constructs a dynamic confidence field that integrates spatial consistency, temporal continuity, and morphological constraints, dynamically defining the confidence level of data from each sensing node through a two-dimensional weight matrix, effectively distinguishing large-area... This approach addresses the issues of traditional solutions failing to dynamically adapt to noise changes and easily confusing interference with weak signals by using a hybrid reconstruction method combining anisotropic diffusion and local surface fitting. This method preserves edge details of the touch signal while further suppressing residual noise, avoiding response delays caused by temporal smoothing filtering and overcoming the insufficient adaptability of static template cancellation to noise distortion. Based on extreme value search and sub-pixel calculation of net capacitance change, precise touch positioning is achieved on a high signal-to-noise ratio (SNR) basis. These methods enable synergistic optimization of noise suppression, weak signal detection, and low-latency response in complex electromagnetic environments, comprehensively improving the SNR and interaction reliability of the touch system.

[0072] In one embodiment, such as Figure 2 As shown, the original touch data frame obtained by the touchscreen sensing array in the current scan frame is acquired. Based on the display refresh status indication signal aligned with the timing of the current scan frame, the original touch data frame is pre-repaired for isolated singularities based on neighborhood median statistics to obtain a pre-repaired touch data frame, including:

[0073] S21. Obtain the original touch data frame. For each sensing node in the original touch data frame, extract the values ​​of all nodes within the spatial neighborhood window centered on the current sensing node, and calculate the median and median absolute deviation of the values ​​within the spatial neighborhood window.

[0074] Specifically, this method can achieve accurate pre-repair of isolated singular points in the original touch data frame through a step-by-step neighborhood statistics and interpolation repair process, ensuring that the pre-repaired data retains the true signal characteristics while effectively eliminating discrete noise interference.

[0075] For example, the raw touch data frame obtained by this method is the basic data collected by the touch screen sensing array within the complete scan cycle of the current scan frame. It captures the electrical signals related to capacitance changes through each sensing node in the array, and arranges them into a two-dimensional matrix according to their spatial positions. Each element in the matrix... Corresponding coordinates are The raw data collected by the sensing nodes, among which For the row index of the sensing node, For each sensing node in the original touch data frame, the method uses the current sensing node as the column index; A spatial neighborhood window is defined around the center. The selection of the spatial neighborhood window can be flexibly set according to the node density and noise distribution characteristics of the touch screen sensing array. For example, a 3×3 or 5×5 square window can be selected. The range of the window can be represented as follows: ,in Half the width of the window (when the window is 3x3) When it is 5×5 This ensures the window covers spatially related neighboring nodes around the current node, avoiding insufficient statistical samples due to an excessively small window or the introduction of irrelevant node data due to an excessively large window. Extract the values ​​of all nodes within this spatial neighborhood window. Then, the values ​​within the window are sorted in ascending order using a sorting algorithm to obtain an ordered numerical sequence. (in This represents the total number of nodes within the window. ), take the value at the middle position as the median of the neighborhood, when When the median is odd, ,when When the median is even, This median reflects the central tendency of the signal within the window and is less affected by extreme outliers; at the same time, it calculates the absolute value of the deviation of all values ​​within the window from this median. Then, these absolute deviation values ​​are sorted in ascending order to obtain a sequence of absolute deviation values. Similarly, the value at the middle position is taken as the median absolute deviation (MAD), which is calculated using the following formula:

[0076]

[0077] in, The median of the absolute deviation. The th in the absolute value sequence of deviations One element, This represents the total number of nodes within the spatial neighborhood window. For the sequence index, the median of the absolute deviation It can quantify the dispersion of data within a window, and its value directly reflects the stability of the signal in the current neighborhood, providing an objective and adaptive reference for subsequent anomaly point judgment, so as to overcome the shortcomings of traditional fixed thresholds that cannot adapt to different noise scenarios.

[0078] S22. Calculate the absolute value of the deviation between the original value of the current sensing node and the median. When the absolute value of the deviation exceeds the dynamic elastic boundary determined based on the median of the absolute deviation, the current sensing node is determined to be an anomaly.

[0079] Specifically, this method calculates the absolute value of the deviation between the original value of the current sensing node and the neighborhood median obtained in step S21. Its core purpose is to quantify the degree of deviation between the current node and the surrounding normal nodes through numerical differences.

[0080] For example, the method can calculate the absolute value of the deviation between the original value of the current sensing node and the neighborhood median obtained in step S21 using the following formula:

[0081]

[0082] in, For the current sensing node The absolute value of the deviation between the original value and the neighborhood median. For the current sensing node The original collected values, The median of the values ​​within the spatial neighborhood window calculated in step S21, and the absolute value of this deviation. The value is used to quantify the degree of deviation of the current node from the surrounding normal nodes. The larger the value, the more likely the current node is to be an anomaly point affected by noise.

[0083] Furthermore, to avoid the problem of insufficient adaptability caused by using a fixed threshold, this method can also be based on the median of absolute deviation. Determine the dynamic elastic boundary. The essence of the dynamic elastic boundary is to adaptively adjust the anomaly judgment criteria based on the discrete characteristics of the current neighborhood data. Its calculation is achieved through the following formula:

[0084]

[0085] in, The dynamic elastic boundary for anomaly point determination. The preset adjustment coefficient (the value can be set to 3~4, and can be adaptively calibrated according to the noise intensity of the actual application scenario, for example, 4 in a strong electromagnetic interference environment and 3 in a normal environment). This is the median of the absolute deviation calculated in step S21. This is the dynamic elastic boundary. It can dynamically change with the dispersion of neighborhood data, when the signal in the neighborhood is stable ( Hour, The corresponding reduction allows for precise identification of minor anomalies; when signal fluctuations within the neighborhood are large ( When (large), The value is increased accordingly to avoid misinterpreting normal signal edge changes. If the absolute value of the deviation of the current sensing node... Exceeding the dynamic elastic boundary ,Right now When the node is disturbed by random noise, it can be determined that the node is an anomaly. This step achieves adaptive identification of anomalies through dynamic thresholds, which is more robust than traditional fixed threshold judgment and effectively solves the problem of low accuracy of anomaly identification under different noise scenarios.

[0086] S23. For the current sensing node that is determined to be an anomaly, obtain the value of the nearest non-anomaly node in the same row or column direction as the current sensing node, perform distance-weighted interpolation based on the spatial geometric distance between the nearest non-anomaly node and the current sensing node, and generate a replacement value to fill in the position of the current sensing node.

[0087] Specifically, for the current sensing node identified as an anomaly... This method works by moving in the same direction as the current sensing node ( fixed, Changes) and same column ( fixed, The search process involves searching for the nearest non-abnormal node in both directions of the change. The search proceeds from nearest to farthest, checking adjacent nodes in the same row or column until a node not identified as an anomalous point is found. If the nearest non-abnormal node is found in the same row direction... Find the nearest non-abnormal node in the same column direction. Then obtain the values ​​of these two nodes. and The selection of the same row or column direction is based on the continuity characteristics of the actual touch signal in the horizontal and vertical directions, ensuring that the nearest non-anomaly node found can reflect the actual signal trend at the current location. Distance-weighted interpolation is performed based on the spatial geometric distance between the nearest non-anomaly node and the current anomaly point. The core principle of distance-weighted interpolation is that the closer a non-anomaly node is to the current anomaly point, the higher its reference value for the anomaly point, and the larger its weight. For example, the interpolation calculation can be achieved through the following formula:

[0088]

[0089] in, The current anomaly The replacement value for the repair The current anomaly Nearest neighbor non-abnormal node in the same column Spatial geometric distance (i.e., the absolute value of the row index difference) ), The current anomaly Nearest neighbor non-abnormal node in the same row Spatial geometric distance (i.e., the absolute value of the column index difference) ), The nearest neighbor non-abnormal node in the same column The value, The nearest neighbor non-abnormal node in the same row The numerical value. If only one nearest non-abnormal node is found in the same row or column, linear interpolation is used to simplify the calculation, for example, only finding the nearest non-abnormal node in the same row. hour, After obtaining the replacement value through the distance-weighted interpolation, the replacement value is filled into the position of the current sensing node to achieve accurate repair of the abnormal point. Compared with directly replacing with the neighborhood median, this step can better preserve the spatial gradient characteristics of the signal, avoid the problem of signal abrupt change or loss of details after repair, and ensure that the pre-repaired data is more in line with the change law of the real touch signal in terms of spatial distribution.

[0090] S24. Traverse all sensing nodes in the original touch data frame until all sensing nodes have executed steps S21 to S23, and obtain the pre-repaired touch data frame.

[0091] Specifically, this method traverses all sensing nodes in the original touch data frame in either row-major or column-major order. The traversal order can be flexibly selected according to the data storage format; for example, when using row-major order, it follows... From 0 to , From 0 to Process each node in the following order (where This represents the total number of rows in the sensor array. (Total number of columns) For each node, the processing steps S21 to S23 are executed sequentially: first, the neighborhood window is extracted and the median and median absolute deviation are calculated; then, it is determined whether it is an outlier. If it is an outlier, it is repaired using distance-weighted interpolation; if it is a normal point, the original value is retained. This process continues until all nodes have completed outlier detection and repair, resulting in a pre-repaired touch data frame. This pre-repaired touch data frame not only effectively eliminates isolated singularities caused by random electromagnetic pulses and transient circuit jitter through adaptive judgment and precise interpolation, but also preserves the spatial distribution pattern and edge details of the original touch signal to the greatest extent through the collaborative design of spatial neighborhood statistics and distance-weighted interpolation. Its core significance lies in providing a clean and reliable data foundation for subsequent steps such as spatial morphology estimation of display noise and extraction of common-mode interference time-domain feature parameters, avoiding interference from discrete outliers on the accuracy of noise feature quantization. If there are a large number of isolated singularities in the original data, outliers may be mistakenly included in the noise components during noise feature extraction, thus affecting the accuracy of the dynamic confidence field construction and reducing the effect of hybrid reconstruction. Therefore, this pre-repair process directly ensures the reliability of all subsequent signal processing flows, laying a crucial foundation for improving the robustness of the entire noise suppression method and the fidelity of the touch signal.

[0092] In one embodiment, based on the pre-repair touch data frame and the display refresh status indication signal, the spatial morphology estimation of the display noise and the temporal characteristic parameters of the common-mode interference of the current frame are extracted to obtain noise level data, including:

[0093] Based on the display refresh status indicator signal, the currently active area of ​​the display row is scanned. For the sensing row data in the pre-repair touch data frame that corresponds to the display refresh activation status, the histogram statistics of the node values ​​in each column of the row are calculated row by row. The representative value corresponding to the highest frequency interval of the histogram statistics is used as the candidate value of the row display noise benchmark.

[0094] Specifically, when extracting the spatial morphology estimation of display noise and the temporal feature parameters of common-mode interference in the current frame, this method first relies on the display refresh status indication signal, which is aligned with the timing of the pre-repaired touch data frame, to accurately locate the currently active area of ​​the display row scan. The display refresh status indication signal contains a row scan enable signal and a row index synchronization identifier. This signal allows for the real-time determination of the sensing row range currently in the display drive active state, ensuring that the subsequent noise extraction and display refresh timing are perfectly matched, avoiding noise estimation deviations caused by timing misalignment. Furthermore, the method targets the display refresh activation state in the pre-repaired touch data frame. Based on the corresponding sensing row data, this method performs histogram statistics on the values ​​of all column nodes in each row. Specifically, it divides the values ​​of all nodes in a single sensing row into several continuous value intervals, counts the frequency of occurrence of node values ​​in each interval, and determines the interval with the highest frequency as the highest frequency interval. The median or the value that appears most frequently in the interval is selected as the candidate value for the row display noise benchmark, so that the display noise shows a concentrated distribution characteristic in a single sensing row. The highest frequency interval of the histogram can accurately reflect the core level of display noise interference in the row, providing a row-by-row noise benchmark for subsequent spatial morphology estimation.

[0095] The baseline candidate values ​​of row display noise for several consecutive rows within the same display refresh half-cycle are smoothly fitted to generate a display noise spatial envelope curve. The sensing rows corresponding to the display refresh blanking period are filled by extrapolation of the envelope curve of the adjacent activation period or by the statistical value of the historical blanking period to obtain the display noise estimation map.

[0096] Specifically, to obtain a complete spatial pattern of display noise, this method smoothly fits the baseline candidate values ​​of row display noise for several consecutive rows within the same display refresh half-cycle. The smoothing fit can employ moving average or low-pass filtering algorithms. By trend-based processing of the candidate values ​​for consecutive rows, the influence of random fluctuations between rows on the noise pattern is eliminated, generating a display noise spatial envelope curve that continuously changes along the sensing row direction. This curve accurately characterizes the spatial distribution trend of display noise. Furthermore, for the sensing row corresponding to the display refresh blanking period (i.e., the sensing row where the display driver is inactive), since there is no direct display driver coupling noise, this method linearly extrapolates the spatial envelope curves of display noise from adjacent active periods, or fills them with the statistical mean of noise from the same blanking period in the most recent few frames. This ensures the continuity of the display noise estimation map between the active and blanking periods, ultimately forming a display noise estimation map of the same size as the pre-repaired touch data frame. The value of each node in this map is the estimated display noise value at the corresponding location, fully presenting the spatial distribution characteristics of the display noise.

[0097] From the pre-repaired touch data frame, select the sensing row whose row display noise baseline candidate value is consistent with the mode of the row and whose data variance of the row is lower than the preset condition as the silent reference row. Calculate the average value of all node values ​​in the silent reference row to obtain the estimated value of common mode interference intensity.

[0098] Specifically, this method filters silent reference rows from pre-repaired touch data frames. The filtering criteria are that the row display noise baseline candidate value is consistent with the mode within the row, and the variance of the data within the row is lower than a preset condition. The mode within the row refers to the value of the node that appears most frequently in the row. Its consistency with the row display noise baseline candidate value indicates that the data in that row is dominated by display noise and is concentrated in the distribution. The variance of the data within the row being lower than the preset condition means that there is no obvious touch signal interference in that row (touch signals will cause the local data variance to increase). This dual-condition filtering ensures that the silent reference rows contain only noise components and no effective touch signals. After obtaining the silent reference rows, the arithmetic mean of all node values ​​in the row is calculated. For example, the common-mode interference intensity estimate can be obtained using the following formula:

[0099]

[0100] in, This is an estimate of the common-mode interference intensity. This represents the total number of column nodes in a single sensing row. For the first Silent reference line number The pre-repair values ​​of column nodes and the estimated common-mode interference intensity can accurately quantify the overall intensity of power supply common-mode interference in the current frame, providing a core quantitative basis for subsequent interference suppression.

[0101] The common-mode interference intensity estimate of the current frame is combined with the common-mode interference intensity estimates of several previous historical frames to form a time series. The first difference and short-term fluctuation variance of the time series are calculated to obtain the common-mode interference time-domain activity index.

[0102] Specifically, to comprehensively characterize the temporal dynamics of common-mode interference, this method arranges the estimated common-mode interference intensity of the current frame and the estimated common-mode interference intensity of several previous historical frames (such as the previous 5 or 10 frames, which can be set according to actual needs) in chronological order to form a common-mode interference intensity time series. Based on this time series, the first-order difference between the estimated values ​​of adjacent frames is calculated to reflect the instantaneous rate of change of common-mode interference; at the same time, the variance of the time series within the short-term window is calculated to characterize the stability of common-mode interference. The absolute mean of the first-order difference is weighted and fused with the short-term variance to obtain a common-mode interference temporal activity index. This index can effectively distinguish whether common-mode interference is in a stationary state or a rapidly changing state, providing temporal noise feature support for the subsequent construction of a dynamic confidence field.

[0103] The root mean square value is calculated based on the residual sequence of each node value in the silent reference row after removing the in-row trend, and the noise floor level is estimated.

[0104] Specifically, when calculating the estimated noise floor level, this method can extract the intra-row trend of each node value in the silent reference row. The intra-row trend is the gradual change characteristic of the node value along the column direction (such as a linear or non-linear trend). The intra-row trend of the row is removed by polynomial fitting or sliding window trend extraction algorithm to obtain a residual sequence containing only the inherent noise floor of the system. For example, the root mean square value can be calculated based on the residual sequence using the following formula:

[0105]

[0106] in, This is an estimate of the noise level. This represents the total number of column nodes in a single sensing row. For silent reference line number The residual after removing in-row trends from column node values, the root mean square value, can accurately quantify the inherent noise level of the system itself, providing a benchmark threshold for distinguishing between signals and noise, and ensuring that weak touch signals are not misjudged as background noise. Through the above steps, this method ultimately integrates noise level data including display noise estimation map, common-mode interference intensity estimate, common-mode interference temporal activity index, and background noise level estimate. It comprehensively and accurately quantifies the mixed noise characteristics of the current frame, providing targeted noise reference for subsequent dynamic confidence field construction and hybrid reconstruction processing, ensuring that the noise suppression process is more targeted.

[0107] In one embodiment, a dynamic confidence field is constructed based on pre-repaired touch data frames and noise level data, including:

[0108] For each sensing node in the pre-repaired touch data frame, the normalized local entropy is calculated in the local spatial neighborhood centered on the sensing node, and the absolute value of the difference between the sensing node value and the corresponding position value in the display noise estimation map is calculated as the local de-display residual. The normalized local entropy and the normalized ratio of the local de-display residual to the estimated value of the background noise level are combined to obtain the spatial consistency factor.

[0109] Specifically, when constructing the dynamic confidence field, this method can calculate the spatial consistency factor for each sensing node in the pre-repaired touch data frame through spatial dimension feature analysis to quantify the spatial reliability of the node data. For example, the method delineates a local spatial neighborhood (such as a 3×3 or 5×5 window) centered on the current sensing node, calculates the normalized local entropy of all node values ​​within this neighborhood. Local entropy characterizes the dispersion of neighborhood data; the more concentrated the data (less noise interference), the lower the entropy value; the more dispersed the data (potentially containing noise or signal abrupt changes), the higher the entropy value. After normalization, the local entropy range is mapped to the [0,1] interval. Simultaneously, the absolute value of the difference between the pre-repaired value of the current sensing node and the corresponding noise estimate in the display noise estimation map is calculated to obtain the local de-display residual. This residual reflects the strength of the effective components of the node data after removing display noise. The ratio of this local de-display residual to the estimated noise level is calculated, and the ratio result is normalized to obtain the normalized residual value. The spatial consistency factor is obtained by combining the normalized local entropy with the normalized residual value using the following formula:

[0110]

[0111] in, , which is the spatial consistency factor (value range [0,1], the closer to 1, the higher the spatial reliability). , which is the weighting coefficient (with a value range of [0.3, 0.7], used to balance the contribution of local entropy and residual). To normalize the local entropy, To locally remove the normalized value of the residual, this method can take into account the effective signal strength after removing the display noise when considering the spatial concentration of neighborhood data, thus ensuring that the spatial consistency factor can fully reflect the spatial reliability characteristics of the node data.

[0112] For each sensing node in the pre-repaired touch data frame, the historical data values ​​of the sensing node after pre-repair processing in several consecutive historical frames are obtained to form a time series sample. The time series sample is used to predict the theoretical expected value of the sensing node in the current frame through an autoregressive moving average model. The recent fluctuation standard deviation of the time series sample is calculated. The time continuity factor is obtained based on the degree of deviation between the absolute value of the deviation between the actual observed value and the theoretical expected value of the current frame and the recent fluctuation standard deviation.

[0113] Specifically, to verify the reliability of node data from a time perspective, this method calculates a time continuity factor and determines whether the data conforms to the temporal evolution pattern of real touch signals by comparing the historical data trends with the current observations. For example, for each sensing node in the pre-repaired touch data frame, this method can extract the pre-repaired historical data values ​​of that node from several consecutive historical frames (e.g., the first 3 to 8 frames, adaptively set according to the touch refresh rate), and construct a time-series sample in chronological order. (in The current frame number. (This refers to the number of historical frames). Furthermore, an autoregressive moving average (ARMA) model can be used to fit and train the time series sample, and the theoretical expected value of the node in the current frame can be obtained through model parameter estimation. This expected value characterizes the signal prediction based on historical trends; simultaneously, the recent volatility standard deviation of the time series sample is calculated. This is used to quantify the normal fluctuation range of historical data, and its calculation formula is:

[0114]

[0115] in, This represents the recent standard deviation of the time series sample. For the number of historical frames, For the first The pre-repair value for this node in the frame. This is the average value of the time series samples. Then, the actual observed value for that node in the current frame (i.e., the pre-repaired value) is calculated. ) and theoretical expected value absolute value of the deviation between The absolute value of this deviation is compared with the recent standard deviation of the fluctuation. The deviation coefficient is calculated by performing a ratio calculation. If the deviation coefficient is less than a preset threshold (e.g., 1.5), the time continuity factor takes a higher value (close to 1), indicating that the current observation is consistent with the historical trend and has high reliability. If the deviation coefficient exceeds the preset threshold, the time continuity factor decreases as the deviation coefficient increases, indicating that the current observation may be affected by instantaneous interference and has low reliability. Finally, a time continuity factor with a value range of [0,1] is obtained. .

[0116] A preliminary reconstruction based on spatial consistency factors and temporal continuity factors is performed on the pre-repaired touch data frame to obtain a coarsely denoised intermediate image. Connectivity analysis is performed on the coarsely denoised intermediate image to extract candidate touch connected regions. The similarity between the geometric morphological parameters of each candidate touch connected region and the reasonable morphological parameter range defined by the finger touch physical model is calculated. The morphological likelihood function value of the candidate touch connected region to which each sensing node belongs is used as a morphological constraint factor.

[0117] Specifically, to introduce physical form constraints for realistic touch, this method quantifies the fit between node data and the physical model of finger touch through a form constraint factor. First, based on the calculated spatial consistency factor... With time continuity factor A preliminary reconstruction of the pre-repaired touch data frames is performed; for example, the data of each node can be reconstructed by... The weights are locally weighted and smoothed to retain the signal features of high-weight nodes and suppress the interference components of low-weight nodes, resulting in a coarsely denoised intermediate image. This image has initially removed most of the noise and highlighted the potential touch area. Further, connected component analysis is performed on the coarsely denoised intermediate image. Through threshold segmentation (the threshold is set based on an estimated noise level), continuous high-signal-intensity regions are extracted as candidate touch connected regions. Each candidate touch connected region consists of spatially adjacent sensing nodes with signal values ​​higher than the threshold. Geometric morphological parameters of each candidate touch connected region are calculated, including region area, equivalent radius, aspect ratio, and roundness. These parameters describe the spatial morphological characteristics of the region. These geometric morphological parameters are then matched with the reasonable morphological parameter range defined by the finger touch physical model. For example, the equivalent radius of finger touch is usually within a preset range, and the roundness is close to 1 (distributed in a near-circular pattern). By calculating the Euclidean distance or cosine similarity between the actual morphological parameters and the reasonable range, the morphological likelihood function value of each candidate touch connected region is obtained. This value, mapped to the [0,1] interval, serves as the morphological constraint factor for all sensing nodes within that region. If the shape of the candidate region to which the node belongs highly matches the touch physics model, then A value close to 1 indicates that the node data is highly likely to originate from actual touch; if the region shape does not conform to the touch model (e.g., it is elongated, too large, or too small), then... A value close to 0 indicates that the node data may be subject to noise interference.

[0118] The spatial consistency factor, temporal continuity factor, and morphological constraint factor are synthesized and normalized through point-by-point multiplication at each sensing node to obtain the dynamic confidence field.

[0119] Specifically, this method obtains the spatial consistency factor. Time continuity factor and morphological constraint factor Then, point-by-point multiplication synthesis can be performed on each sensing node, and the product of the three factors can be normalized to ensure that the final weight values ​​are in the range [0,1], thus obtaining a dynamic confidence field. For example, this can be achieved through the following formula:

[0120]

[0121] in, Coordinates in the dynamic confidence field The weight values ​​of the sensing nodes, , , These are the spatial consistency factor, temporal continuity factor, and morphological constraint factor of the node, respectively. The maximum value of the product of the three factors of all nodes in the entire pre-repaired touch data frame is the dynamic confidence field, which is the same size as the touch screen sensing array. Each weight value accurately represents the credibility of the corresponding sensing node data in reflecting the real touch state, providing differentiated weight guidance for subsequent hybrid reconstruction processing. Specifically, high-weight nodes will focus on preserving signal details, while low-weight nodes will strengthen noise suppression, thereby achieving a precise balance between noise suppression and weak signal fidelity.

[0122] In the above embodiments, the spatial consistency factor verifies whether nodes conform to the spatial continuity characteristics of the signal from the perspective of local data distribution; the temporal continuity factor verifies whether nodes conform to the temporal stability law of the signal from the perspective of historical data trends; and the morphological constraint factor verifies whether nodes belong to a reasonable morphological region of real touch from the perspective of physical model matching. The three factors respectively cover the spatial, temporal, and physical attributes of the signal, forming a complementary verification system. By fusing the three factors through multiplication, noise nodes that are reliable in a single dimension but contradictory in multiple dimensions (such as interference nodes with normal spatial distribution but abrupt temporal changes) can be effectively filtered out, ensuring that only nodes that truly conform to the characteristics of real touch signals receive high weights. The global normalization process, by using the maximum fusion weight of all nodes as a benchmark, unifies the weights of all nodes to the [0,1] interval, which not only avoids fusion deviations caused by differences in factor numerical ranges but also gives the weight values ​​a clear physical meaning (i.e., the proportion relative to the maximum reliability), providing a unified standard for precise control of processing intensity in the subsequent hybrid reconstruction process, ensuring the scientific nature and operability of the entire noise suppression process.

[0123] In one embodiment, a hybrid reconstruction process combining anisotropic diffusion and local surface fitting is performed on the pre-repaired touch data frame based on a dynamic confidence field to obtain a high signal-to-noise ratio touch signal reconstruction frame, including:

[0124] S61. Use the pre-repaired touch data frame as the current reconstruction frame.

[0125] Specifically, when performing hybrid reconstruction, the method uses the pre-repaired touch data frame as the initial current reconstruction frame. This initial frame has completed the pre-repair of isolated singularities, providing a preliminary cleaned data foundation for hybrid reconstruction and ensuring that subsequent diffusion and fitting processes can focus on suppressing residual hybrid noise and preserving weak signals.

[0126] S62. For each sensing node in the current reconstructed frame, calculate the local gradient vector in the spatial neighborhood. Based on the magnitude of the local gradient vector and the corresponding sensing node confidence value extracted from the dynamic confidence field, construct a joint diffusion control function. The joint diffusion control function is composed of the edge stopping function multiplied by the confidence modulation factor.

[0127] Specifically, for each sensing node in the current reconstructed frame, a local gradient vector within its spatial neighborhood is calculated to characterize the drastic changes in the signal around the node. A larger magnitude of the local gradient vector indicates that the node is located at a signal edge or in a noise-interference region, requiring more attention to detail preservation or enhanced suppression; a smaller magnitude indicates a smoother signal, allowing for more moderate smoothing to suppress noise. For example, taking the current sensing node... Construct a 3×3 spatial neighborhood around the center, and calculate the first-order difference in the horizontal direction (column index changes) and the vertical direction (row index changes) to obtain the local gradient vector. ,in The gradient component is in the horizontal direction. The gradient component in the vertical direction has its magnitude calculated using the following formula:

[0128]

[0129] in, The magnitude of the local gradient vector. , , For the nodes in the current reconstructed frame The numerical value; extract the confidence value corresponding to the sensing node from the dynamic confidence field. A joint diffusion control function is constructed by multiplying an edge stopping function by a confidence modulation factor. The edge stopping function adaptively adjusts the diffusion intensity based on the local gradient magnitude, while the confidence modulation factor modulates the diffusion behavior based on the node confidence level. For example, the edge stopping function can take an exponential decay form, expressed as:

[0130]

[0131] in, This is the edge stopping function, with values ​​ranging from (0,1]. The gradient threshold is set adaptively based on the dynamic range of the signal. The magnitude of the local gradient vector is used. When the gradient magnitude is large, the edge stopping function value approaches 0, suppressing diffusion to preserve edge details; when the gradient magnitude is small, the function value approaches 1, allowing stronger diffusion to suppress noise. The confidence modulation factor directly adopts the weight values ​​of the dynamic confidence field. (Value range [0,1]), the higher the node confidence, the closer the modulation factor is to 1, and the more closely the diffusion behavior conforms to the control of the edge stopping function; the lower the confidence, the smaller the modulation factor, further weakening the diffusion to avoid noise diffusion. Finally, the joint diffusion control function is as follows:

[0132]

[0133] in, This is a joint diffusion control function that comprehensively reflects the dual constraints of signal edge characteristics and node credibility on diffusion.

[0134] S63. Perform an anisotropic diffusion filter on the current reconstructed frame based on the joint diffusion control function to obtain the diffusion intermediate frame.

[0135] Specifically, this method performs an anisotropic diffusion filter on the current reconstructed frame based on a joint diffusion control function. The core of anisotropic diffusion is to diffuse along the direction of gradual signal change and stop diffusion along the signal edge direction. The joint diffusion control function provides personalized control over the diffusion intensity and direction for each node. Specifically, for each node in the current reconstructed frame... Calculate the numerical differences among its 4 or 8 neighboring nodes, and combine this with the joint diffusion control function. The diffusion update is calculated by weighting the control function values ​​of neighboring nodes. The current node value is then superimposed with the diffusion update to obtain the diffused node value. This process is repeated for all nodes to form a diffusion intermediate frame. This diffusion process preserves the edge details of the real touch signal through the edge stopping function and avoids the diffusion of noise from low-confidence nodes to high-confidence regions through the confidence modulation factor, achieving targeted noise suppression and detail preservation.

[0136] S64. In the diffusion intermediate frame, for continuous regions in the dynamic confidence field where the confidence value is lower than the dynamic baseline, search for reliable nodes with a confidence value higher than the dynamic baseline. Based on the diffusion intermediate frame values ​​and spatial coordinates of the reliable nodes, perform moving least squares local polynomial surface fitting. Use the fitted surface to perform interpolation estimation on the nodes in the low confidence region to obtain the surface-fitted repair frame.

[0137] Specifically, this method addresses the issue of noise residue or signal distortion in low-confidence continuous regions within the diffuse intermediate frames by performing surface fitting and repair. For example, this method can filter out continuous regions in the dynamic confidence field with confidence values ​​lower than the benchmark by setting a dynamic benchmark (e.g., a confidence value of 0.5, which can be adaptively adjusted according to noise intensity). These regions are often characterized by strong interference or weak signals, which are difficult to completely eliminate using diffusion filtering alone. For each such low-confidence continuous region, nearby sensing nodes with confidence values ​​higher than the dynamic benchmark are searched as reliable nodes. Reliable nodes, after diffusion filtering, exhibit high signal quality and strong reliability, providing a true signal reference for low-confidence regions. Based on the diffuse intermediate frame values ​​and spatial coordinates of the reliable nodes, a moving least squares algorithm is used for local polynomial surface fitting. This algorithm constructs a polynomial surface model that fits the local signal trend by assigning weights to the reliable node data that decay with spatial distance. The expression for the fitted model is:

[0138]

[0139] in, The output value for fitting the surface, The spatial coordinates of the nodes to be repaired in the low-confidence region. The coefficients of the fitting polynomial are obtained by solving the moving least squares algorithm. This fitted surface is used to interpolate and estimate the values ​​of each node in the low-confidence region, resulting in repaired values ​​for each node. These values ​​are then updated to the corresponding positions in the diffusion intermediate frames, forming a surface-fitted repaired frame. This step, guided by reliable nodes, not only compensates for signal distortion in the low-confidence region but also ensures that the repaired signal maintains consistency with the signal trend of the surrounding high-confidence region, avoiding signal abrupt changes.

[0140] S65. Use the surface fitting and repair frame as the updated current reconstruction frame, and repeat S62 to S64 until the overall change of the reconstruction frame between two consecutive iterations is lower than the convergence tolerance or the preset maximum number of iterations is reached. Then, determine the current reconstruction frame when the iteration stops as the high signal-to-noise ratio touch signal reconstruction frame.

[0141] Specifically, this method uses the surface-fitted repair frame as the updated current reconstructed frame and repeats the process from S62 to S64, namely, recalculating the local gradient vector and joint diffusion control function, performing a new round of anisotropic diffusion, and performing surface-fitting repair on low-confidence regions. During the iteration process, each round of processing further suppresses noise and optimizes signal details based on the previous one, until the stopping condition is met. For example, if the overall change of the reconstructed frame between two consecutive iterations is lower than the preset convergence tolerance (the overall change is determined by calculating the root mean square value of the difference between the corresponding node values ​​of the two rounds of frames, and the tolerance is set to 0.1%~1% of the signal dynamic range), it indicates that the signal has become stable; or, if the preset maximum number of iterations is reached (e.g., 5~10 times, set according to the balance between touch latency requirements and noise suppression requirements), it avoids response delay caused by excessive iteration. Furthermore, the current reconstructed frame at the time of stopping iteration is the high signal-to-noise ratio touch signal reconstruction frame. This frame not only suppresses mixed noise through anisotropic diffusion depth, but also repairs signal distortion in low confidence areas through local surface fitting. At the same time, it fully preserves the edge details and subtle change features of the real touch signal, laying a high-quality data foundation for the subsequent accurate extraction of net capacitance change and touch coordinates.

[0142] In summary, the noise suppression method for high signal-to-noise ratio touch signals provided in this application acquires the original touch data frames of the touch screen sensing array, combines them with the time-aligned display refresh status indicator signal, and uses a combination of neighborhood median statistics and distance-weighted interpolation to pre-repair isolated singular points, initially purifying the data while preserving the spatial characteristics of the signal. Subsequently, based on the pre-repaired data and the display synchronization signal, it accurately extracts the spatial morphology estimation of display noise, the temporal characteristic parameters of common-mode interference, and the estimated value of the noise floor level, comprehensively quantifying the characteristics of mixed non-stationary noise. Furthermore, it integrates three factors—spatial consistency, temporal continuity, and morphological constraints—to construct a dynamic confidence field, dynamically defining the credibility of the data from each sensing node, thus overcoming the bottleneck of traditional schemes that struggle to distinguish between background interference and weak touch signals. On this basis, through the synergistic effect of anisotropic diffusion guided by a joint diffusion control function and moving least squares local surface fitting, it achieves a balance between noise depth suppression and weak signal detail preservation, generating a high signal-to-noise ratio touch signal reconstruction frame. Finally, it extracts the net capacitance change matrix by combining the silent baseline reference value, and obtains the precise touch position through local extremum search and sub-pixel coordinate calculation. This method enables the synergistic optimization of background noise suppression and instantaneous response sensitivity to weak touch events, solving the problem of difficulty in balancing noise suppression and weak signal fidelity. It significantly improves the touch accuracy and smoothness of touchscreens under complex electromagnetic conditions, while avoiding the touch response delay caused by traditional methods such as time-domain smoothing. This provides technical support for the reliable application of high refresh rate, high resolution, and ultra-thin stacked touchscreens in complex scenarios such as mobile terminals, vehicle cockpits, and industrial control.

[0143] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0144] Based on the same inventive concept, this application also provides a noise suppression system 10 for high signal-to-noise ratio (SNR) touch signals of a touchscreen, used to implement the noise suppression method for high SNR touch signals described above. The solution provided by this system is similar to the implementation described in the above method. Therefore, the specific limitations of one or more embodiments of the noise suppression system 10 for high SNR touch signals provided below can be found in the limitations of the noise suppression method for high SNR touch signals described above, and will not be repeated here.

[0145] In one exemplary embodiment, such as Figure 3 As shown, a noise suppression system for high signal-to-noise ratio touchscreen signals is provided, comprising:

[0146] The pre-repair module 11 is used to obtain the original touch data frame obtained by the touch screen sensing array in the current scanning frame, and perform isolated singularity pre-repair on the original touch data frame based on the display refresh status indication signal aligned with the timing of the current scanning frame, so as to obtain the pre-repaired touch data frame.

[0147] The noise analysis module 12 is used to extract the spatial morphology estimation of display noise and the temporal characteristic parameters of common-mode interference of the current frame based on the pre-repaired touch data frame and the display refresh status indication signal, and to obtain noise level data, which includes display noise estimation map, common-mode interference intensity estimation value and noise floor level estimation value.

[0148] The confidence field construction module 13 is used to construct a dynamic confidence field that integrates spatial consistency factor, temporal continuity factor and morphological constraint factor based on the pre-repaired touch data frame and noise level data; wherein, the dynamic confidence field is a two-dimensional weight matrix of the same size as the touch screen sensing array, and each weight value represents the degree of confidence of the corresponding sensing node data in reflecting the real touch state.

[0149] The hybrid reconstruction module 14 is used to perform hybrid reconstruction processing on the pre-repaired touch data frame by combining anisotropic diffusion and local surface fitting according to the dynamic confidence field, so as to obtain a high signal-to-noise ratio touch signal reconstruction frame.

[0150] The touch positioning module 15 is used to extract the net capacitance change matrix caused by touch based on the reconstructed frame of the high signal-to-noise ratio touch signal and the pre-established silent baseline reference value, and to perform local extremum search and sub-pixel coordinate calculation on the net capacitance change matrix to obtain the touch position coordinate information of the current frame.

[0151] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a noise suppression method for a high signal-to-noise ratio touchscreen signal as described above.

[0152] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0153] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0154] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A noise suppression method for touch signals of a high signal-to-noise ratio touchscreen, characterized in that, The method includes: The original touch data frame obtained by the touch screen sensing array in the current scan frame is acquired. Based on the display refresh status indication signal that is aligned with the timing of the current scan frame, the original touch data frame is pre-repaired for isolated singular points based on neighborhood median statistics to obtain a pre-repaired touch data frame. Based on the pre-repaired touch data frame and the display refresh status indication signal, the display noise spatial morphology estimation and common-mode interference temporal feature parameters of the current frame are extracted to obtain noise level data, which includes display noise estimation map, common-mode interference intensity estimation value and noise floor level estimation value. Based on the pre-repaired touch data frame and the noise level data, a dynamic confidence field is constructed that integrates spatial consistency factor, temporal continuity factor and morphological constraint factor; wherein, the dynamic confidence field is a two-dimensional weight matrix of the same size as the touch screen sensing array, and each weight value represents the degree of confidence of the corresponding sensing node data in reflecting the real touch state; Based on the dynamic confidence field, the pre-repaired touch data frame is subjected to a hybrid reconstruction process combining anisotropic diffusion and local surface fitting to obtain a high signal-to-noise ratio touch signal reconstruction frame. Based on the reconstructed frame of the high signal-to-noise ratio touch signal and the pre-established silent baseline reference value, the net capacitance change matrix caused by touch is extracted, and the local extremum search and sub-pixel coordinate solution are performed on the net capacitance change matrix to obtain the touch position coordinate information of the current frame.

2. The method according to claim 1, characterized in that, The process of acquiring the original touch data frame obtained by the touchscreen sensing array in the current scan frame, and performing isolated singularity pre-repair on the original touch data frame based on neighborhood median statistics, based on a display refresh status indication signal aligned with the timing of the current scan frame, to obtain a pre-repaired touch data frame, includes: S21. Obtain the original touch data frame. For each sensing node in the original touch data frame, extract the values ​​of all nodes in the spatial neighborhood window centered on the current sensing node, and calculate the median and median absolute deviation of the values ​​in the spatial neighborhood window. S22. Calculate the absolute value of the deviation between the original value of the current sensing node and the median. When the absolute value of the deviation exceeds the dynamic elastic boundary determined based on the median of the absolute deviation, determine that the current sensing node is an anomaly. S23. For the current sensing node that is determined to be an anomaly, obtain the value of the nearest non-anomaly node in the same row or column direction as the current sensing node, perform distance-weighted interpolation based on the spatial geometric distance between the nearest non-anomaly node and the current sensing node, and generate a replacement value to fill in the position of the current sensing node. S24. Traverse all the sensing nodes in the original touch data frame until all the sensing nodes have executed steps S21 to S23, and obtain the pre-repaired touch data frame.

3. The method according to claim 1, characterized in that, The step of extracting display noise spatial morphology estimation and common-mode interference temporal characteristic parameters of the current frame based on the pre-repaired touch data frame and the display refresh status indication signal to obtain noise level data includes: According to the display row scan currently active area identified by the display refresh status indication signal, the histogram statistics of the node values ​​of each column in the row are calculated for the sensing row data corresponding to the display refresh activation status in the pre-repair touch data frame, and the representative value corresponding to the highest frequency interval of the histogram statistics is used as the candidate value of the row display noise benchmark. The candidate values ​​of the row display noise baseline for several consecutive rows within the same display refresh half-cycle are smoothly fitted to generate the display noise spatial envelope curve. The sensing rows corresponding to the display refresh blanking period are filled by extrapolation of the envelope curve of the adjacent activation period or by the statistical value of the historical blanking period to obtain the display noise estimation map. From the pre-repaired touch data frame, select the sensing row whose row display noise baseline candidate value is consistent with the mode of the row and whose data variance of the row is lower than the preset condition as the silent reference row, calculate the average value of all node values ​​in the silent reference row, and obtain the common mode interference intensity estimate. The common-mode interference intensity estimate of the current frame is combined with the common-mode interference intensity estimate of several previous historical frames to form a time series. The first difference and short-term fluctuation variance of the time series are calculated to obtain the common-mode interference time-domain activity index. The root mean square value is calculated based on the residual sequence of each node value in the silent reference row after removing the in-row trend, and the estimated value of the noise floor level is obtained.

4. The method according to claim 1, characterized in that, The step of constructing a dynamic confidence field based on the pre-repaired touch data frame and the noise level data includes: For each sensing node in the pre-repaired touch data frame, a normalized local entropy is calculated in the local spatial neighborhood centered on the sensing node, and the absolute value of the difference between the sensing node value and the corresponding position value in the display noise estimation map is calculated as the local de-display residual. The normalized local entropy and the normalized ratio of the local de-display residual to the estimated value of the background noise level are combined to obtain the spatial consistency factor. For each sensing node in the pre-repaired touch data frame, the historical data values ​​of the sensing node after pre-repair processing in several consecutive historical frames are obtained to form a time series sample. The time series sample is used to predict the theoretical expected value of the sensing node in the current frame through an autoregressive moving average model. The recent fluctuation standard deviation of the time series sample is calculated. The time continuity factor is obtained based on the degree of deviation of the absolute value of the deviation between the actual observed value and the theoretical expected value in the current frame relative to the recent fluctuation standard deviation. The pre-repaired touch data frame is reconstructed based on the spatial consistency factor and the temporal continuity factor to obtain a coarsely denoised intermediate image. Connectivity analysis is performed on the coarsely denoised intermediate image to extract candidate touch connected regions. The similarity between the geometric morphological parameters of each candidate touch connected region and the reasonable morphological parameter range defined by the finger touch physical model is calculated. The morphological likelihood function value of the candidate touch connected region to which each sensing node belongs is used as the morphological constraint factor. The spatial consistency factor, the temporal continuity factor, and the morphological constraint factor are multiplied point-by-point at each sensing node and then normalized to obtain the dynamic confidence field.

5. The method according to claim 1, characterized in that, The step of performing a hybrid reconstruction process combining anisotropic diffusion and local surface fitting on the pre-repaired touch data frame based on the dynamic confidence field to obtain a high signal-to-noise ratio touch signal reconstruction frame includes: S61. Use the pre-repaired touch data frame as the current reconstruction frame; S62. For each sensing node in the current reconstructed frame, calculate the local gradient vector in the spatial neighborhood. Based on the magnitude of the local gradient vector and the corresponding sensing node confidence value extracted from the dynamic confidence field, construct a joint diffusion control function. The joint diffusion control function is composed of an edge stopping function multiplied by a confidence modulation factor. S63. Perform an anisotropic diffusion filter on the current reconstructed frame based on the joint diffusion control function to obtain a diffusion intermediate frame; S64. In the diffusion intermediate frame, for the continuous region in the dynamic confidence field where the confidence value is lower than the dynamic reference, search for reliable nodes with a confidence value higher than the dynamic reference. Based on the diffusion intermediate frame value and spatial coordinates of the reliable nodes, perform moving least squares local polynomial surface fitting. Use the fitted surface to perform interpolation estimation on the nodes in the low confidence region to obtain the surface fitting repair frame. S65. The surface fitting and repair frame is used as the updated current reconstruction frame. S62 to S64 are repeated until the overall change of the reconstruction frame between two consecutive iterations is lower than the convergence tolerance or the preset maximum number of iterations is reached. The current reconstruction frame at the time of stopping iteration is determined as the high signal-to-noise ratio touch signal reconstruction frame.

6. A noise suppression system for high signal-to-noise ratio touchscreen signals, used to implement the method according to any one of claims 1 to 5, characterized in that, The system includes: The pre-repair module is used to acquire the original touch data frame obtained by the touch screen sensing array in the current scan frame, and perform isolated singularity pre-repair on the original touch data frame based on neighborhood median statistics based on the display refresh status indication signal aligned with the timing of the current scan frame to obtain the pre-repaired touch data frame. The noise analysis module is used to extract the display noise spatial morphology estimation and common-mode interference temporal feature parameters of the current frame based on the pre-repaired touch data frame and the display refresh status indication signal, and obtain noise level data, which includes display noise estimation map, common-mode interference intensity estimation value and noise floor level estimation value. The confidence field construction module is used to construct a dynamic confidence field that integrates spatial consistency factor, temporal continuity factor and morphological constraint factor based on the pre-repaired touch data frame and the noise level data; wherein, the dynamic confidence field is a two-dimensional weight matrix of the same size as the touch screen sensing array, and each weight value represents the degree of confidence of the corresponding sensing node data in reflecting the real touch state; The hybrid reconstruction module is used to perform hybrid reconstruction processing on the pre-repaired touch data frame based on the dynamic confidence field, combining anisotropic diffusion and local surface fitting, to obtain a high signal-to-noise ratio touch signal reconstruction frame. The touch positioning module is used to extract the net capacitance change matrix caused by touch based on the reconstructed frame of the high signal-to-noise ratio touch signal and the pre-established silent baseline reference value, and to perform local extremum search and sub-pixel coordinate calculation on the net capacitance change matrix to obtain the touch position coordinate information of the current frame.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

Images