Touch data processing method based on organic display

Abstract

The invention relates to the technical field of organic display data processing, in particular to a touch data processing method based on an organic display, which comprises the following steps of: monitoring a flexible deformation acceleration component and an environmental parameter change rate in real time through a dynamic trigger function, and triggering high-priority baseline calibration when a threshold value exceeds a limit; and generating a self-adaptive compensation amount matched with the current touch signal noise spectrum. And constructing a dielectric constant offset prediction model by using a long short-term memory neural network or a Kalman filter, deducing the dynamic change trend of the dielectric property of the material, and correcting a touch signal compensation coefficient in a grading manner according to the predicted offset. And a closed-loop error correction mechanism is formed based on the spatial distribution entropy reverse optimization model weight parameter and the trigger threshold boundary of the contact coordinate residual matrix. According to the method, the problem of time domain mismatch of fixed period calibration and organic material nonlinear offset is solved, the accumulation of touch coordinate analysis errors in a high-curvature deformation scene is effectively inhibited, and the touch positioning precision and the interaction response real-time performance are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of organic display data processing, and particularly to a touch data processing method based on an organic display. Background Art

[0002] The touch data processing of an organic display is based on a capacitive or resistive sensing principle, and the charge change or piezoresistive signal at the touch position is captured in real time through a transparent electrode array integrated on the panel surface. After the original touch signal is sampled and analog-to-digital converted by a front-end analog circuit, a dedicated touch chip performs digital filtering to eliminate environmental electromagnetic interference, and an interpolation algorithm is used to improve the coordinate resolution accuracy to the sub-pixel level. In view of the possible dielectric constant fluctuations or flexible deformation characteristics of organic display materials, the system introduces a dynamic baseline calibration technology, and compensates for signal drift caused by temperature, humidity or mechanical stress by periodically updating the reference potential threshold in the non-touch state. The processed touch coordinate data is transmitted to the main control unit through an SPI or I2C interface, and the spatio-temporal consistency of the touch trajectory and visual rendering is achieved by combining the display timing synchronization mechanism. Finally, the distribution and response of multi-touch events are completed through a human-computer interaction framework.

[0003] In the touch data processing of an organic display, the dynamic baseline calibration technology updates the reference potential threshold at a fixed period, and it is difficult to adapt to the non-linear offset of the dielectric constant caused by instantaneous mechanical deformation or rapid jump of environmental parameters of organic materials in real time, resulting in signal compensation lagging behind the actual physical state of the touch interface, and causing the touch coordinate parsing error to increase exponentially with the increase of the deformation rate. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a touch data processing method based on an organic display, which solves the time-domain mismatch problem between the fixed-period update mechanism of dynamic baseline calibration and the non-linear offset of the dielectric constant of organic materials, so as to suppress the exponential growth of touch coordinate parsing errors in a flexible deformation scenario.

[0005] To solve the above technical problems, the specific technical solution of the present invention is as follows: The touch data processing method based on an organic display provided by the present invention includes: Receiving the original touch signal collected by the touch electrode array, synchronously obtaining the deformation gradient parameter output by the flexible deformation sensor and the temperature and humidity parameters collected by the environmental sensor, and after performing analog-to-digital conversion on the touch signal, jointly inputting the deformation gradient parameter and temperature and humidity parameter aligned with the time stamp into a multi-source data fusion module; In the multi-source data fusion module, a dynamic trigger function is constructed based on the acceleration component of the deformation gradient parameter and the change rate of the temperature and humidity parameters. When the acceleration component or the change rate of the temperature and humidity exceeds the corresponding preset threshold, a high-priority baseline calibration process is triggered, and an adaptive compensation amount for the reference potential threshold is generated according to the current touch signal characteristics and historical baseline data; The touch signal after baseline calibration, the deformation gradient parameter, and the temperature and humidity parameters are input into the dielectric constant offset prediction model. The trend of the dielectric constant offset of the organic material is deduced through a time series prediction algorithm, and the touch signal compensation coefficient is dynamically corrected according to the predicted offset amount; Based on the touch signal after correction and compensation, an interpolation algorithm is used to calculate the sub-pixel-level touch coordinates, and a contact coordinate residual matrix is generated. When the spatial distribution statistical value of the residual matrix exceeds the preset tolerance range, the weight parameters of the dielectric constant offset prediction model are reversely optimized, and the threshold boundary of the dynamic trigger function is synchronously adjusted; The optimized touch coordinate data is frame-synchronized and aligned with the display driving timing. Combining the touch pressure sensor data and the contact movement rate parameters, a touch event priority queue is generated through the event distribution engine and the response result is output.

[0006] Further, in the touch data processing method based on an organic display of the present invention, after receiving the original touch signal collected by the touch electrode array, it includes: Perform adaptive notch filtering on the touch signal to eliminate the periodic interference generated by the display driving circuit, perform Butterworth low-pass filtering on the deformation gradient parameter output by the flexible deformation sensor to strip high-frequency vibration noise, and perform exponentially weighted moving average processing on the temperature and humidity parameters collected by the environmental sensor to smooth transient fluctuations; Extract the effective amplitude envelope of the filtered touch signal, the second derivative feature of the deformation gradient parameter, and the change trend slope of the temperature and humidity parameters, and input the feature vector into the feature input port of the dynamic trigger function described in claim 1.

[0007] Further, in the touch data processing method based on an organic display of the present invention, the trigger baseline calibration includes: When generating a compensation amount according to the noise spectral density of the current touch signal and the reference value stored in the historical baseline database, if it is detected that the output value of the deformation acceleration sensor exceeds the material deformation critical threshold, a compensation strategy dominated by the deformation gradient parameter is preferentially adopted, and the compensation amount weight is assigned to the deformation compensation sub-module; When the change rate collected by the temperature and humidity sensor crosses the environmental gradient threshold, the environmental compensation sub-model is activated to execute in parallel with the baseline calibration main process, and the output of the environmental compensation sub-model is weighted and fused with the compensation amount of the deformation compensation sub-module.

[0008] Further, in the touch data processing method based on an organic display according to the present invention, the processing of the dielectric constant offset prediction model includes: Dynamically allocate the input weights of the prediction model according to the real-time deformation gradient parameter and the temperature and humidity parameter. When the output value of the deformation acceleration sensor exceeds the set threshold, increase the weight of the deformation factor in the input layer of the model to more than 0.8. After completing the parsing of the effective contact coordinates for a preset number of times, update the weights of the hidden layer of the LSTM network of the time series prediction algorithm or the process noise covariance matrix of the Kalman filter using the statistical distribution data of the contact coordinate residual matrix.

[0009] Further, in the touch data processing method based on an organic display according to the present invention, the dynamic correction of the touch signal compensation coefficient includes: Set the compensation boundary according to the material deformation interval where the dielectric constant prediction offset is located, and the deformation interval is quantified and graded by the curvature change rate collected by the flexible deformation sensor. When the predicted offset enters the non-linear deformation region, enable the saturation compensation mode to limit the maximum compensation amplitude, and the non-linear deformation region corresponds to the state where the deformation gradient parameter exceeds the preset curvature threshold. The compensation boundary value is generated based on the statistical learning results stored in the historical baseline database and is periodically iteratively updated with the optimization results of the contact coordinate residual matrix.

[0010] Further, in the touch data processing method based on an organic display according to the present invention, after generating the contact coordinate residual matrix, it includes: Analyze the spatial distribution entropy value of the contact coordinate residual matrix using the sliding window statistical method. When the entropy value exceeds the set model optimization threshold, trigger the parameter retraining process of the dielectric constant offset prediction model. Limit the adjustment step size of the weights of the hidden layer of the LSTM network or the parameters of the Kalman filter in a single optimization iteration to prevent overfitting of the model.

[0011] Further, in the touch data processing method based on an organic display according to the present invention, the generation of the touch event queue includes: Calculate the touch event confidence score based on the statistical distribution of the contact coordinate residual matrix, the consistency check result of the pressure sensor data, and the curvature change rate of the sliding trajectory. When the confidence scores of consecutive touch events are lower than the set environmental gradient threshold, freeze the current event queue and start the fast recalibration process, re-parse the buffered data after preprocessing and filter out low-confidence contacts.

[0012] Further, in the touch data processing method based on an organic display according to the present invention, the fast recalibration process includes: Synchronously update the compensation boundary parameters of the dielectric constant offset prediction model within the preset time of the baseline calibration process; During the recalibration process, pause the touch event distribution thread, and resume the optimized touch coordinate data processing after the calibration is completed.

[0013] Furthermore, for the touch data processing method based on an organic display according to the present invention, the frame synchronization alignment process includes: Adjust the timing phase of the touch coordinate output thread according to the vertical synchronization signal of the display driving timing; Insert dynamic delay compensation between the touch trajectory and the display rendering, and the delay compensation amount is calculated based on the deformation gradient data collected by the strain sensor and the current display frame rate.

[0014] Furthermore, the touch data processing method based on an organic display according to the present invention further includes: Set distributed microelectromechanical system strain sensors in the flexible substrate, and the sensors share the timing signals of the display driving circuit with the touch electrode array; The X / Y / Z axis deformation gradient data collected by the strain sensor is transmitted through a time-division multiplexing mechanism to reduce the crosstalk amplitude with the signals collected by the touch electrode array.

[0015] Advantages of the present invention; The present invention uses a dynamic trigger function to monitor the deformation acceleration and the jump amplitude of environmental parameters in real time, preferentially triggers baseline calibration during flexible deformation or environmental mutation, and combines the timing deduction ability of the dielectric constant offset prediction model to achieve real-time matching of the touch signal compensation coefficient and the drift of material properties. Using the spatial distribution characteristics of the contact coordinate residual matrix to reversely optimize the weights and trigger threshold boundaries of the prediction model, a closed-loop control link from signal acquisition, error prediction to parameter self-adaptation is formed, effectively suppressing the cumulative effect of touch coordinate errors caused by the non-linear deformation of organic materials. The time-division multiplexing mechanism and the dynamic delay compensation strategy cooperate to reduce the transmission crosstalk between the touch signal and the deformation sensing data, and eliminate the visual interaction delay caused by the deformation of the flexible screen through frame synchronization alignment and spatio-temporal coordinate mapping, overall improving the touch positioning accuracy and response real-time performance in high-curvature deformation scenarios. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, other drawings can be obtained according to the drawings without creative efforts.

[0017] Figure 1 It is a flowchart of the touch data processing method based on an organic display provided by an embodiment of the present invention. Detailed implementation manners

[0018] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without any creative efforts shall fall within the protection scope of the present invention. The technical solutions provided by the embodiments of the present invention will be described in detail below with reference to the drawings. To better understand the objectives of the present invention, the present invention will be further described in detail below.

[0019] Please refer to Figure 1 , the touch data processing method based on an organic display provided by the present invention includes: Step S101: Receive the original touch signal collected by the touch electrode array, synchronously obtain the deformation gradient parameter output by the flexible deformation sensor and the temperature and humidity parameters collected by the environmental sensor. After performing analog-to-digital conversion on the touch signal, the deformation gradient parameter and temperature and humidity parameter aligned with the time stamp are jointly input into the multi-source data fusion module; Step S102: In the multi-source data fusion module, construct a dynamic trigger function based on the acceleration component of the deformation gradient parameter and the change rate of the temperature and humidity parameter. When the acceleration component or the change rate of the temperature and humidity exceeds the corresponding preset threshold, trigger a high-priority baseline calibration process, and generate an adaptive compensation amount for the reference potential threshold according to the current touch signal characteristics and historical baseline data; Step S103: Input the touch signal, deformation gradient parameter, and temperature and humidity parameter after baseline calibration into the dielectric constant offset prediction model, deduce the dielectric constant offset trend of the organic material through the time series prediction algorithm, and dynamically correct the touch signal compensation coefficient according to the predicted offset amount; Step S104: Based on the touch signal after correction and compensation, use the interpolation algorithm to calculate the sub-pixel-level touch coordinates, generate a contact coordinate residual matrix. When the spatial distribution statistical value of the residual matrix exceeds the preset tolerance range, reversely optimize the weight parameters of the dielectric constant offset prediction model, and synchronously adjust the threshold boundary of the dynamic trigger function; Step S105: Perform frame synchronization alignment processing on the optimized touch coordinate data and the display driving timing, combine the touch pressure sensor data and the contact movement speed parameter, and generate a touch event priority queue and output a response result through the event distribution engine.

[0020] In the touch data processing method based on organic display, after the touch electrode array collects the original touch signal, the micro-electromechanical system strain sensor and environmental sensor embedded in the flexible substrate are synchronously started to obtain the three-dimensional deformation gradient parameters and environmental temperature and humidity parameters respectively. The touch signal is amplified and converted to analog by the analog front-end circuit, the deformation gradient parameter is eliminated by the Butterworth low-pass filter to eliminate high-frequency mechanical vibration noise, and the temperature and humidity parameters are smoothed by the exponential weighted sliding average algorithm to smooth transient fluctuations. After the multi-source data is aligned with the timestamp by the hardware clock synchronization module, it is input into the multi-source data fusion module for time and space registration to form a heterogeneous sensor data stream with a unified time reference.

[0021] In the multi-source data fusion module, the first-order derivative of the deformation gradient parameter is analyzed in real time based on the deformation acceleration component calculation unit, and the input vector of the dynamic trigger function is constructed in combination with the environmental parameter gradient value output by the temperature and humidity change rate detection unit. When the deformation acceleration exceeds the critical threshold of the material elastic deformation or the temperature and humidity gradient value exceeds the environmental stability boundary, the baseline calibration process of the interrupt priority is triggered. This calibration process calls the reference potential characteristic spectrum stored in the historical baseline database, generates a compensation amount that matches the current touch signal noise spectrum through an adaptive filtering algorithm, and updates the touch signal recognition threshold.

[0022] The touch signal after baseline calibration is input into the dielectric constant offset prediction model together with deformation and environmental parameters. The model adopts a long short-term memory neural network architecture. After receiving the continuous sensor data stream, it deduces the dynamic offset trend of the organic material dielectric constant through the hidden layer state update mechanism. The output of the prediction model is connected to the signal compensation coefficient calculation unit, which dynamically adjusts the amplitude compensation coefficient and phase correction parameter of the touch signal according to the relative position of the dielectric constant offset in the material characteristic curve.

[0023] The compensated touch signal enters the coordinate analysis stage, and the sub-pixel touch coordinate data is calculated by using a bilinear interpolation algorithm combined with the topological structure of the touch electrode array. The synchronously generated touch coordinate residual matrix contains the spatial deviation distribution of the actual touch point and the theoretical predicted position. When the statistical variance of the residual matrix exceeds the preset tolerance threshold, the model parameter optimization thread is triggered. This optimization thread updates the convolution kernel weights of the dielectric constant offset prediction model through the back-propagation algorithm, and jointly adjusts the deformation acceleration and environmental gradient threshold parameters in the dynamic trigger function to form a closed-loop error correction mechanism.

[0024] The optimized touch coordinate data input and display synchronization control module adjusts the touch coordinate output timing phase according to the vertical synchronization signal of the display panel to eliminate the timing deviation between touch response and frame rendering. The contact pressure data collected by the touch pressure sensor and the contact movement rate parameters detected by the photoelectric sensor are jointly input into the event priority evaluation unit to generate a multi-dimensional event feature vector including the contact trajectory, pressure level, and interaction timeliness. The event distribution engine constructs a touch event queue with execution priorities based on the comprehensive scoring results of the feature vector and outputs a touch response data stream with low latency and high precision through the human-computer interaction interface.

[0025] Each processing stage realizes information interaction through the data bus and status register. The real-time monitoring data of the deformation sensor continuously updates the input parameters of the multi-source fusion module. The change of environmental parameters triggers the threshold adjustment of the calibration process. The residual optimization result is real-time fed back to the model prediction and signal compensation unit, and the display synchronization control signal shares the timing reference with the event distribution engine. The closed-loop control mechanism of this data processing link enables the touch signal processing system to adapt to the changes of organic material characteristics and environmental condition fluctuations and maintain stable touch recognition accuracy.

[0026] Specifically, for the touch data processing method based on an organic display according to the present invention, after receiving the original touch signal collected by the touch electrode array, it includes: Performing adaptive notch filtering on the touch signal to eliminate the periodic interference generated by the display driving circuit, performing Butterworth low-pass filtering on the deformation gradient parameters output by the flexible deformation sensor to strip high-frequency vibration noise, and performing exponentially weighted moving average processing on the temperature and humidity parameters collected by the environmental sensor to smooth transient fluctuations; Extracting the effective amplitude envelope of the filtered touch signal, the second derivative feature of the deformation gradient parameter, and the change trend slope of the temperature and humidity parameters, and inputting the feature vector into the feature input port of the dynamic trigger function described in claim 1.

[0027] After the original signal collected by the touch electrode array is input into the signal preprocessing module, the display driving interference is first processed by the adaptive notch filter bank. This filter bank real-time monitors the fundamental frequency and harmonic components of the display driving circuit and dynamically adjusts the notch center frequency and bandwidth parameters according to the detected interference frequency, effectively suppressing the periodic noise overlapping with the touch signal frequency band. The processed touch signal retains the high-frequency feature components of valid touch events and eliminates the fixed pattern noise introduced by the display refresh operation.

[0028] The deformation gradient parameter output by the flexible deformation sensor enters the mechanical noise suppression unit, and a fourth-order Butterworth low-pass filter is used for signal conditioning. The cut-off frequency of the filter is dynamically configured according to the material resonance frequency characteristics of the organic substrate to filter out the transient noise components generated by external mechanical shocks or high-frequency vibrations. The filtered deformation gradient parameter retains the low-frequency characteristic signal reflecting the bending deformation of the flexible display, providing a stable data input for subsequent deformation state analysis.

[0029] The temperature and humidity parameters collected by the environmental sensor are input into the data smoothing processing unit, and an exponentially weighted moving average algorithm with a variable time window length is used. The algorithm automatically adjusts the time window length according to the fluctuation amplitude of the environmental parameters, extends the window to improve the smoothing effect during the stable stage of temperature and humidity, and shortens the window to retain the change trend during the environmental mutation stage. The processed environmental parameters eliminate the influence of instantaneous sampling outliers and accurately reflect the gradual change process of the environmental conditions.

[0030] The preprocessed multi-source data is input into the feature extraction unit. The effective amplitude envelope of the touch signal is extracted through an envelope detection circuit to retain the energy distribution characteristics of the touch event. The second derivative feature of the deformation gradient parameter is obtained by calculation with a digital differentiator, which characterizes the change trend of the deformation acceleration. The change trend slope of the temperature and humidity parameters is obtained by fitting the linear component with the least squares method to quantify the dynamic change rate of the environmental conditions. After the above feature parameters are normalized, they are aligned according to the time stamp to form a multi-dimensional feature vector.

[0031] The feature vector is transmitted to the dynamic trigger function processing unit through a dedicated data bus, and a feature dimension verification mechanism is set at the input port. This verification mechanism verifies that the dimension of the feature vector matches the number of nodes in the input layer of the dynamic trigger function to avoid abnormal function calculation caused by incorrect data formats. The feature vector that passes the verification enters the feature buffer of the dynamic trigger function and waits for the function call cycle to trigger the calculation process.

[0032] Specifically, for the touch data processing method based on an organic display according to the present invention, the trigger baseline calibration includes: When generating a compensation amount based on the noise spectral density of the current touch signal and the reference value stored in the historical baseline database, if it is detected that the output value of the deformation acceleration sensor exceeds the material deformation critical threshold, a compensation strategy dominated by the deformation gradient parameter is preferentially adopted, and the compensation amount weight is allocated to the deformation compensation sub-module; When the change rate collected by the temperature and humidity sensor crosses the environmental gradient threshold, the environmental compensation sub-model is activated to execute in parallel with the baseline calibration main process, and the output of the environmental compensation sub-model is weighted and fused with the compensation amount of the deformation compensation sub-module.

[0033] After the touch signal enters the baseline calibration module, the noise analysis unit calculates the noise spectral density distribution characteristics of the current touch signal in real time, and obtains the total noise energy value in the medium and high frequency bands through the spectral energy integration algorithm. This noise energy value is compared with the reference noise spectrum stored in the historical baseline database to generate an initial compensation amount based on the frequency-domain energy difference. During the generation of the compensation amount, the deformation state monitoring unit synchronously receives the acceleration component data of the flexible deformation sensor. When it detects that the acceleration value exceeds the deformation critical threshold corresponding to the material yield strength, it triggers the compensation strategy switching mechanism.

[0034] After the compensation strategy switching mechanism is started, the deformation compensation sub-module takes over the main control right and raises the compensation amount weight distribution coefficient to the preset upper limit value. This sub-module calls the historical change curve of the deformation gradient parameter and finds the optimal compensation parameter combination in the similar deformation mode through the sliding window matching algorithm. The weight distribution process adopts dynamic priority scheduling. The compensation amount output by the deformation compensation sub-module covers more than 60% of the noise compensation base amount, and the remaining weight is reserved for the basic noise suppression module.

[0035] The environmental monitoring thread continuously detects the gradient change data of the temperature and humidity sensor. When the change rate breaks through the gradient threshold set in the environmental stability model, the activation signal of the environmental compensation sub-model triggers the interrupt service program. After this sub-model is started, it forms a parallel execution link with the baseline calibration main process, obtains the current environmental parameters and historical environmental feature data through an independent data buffer, and calculates the environmental compensation component based on the humidity penetration effect and the temperature expansion coefficient.

[0036] The compensation amount fusion unit receives the output data of the deformation compensation sub-module and the environmental compensation sub-model, and uses the variable weight fusion algorithm to integrate the data. When the deformation compensation amount occupies the dominant weight, the environmental compensation component is superimposed as an offset correction term; when the environmental change rate continuously exceeds the threshold, the fusion algorithm automatically adjusts the weight distribution ratio so that the environmental compensation amount obtains a weight ratio of not less than 30%. The fused comprehensive compensation amount is input into the reference potential adjustment circuit to update the touch signal recognition threshold voltage.

[0037] During the parallel processing process, the deformation compensation sub-module and the environmental compensation sub-model exchange real-time status data through the shared memory area. The excessive state of the deformation acceleration will trigger the warning mechanism of the environmental compensation sub-model to pre-load the environmental parameter compensation template that may be required; conversely, when the environmental parameters mutate, the deformation compensation sub-module automatically enters the high-sensitivity monitoring mode. The cooperative working mechanism of the dual modules realizes the real-time synchronization of the running state through the interrupt signal and the status register, maintaining the dynamic balance of the compensation system.

[0038] Specifically, for the touch data processing method based on the organic display described in the present invention, the processing of the dielectric constant offset prediction model includes: Dynamically allocate the input weights of the prediction model according to the real-time deformation gradient parameters and temperature and humidity parameters. When the output value of the deformation acceleration sensor exceeds the set threshold, increase the weight of the deformation factor in the input layer of the model to more than 0.8; After completing the parsing of the effective contact coordinates for the preset number of times, update the weights of the hidden layer of the LSTM network of the time series prediction algorithm or the process noise covariance matrix of the Kalman filter using the statistical distribution data of the contact coordinate residual matrix.

[0039] After receiving the preprocessed deformation gradient parameters and environmental temperature and humidity parameters, the dielectric constant offset prediction model has a dynamic weight allocation unit that parses the sensor data features in real time. The deformation acceleration component is calculated by a digital differential circuit. When it is detected that this component exceeds the elastic deformation threshold in the material fatigue characteristic database, the weight controller increases the connection weight of the deformation factor in the input layer of the model to the preset upper limit value. During the weight adjustment process, the environmental parameter channel maintains a basic weight ratio of not less than 15% to maintain the monitoring ability of environmental factors on the change of the dielectric constant.

[0040] The data preprocessing module in the input layer of the model normalizes the deformed parameters with increased weights to eliminate the influence of dimensional differences on the prediction accuracy. The normalized data stream is input into the time series processing unit of the long short-term memory neural network, which captures the time-varying characteristics of the dielectric constant through the memory cell state update mechanism. When it is detected during the system operation that the statistical variance of the contact coordinate residual matrix continuously exceeds the standard, the residual analysis module generates a model update trigger signal.

[0041] After the model parameter update thread is started, the spatial distribution data of the residual matrix is calculated by the statistical analysis module to obtain the covariance matrix eigenvalue, and the dominant error component is extracted. This component is input into the LSTM network weight adjustment algorithm, and the connection weights of the hidden layer neurons are updated by the gradient descent method, and the update step size is limited by the preset maximum adjustment amplitude value. When the system is configured in the Kalman filter mode, the process noise covariance matrix is iteratively corrected according to the autocorrelation characteristics of the residual data, and the correction amount controls the convergence speed through an exponential decay function.

[0042] The dielectric constant offset prediction model with updated weights enters the verification stage, and the latest touch signal compensation result is input into the offline verification data set for accuracy comparison. The model parameters that pass the verification are written into the non-volatile memory through the bus to form a new set of reference parameters; the parameter versions that do not pass the verification trigger the rollback mechanism, and the model parameters of the previous stable version are loaded to continue running. This process maintains the continuous learning ability of the prediction model in the scenario of material property drift, and at the same time prevents model degradation caused by abnormal data.

[0043] The data intercommunication between the model input weight allocation mechanism and the parameter update process is realized through the status register. The over-standard status of the deformation acceleration will temporarily freeze the function of adjusting the environmental parameter weights, and the model convergence speed in the deformation-dominated mode is preferentially ensured. The statistical characteristic data of the residual matrix is synchronously fed back to the weight allocation unit. When it is detected that the proportion of the prediction error caused by the environmental parameters increases, the function of adjusting the weights of the environmental channels is automatically restored. This closed-loop control mechanism realizes the dynamic balance of the influence of deformation factors and environmental factors on the dielectric constant, and improves the model adaptability under different working conditions.

[0044] Specifically, for the touch data processing method based on the organic display according to the present invention, the dynamic correction of the touch signal compensation coefficient includes: Setting a compensation boundary according to the material deformation interval where the dielectric constant prediction offset is located, and the deformation interval is quantitatively graded by the curvature change rate collected by the flexible deformation sensor; When the predicted offset enters the non-linear deformation area, the saturation compensation mode is enabled to limit the maximum compensation amplitude, and the non-linear deformation area corresponds to the state where the deformation gradient parameter exceeds the preset curvature threshold; The compensation boundary value is generated based on the statistical learning results stored in the historical baseline database, and is periodically iteratively updated with the optimization results of the contact coordinate residual matrix.

[0045] After the touch signal compensation coefficient adjustment module receives the output data of the dielectric constant offset prediction model, the deformation state classification unit first analyzes the real-time curvature change rate parameter collected by the flexible deformation sensor. After being quantified by the analog-to-digital conversion circuit, this parameter is pattern-matched with the classification standard in the material deformation characteristic database, and the current deformation state is divided into a linear deformation area, a transition deformation area or a non-linear deformation area. The compensation boundary values corresponding to each deformation area are retrieved from the statistical learning records of the historical baseline database and loaded into the register bank of the compensation coefficient calculation unit.

[0046] When it is detected that the predicted offset enters the non-linear deformation area, the boundary control module starts the saturation compensation mechanism. This mechanism activates the piecewise linear approximation algorithm by comparing the relationship between the current curvature change rate and the preset threshold. According to the slope change characteristics of the non-linear interval, the algorithm applies an amplitude limit condition in the compensation coefficient calculation process, and constrains the maximum compensation amount within the safe range of the material stress-strain curve. The limit parameters are stored in the protected sector of the non-volatile memory to prevent data loss caused by abnormal power-off.

[0047] The compensation boundary update thread periodically reads the optimization results of the contact coordinate residual matrix, and extracts the statistical characteristics of the spatial error through the residual distribution feature analysis module. After the characteristic parameters are input into the statistical learning engine, they are matched with the reference data in the historical deformation pattern library to generate a new compensation boundary recommended value. After the recommended value is verified for accuracy through the offline verification data set, it is updated to the dynamic compensation boundary lookup table to complete the iterative optimization of the boundary parameters.

[0048] During the boundary parameter update process, the deformation state classification unit continuously monitors the gradient trend of the curvature change rate. When it is detected that the deformation state returns from the non-linear region to the linear region, a compensation mode switching signal is triggered. This signal controls the compensation coefficient calculation unit to gradually release the amplitude limit and restore the weight distribution ratio of the conventional compensation algorithm. The exponential decay transition strategy is adopted during the mode switching process to avoid the jump of the touch coordinate caused by the step change of the compensation amount.

[0049] A two-way communication is established between the historical baseline database and the real-time processing module through a dedicated data bus, and the deformation state classification result is fed back to the statistical learning engine in real time. The residual optimization data synchronously updates the grading standard of the material deformation characteristic database, forming a co-evolution mechanism of the compensation boundary parameters and the material characteristic data. This design enables the compensation system to adapt to the material aging process of the organic display and maintain the stability of the touch signal processing accuracy over time.

[0050] Specifically, for the touch data processing method based on an organic display described in the present invention, after generating the contact coordinate residual matrix, it includes: Analyze the spatial distribution entropy value of the contact coordinate residual matrix by using the sliding window statistical method. When the entropy value exceeds the set model optimization threshold, trigger the parameter retraining process of the dielectric constant offset prediction model; Limit the adjustment step size of the weights of the hidden layer of the LSTM network or the parameters of the Kalman filter in a single optimization iteration to prevent model overfitting.

[0051] After the contact coordinate residual matrix is generated, the spatial distribution analysis module starts the sliding window scanning mechanism. The window divides the two-dimensional grid units based on the topological structure of the touch electrode array, and statistically analyzes the distribution density of the residual points in each unit. The entropy value calculation unit quantifies the spatial disorder degree of the residual matrix by using the Shannon entropy formula according to the probability characteristics of the density distribution. When the calculated entropy value exceeds the model optimization threshold, the signal generation module sends an interrupt request to the parameter update subsystem.

[0052] After the parameter retraining process is started, the training data preparation module extracts the touch signal sequence, deformation gradient parameters, and environmental data within a preset time window from the circular buffer. After the data set is normalized, it is input into the LSTM network weight update unit or the Kalman filter parameter adjustment unit. The network weight update uses a gradient descent algorithm with constraints, setting an absolute upper limit on the amount of weight adjustment per time to suppress the impact of high-frequency parameter fluctuations on the model stability.

[0053] During the weight update of the hidden layer of the LSTM network, the gradient clipping module monitors the magnitude of the gradient calculated by backpropagation in real time. When it detects that the gradient value exceeds the preset safety threshold, it starts a dynamic scaling mechanism to compress the gradient vector proportionally within the allowable range. The scaled gradient is input into the weight update formula to maintain the evolution of network parameters within a reasonable range. When adjusting the parameters of the Kalman filter, the update step size of the process noise covariance matrix is limited by an exponential decay function to prevent drastic changes in the covariance value during a single iteration.

[0054] After the model parameter update is completed, the verification subsystem loads the verification data set for forward inference testing. The inference result is compared with the actual measurement value of the touch coordinates to generate an accuracy evaluation report. When the evaluation index reaches the acceptance standard, the updated parameter group is written into the model parameter area of the non-volatile memory. If the acceptance fails, the parameter rollback mechanism automatically loads the model parameters of the previous stable version to maintain the continuous operation ability of the system.

[0055] The parameter adjustment step size limit mechanism and the model verification process form a double safety protection. The statistical result of the sliding window is fed back to the step size control unit in real time. When it detects that the entropy value of multiple consecutive training cycles remains at a high level, the maximum allowable step size value is dynamically reduced. This protection mechanism is synchronized with the model training thread through the status register, effectively balancing the parameter update speed and system stability while maintaining the model's adaptability.

[0056] Specifically, for the touch data processing method based on an organic display according to the present invention, the generation of the touch event queue includes: Calculating a touch event confidence score based on the statistical distribution of the contact coordinate residual matrix, the consistency check result of the pressure sensor data, and the change rate of the sliding trajectory curvature; When the confidence scores of consecutive touch events are lower than the set environmental gradient threshold, freeze the current event queue and start a fast recalibration process to re-parse the preprocessed buffered data and filter out low-confidence contacts.

[0057] After the statistical distribution data of the residual matrix of the touch point coordinates is input into the confidence assessment module, the spatial distribution analysis unit calculates the concentration index of the residual points in the touch plane. This index is time-series aligned with the contact area change rate data collected by the pressure sensor, and the physical correlation between the contact pressure distribution and the coordinate position is verified by the consistency verification algorithm. The sliding trajectory processing unit synchronously analyzes the curvature change rate of the contact point movement trajectory, and uses the differential method to calculate the curvature radius change gradient between adjacent touch points to form motion characteristic parameters that characterize the continuity of the touch operation.

[0058] After receiving the above three types of input data, the confidence score calculation engine uses a weighted fusion algorithm to generate a comprehensive score value. The residual aggregation index accounts for 50% of the weight, the pressure consistency verification result is assigned a 30% weight, and the trajectory curvature change rate accounts for 20%. The score value is mapped to the 0-1 interval through normalization processing and updated to the touch event attribute register in real time. When the score values ​​of three consecutive touch events are detected to be lower than the dynamic threshold set by the environmental adaptation module, the event management unit triggers the abnormal status flag.

[0059] After the abnormal status flag is activated, the event queue control module immediately freezes the current pending event queue and saves the queue pointer to the recovery area of ​​the non-volatile memory. The fast recalibration process start signal is synchronously sent to the preprocessing module to call the original touch signal, deformation gradient and environmental parameter data set stored in the ring buffer. The data re-analysis process enables an enhanced filtering algorithm. On the basis of maintaining the original preprocessing process, an adaptive filtering coefficient based on the confidence score is added to suppress the interference of low-scoring touch points on signal analysis.

[0060] The low-confidence touch point filtering unit adopts a two-level screening mechanism. The first level removes touch point data below the first quantile according to the confidence score histogram distribution, and the second level removes isolated noise points through spatial clustering analysis. The filtered touch point data set is input into the fast calibration thread, which calls the average of the last three valid calibration parameters in the historical baseline database as the initial value and executes the accelerated convergence algorithm to update the touch signal compensation coefficient.

[0061] After the calibration parameters are updated, the event queue control module unfreezes the state, loads the queue pointer from the recovery area, and reinitializes the event distribution engine. The event processing flow after restart uses the updated compensation coefficient to process subsequent touch data, and enables temporary monitoring mode, lowering the confidence score threshold by 20% within the preset time to increase the sensitivity of abnormal detection until the system returns to a stable state. The freezing and recovery process of the event queue is atomically operated through the status register and interrupt service routine to maintain the timing continuity of touch event processing.

[0062] Specifically, the touch data processing method based on an organic display of the present invention, the rapid recalibration process includes: During the preset time of the baseline calibration process, synchronously update the compensation boundary parameters of the dielectric constant offset prediction model; During the recalibration process, pause the touch event distribution thread, and resume the optimized touch coordinate data processing after calibration is completed.

[0063] After the fast recalibration process is started, the baseline calibration main thread and the model parameter update thread establish a time synchronization mechanism. The calibration clock management unit allocates processing time periods within the preset time window. The compensation boundary parameter update sub-thread of the dielectric constant offset prediction model obtains real-time sampling values of the current deformation gradient parameters and environmental parameters, and calls the standard deviation data of the last five valid calibration records in the historical baseline database to dynamically calculate the upper and lower limit adjustment amounts of the compensation boundary. After being verified by the data verification module, this adjustment amount is written into the compensation boundary parameter register group to complete the online update of the model parameters.

[0064] After the touch event distribution thread receives the pause instruction, the event queue control module immediately saves the current distribution pointer position to the breakpoint protection area of the non-volatile memory. The event buffer enters the read-only mode, and the newly arrived touch event data is temporarily stored in the high-priority cache queue to maintain the continuity of data acquisition. The interrupt service program updates the thread freeze flag bit of the status register to notify each processing unit to enter the calibration mode.

[0065] During the update process of the compensation boundary parameters, the parameter verification unit continuously monitors the data fluctuation ranges of the deformation sensor and the environmental sensor. When it detects that the parameter adjustment amount exceeds the allowable threshold of the material property database, it triggers the safety protection mechanism to abort the current calibration process and roll back to the previous valid parameter configuration. The newly verified parameters are synchronized to the coefficient memory of the signal compensation module through the high-speed data bus to overwrite the original compensation boundary values.

[0066] After the calibration completion signal is triggered, the event queue control module restores the distribution pointer position from the breakpoint protection area and initializes the restart sequence of the event distribution engine. The data to be processed in the high-priority cache queue is sorted by time stamps and merged into the main processing queue to restore the continuity of the data stream. The optimized touch coordinate data is subjected to display coordinate mapping through the coordinate conversion module to eliminate the spatio-temporal reference deviation generated during the calibration process.

[0067] After the calibration process is completed, the calibration status monitoring thread starts the anomaly detection mechanism, and compares the statistical characteristics of the contact coordinate residual matrix before and after calibration. When it detects that the decrease amplitude of the residual distribution entropy value does not reach the preset standard, it triggers the secondary calibration process and calls earlier calibration parameter combinations in the historical database for iterative optimization. This monitoring mechanism is synchronized with the main processing thread through the verification flag bit of the status register to maintain the self-repair ability of the system in the abnormal state.

[0068] Specifically, for the touch data processing method based on an organic display according to the present invention, the frame synchronization alignment process includes: Adjusting the timing phase of the touch coordinate output thread according to the vertical synchronization signal of the display driving timing; Inserting dynamic delay compensation between the touch trajectory and the display rendering, where the delay compensation amount is calculated based on the deformation gradient data collected by the strain sensor and the current display frame rate.

[0069] After the touch coordinate output thread receives the vertical synchronization signal sent by the display drive controller, the timing alignment module starts the phase-locked loop circuit for clock phase matching. This circuit dynamically adjusts the clock division coefficient of the coordinate output thread by comparing the phase difference between the touch sampling clock and the display refresh clock, so that the output moment of the touch coordinate data is synchronized with the row scanning period of the display panel. During the phase adjustment process, the timing monitoring unit tracks the leading edge jitter characteristics of the vertical synchronization signal in real time and smooths the clock deviation through a digital filtering algorithm.

[0070] The deformation gradient data processing thread continuously receives the three-axis deformation data collected by the strain sensor. The deformation acceleration component is extracted by a differential circuit and input into the delay compensation calculation unit. The display frame rate detection module synchronously monitors the configuration register of the display controller to obtain the actual refresh rate value in the current display mode. The compensation calculation unit constructs a non-linear mapping relationship based on the magnitude of the deformation acceleration and the reciprocal of the frame rate to generate a dynamic delay compensation coefficient matrix.

[0071] The delay compensation execution module receives the touch coordinate data stream after timing alignment, and uses a motion trajectory prediction algorithm to deduce the displacement trend of the touch point. The compensation coefficient matrix acts on the trajectory prediction result, and virtual compensation points are inserted between adjacent touch coordinates through an interpolation algorithm. Before the compensated coordinate data set is input into the display rendering pipeline, the buffer management unit performs a timestamp check to eliminate the timing misalignment error introduced by the compensation calculation.

[0072] The coordinated control of phase alignment and delay compensation is realized by a state machine. When it is detected that the mutation of the deformation gradient data causes the compensation amount to exceed the preset range, the phase resynchronization process is triggered. This process pauses the touch coordinate output thread, re-executes the clock matching operation of the phase-locked loop, and introduces a deformation mutation correction term into the delay compensation coefficient. After the compensation coefficient is updated, the touch coordinate data stream resumes output and is directly written into the input buffer of the display controller through the hardware acceleration interface.

[0073] After the display rendering engine receives the compensated touch coordinate data, the geometric transformation unit in the rendering pipeline performs a surface coordinate mapping on the touch trajectory according to the current deformation gradient data. The mapped contact position information is spatially superimposed with the display content to eliminate the visual deviation of the touch position caused by the deformation of the flexible screen. Before the final rendering data is output to the display panel, the timing verification module verifies the timing consistency between the touch event and the display frame to prevent the display tearing phenomenon caused by the compensation calculation error.

[0074] Specifically, the touch data processing method based on the organic display according to the present invention further includes: A distributed microelectromechanical system strain sensor is arranged in the flexible substrate, and the sensor shares the timing signal of the display driving circuit with the touch electrode array; The X / Y / Z axis deformation gradient data collected by the strain sensor is transmitted through a time-division multiplexing mechanism to reduce the crosstalk amplitude with the signal collected by the touch electrode array.

[0075] In the manufacturing process of the flexible substrate, a distributed microelectromechanical system strain sensor array is integrated. The sensor nodes are embedded in the organic material layer in a grid layout, forming a spatially staggered distribution structure with the touch electrode array. The sensor driving circuit shares the row scanning timing signal of the display panel with the touch electrode driving unit, and alternately activates the touch signal acquisition period and the sensor data reading period through a time-division multiplexing control module. The timing synchronization signal is generated by dividing the main clock of the display driving controller to eliminate the clock phase difference between different functional modules.

[0076] The X / Y / Z axis data acquisition channels of the strain sensor nodes are configured with independent signal conditioning circuits. After the output signals of the axial sensing units are adjusted in amplitude by a programmable gain amplifier, they are input to different channels of the multiplexer. The time-division multiplexing controller polls and switches the channel selection status of the multiplexer according to a preset time slice, and completes the transmission of the sensor data during the charge accumulation stage of the touch electrode. This mechanism completely avoids the sensor signal transmission period from the touch signal sampling window, isolating the transmission paths of the two types of signals in the time dimension.

[0077] The sensor data packet is transmitted to the data processing module through a dedicated low-voltage differential signal link, and an electromagnetic shielding layer is enabled to cover the signal trace area of the flexible substrate during the transmission period. The data receiving end is configured with an adaptive equalization circuit to compensate for the high-frequency attenuation characteristics during the signal transmission process. After the X / Y / Z axis deformation gradient data is timestamp-aligned in the receive buffer, it is input to the multi-source data fusion module for spatial registration to generate a set of deformation state parameters unified with the touch signal spatio-temporal reference.

[0078] The status register of the time-division multiplexing control module monitors the driving timing matching degree between the touch electrode and the sensor in real time. When it detects that the display refresh rate is dynamically adjusted, resulting in timing out-of-step, it triggers the timing resynchronization process. This process pauses the current data transmission task, recalibrates the phase relationship between the touch driving pulse and the sensor sampling clock, and resumes the time-division multiplexing operation at the beginning of the next vertical synchronization period. The timing protection mechanism is implemented through a hardware watchdog circuit to prevent data acquisition failure caused by long-term out-of-step.

[0079] The collaborative processing of the deformation gradient data and the touch signal is achieved through a shared storage architecture. The output data of the sensor data preprocessing unit is written into a specific address area of the dual-port memory, and the touch signal processing thread directly reads it through memory mapping. This architecture avoids the transmission delay caused by bus contention and maintains the timing consistency of multi-source data processing. The storage area access control unit implements a priority scheduling strategy to temporarily increase the touch data access bandwidth during the peak period of touch signal processing and balance the system resource allocation.

[0080] In the specific implementation of the present invention, the touch electrode array collects the original touch signal in real time, and synchronously obtains the three-dimensional deformation gradient parameters output by the microelectromechanical system strain sensor embedded in the flexible substrate and the temperature and humidity parameters of the environmental sensor. After the touch signal is amplified and analog-to-digital converted by the analog front-end circuit, it is time-stamped and aligned with the deformation and environmental parameters through the hardware clock synchronization module, forming a spatio-temporally unified multi-source data stream and inputting it into the fusion processing module. The deformation gradient parameters use a fourth-order Butterworth low-pass filter to strip the high-frequency mechanical vibration noise, and the temperature and humidity parameters are smoothed by an exponentially weighted moving average algorithm with a variable time window length to smooth the transient fluctuations. The preprocessed data retains the low-frequency characteristics of material deformation and the environmental gradual change trend.

[0081] In the multi-source data fusion module, the dynamic trigger function calculates the weighted sum of the deformation acceleration component and the environmental parameter change rate in real time. When the deformation acceleration exceeds the critical threshold of material elastic deformation or the temperature and humidity change rate breaks through the preset gradient, it triggers the interrupt-level baseline calibration process. The calibration process calls the reference potential characteristic spectra of the last five times in the historical baseline database, generates a compensation amount that matches the current touch signal noise spectrum through an adaptive filtering algorithm. When the deformation acceleration exceeds the standard, it preferentially adopts the compensation strategy dominated by the deformation compensation sub-module. When the environment suddenly changes, it activates the parallelly executed environmental compensation sub-model. The outputs of both are fused with variable weights to update the touch signal threshold.

[0082] The dielectric constant offset prediction model adopts a long short-term memory neural network architecture. It inputs calibrated touch signals, deformation gradients, and environmental parameters, and deduces the offset trend of material dielectric properties through the hidden layer state. When the predicted offset enters the non-linear deformation region, the saturation compensation mode is activated to limit the maximum compensation amplitude, and the compensation boundary is dynamically updated based on the historical statistical learning results. When the spatial distribution entropy value of the contact coordinate residual matrix exceeds the optimization threshold, the LSTM network weight adjustment process is triggered, and the gradient clipping technique is used to limit the single iteration step size to prevent overfitting.

[0083] During the touch event queue generation stage, the confidence score is calculated based on the residual distribution, pressure consistency verification, and trajectory curvature change rate. Continuously low-score events trigger the fast recalibration process. During calibration, the event distribution is paused, the original data in the buffer is called for re-parse, and low-confidence contacts are filtered. After updating the compensation parameters, the processing is resumed. The display synchronization module adjusts the touch coordinate output timing according to the vertical synchronization signal, calculates the dynamic delay compensation amount by combining the deformation gradient data and the frame rate, and eliminates the visual coordinate deviation caused by the flexible screen deformation through geometric transformation.

[0084] The time-division multiplexing mechanism controls the driving timing of the touch electrodes and the strain sensors, transmits the deformation gradient data during the display line scanning interval, and the low-voltage differential signal link cooperates with the electromagnetic shielding layer to reduce signal crosstalk. The deformation data and the touch signal are processed collaboratively through a dual-port memory, and the access control unit dynamically allocates the bus bandwidth to ensure the timing consistency under high load. This implementation effectively suppresses the accumulation of touch errors in the flexible deformation scenario through dynamic trigger calibration, prediction compensation, and residual feedback optimization, and improves the interaction accuracy and response real-time performance of the organic display.

[0085] The explanations of the technical feature terms of the present invention are as follows: Dynamic Trigger Function: A real-time decision-making algorithm constructed based on the deformation acceleration component and the environmental parameter variation rate. By weighted fusion of the curvature change rate and the temperature-humidity gradient output by the deformation sensor, the baseline calibration is triggered when the flexible substrate undergoes rapid deformation or the environment changes suddenly, replacing the traditional fixed-period update mechanism.

[0086] Dielectric Constant Offset Prediction Model: A time series prediction model constructed using a Long Short-Term Memory (LSTM) neural network or a Kalman Filter. It takes calibrated touch signals, deformation gradients, and environmental parameters as inputs, and infers the dynamic offset trend of the dielectric constant of organic materials through hidden layer states, providing a prediction basis for touch signal compensation.

[0087] Baseline Calibration Process: A high-priority interrupt service routine that is initiated when the dynamic trigger function determines that the deformation or environmental parameters exceed the threshold. It calls the reference potential spectrum stored in the historical baseline database to generate an adaptive compensation value that matches the current noise spectrum, and updates the recognition threshold of the touch signal.

[0088] Contact Coordinate Residual Matrix: A two-dimensional spatial error distribution matrix that records the deviation data between the actual contact coordinates and the model-predicted coordinates. It calculates the spatial distribution entropy value through the sliding window statistics method, which is used to feedback and optimize the model weight parameters and the trigger threshold boundary.

[0089] Time-Division Multiplexing: A signal transmission strategy in which the touch electrode array and the MEMS strain sensor share the display driving timing. It transmits the deformation gradient data during the sampling gap of the touch signal, reducing signal crosstalk through timing isolation.

[0090] Frame Synchronization Alignment: Adjust the timing phase of the touch coordinate output according to the Vertical Sync Signal, and combine Dynamic Delay Compensation to eliminate the visual deviation between the touch trajectory and the display rendering caused by the deformation of the flexible screen.

[0091] LSTM Hidden Layer Weights: The parameter matrix that stores the timing correlation information in the long short-term memory neural network, updated by the Backpropagation and Gradient Clipping techniques, and used to capture the long-term dependencies of the dielectric constant offset.

[0092] Kalman Filter Process Noise Covariance Matrix: The matrix parameter that describes the statistical characteristics of the system process noise, iteratively corrected by the autocorrelation characteristics of the residual data, and optimizes the tracking accuracy of the prediction model for the drift of the organic material characteristics.

[0093] Confidence Score: The event credibility index calculated by integrating the residual distribution of the contacts, pressure consistency, and trajectory curvature, used to trigger the Fast Recalibration process and filter out low-confidence contacts.

[0094] Dual-Port Memory: The storage architecture that supports parallel access to the deformation data and touch signals, realizes the collaborative processing of multi-source data through memory mapping, and avoids the transmission delay caused by bus contention.

[0095] MEMS (Micro-Electro-Mechanical Systems): Micro-electromechanical systems, referring to the micro strain sensors integrated in the flexible substrate, used for high-precision measurement of the three-dimensional deformation gradient.

[0096] ADC (Analog-to-Digital Converter): Analog-to-Digital Converter, which converts the analog signals collected by the touch electrodes into digital signals.

[0097] SPI / I2C: Serial Peripheral Interface / Inter-Integrated Circuit Bus, which is a data transfer protocol used for the touch chip and the main control unit.

[0098] FFT (Fast Fourier Transform): Fast Fourier Transform, which is an algorithm used for the frequency-domain analysis of touch signals.

[0099] PCA (Principal Component Analysis): Principal Component Analysis, which is used to extract the main feature components of the deformation gradient parameters.

[0100] The dynamic trigger function monitors the flexible deformation acceleration (through the MEMS sensor) and the environmental parameter jumps in real time, triggering the baseline calibration to generate the adaptive compensation amount; the dielectric constant offset prediction model (LSTM / Kalman Filter) uses the compensated data to deduce the material property changes and outputs the compensation coefficient; the contact coordinate residual matrix optimizes the model parameters through the spatial entropy value analysis to form a closed-loop control; the time-division multiplexing mechanism and the frame synchronization strategy solve the signal interference and timing deviation problems. Each technical feature realizes cooperation through the data bus, the status register and the interrupt mechanism, constituting a complete adaptive touch data processing link.

[0101] The present invention constructs a dynamic trigger mechanism to replace the fixed-period calibration, solving the time-domain mismatch problem. Based on the acceleration component of the deformation gradient parameter and the environmental parameter change rate, a dynamic trigger function is constructed to monitor the flexible deformation acceleration and the temperature and humidity jump amplitude in real time. When the deformation rate or the environmental parameter change exceeds the preset threshold, the high-priority baseline calibration process is immediately triggered to generate an adaptive compensation amount that matches the current touch signal noise spectrum, realizing the dynamic adaptation of the calibration frequency to the actual deformation rate of the organic material.

[0102] The active suppression of the non-linear offset is realized through the dielectric constant offset prediction model. The real-time deformation gradient, environmental parameters and touch signals are input into the long short-term memory neural network or the Kalman filter to deduce the dynamic offset trend of the dielectric constant. According to the material deformation interval where the predicted offset amount is located, the touch signal compensation coefficient is dynamically corrected: in the linear deformation region, the compensation amplitude is scaled proportionally, and in the non-linear deformation region, the saturation compensation mode is enabled to limit the maximum compensation amount to prevent the accumulation of overshoot errors.

[0103] Introduce a residual feedback closed-loop optimization mechanism to form continuous adaptive capabilities. Based on the spatial distribution entropy value of the contact coordinate residual matrix, reverse-optimize the weight parameters and dynamic trigger threshold boundaries of the prediction model. When the residual statistical value exceeds the limit, use the sliding window statistical method to trigger model retraining, and limit the parameter adjustment step size of a single iteration to prevent overfitting. The optimized parameters act on the touch coordinate output in real time through the frame synchronization alignment mechanism, and combine with delay compensation to eliminate the spatio-temporal deviation caused by deformation, ultimately suppressing the exponential growth of the parsing error.

Claims

1. A touch data processing method based on an organic display, characterized in that: include: Receive the original touch signal collected by the touch electrode array, synchronously obtain the deformation gradient parameters output by the flexible deformation sensor and the temperature and humidity parameters collected by the environmental sensor, perform analog-to-digital conversion on the touch signal, and input the deformation gradient parameters and temperature and humidity parameters aligned with the timestamp into the multi-source data fusion module; In the multi-source data fusion module, a dynamic trigger function is constructed based on the acceleration component of the deformation gradient parameter and the change rate of the temperature and humidity parameters. When the acceleration component or the temperature and humidity change rate exceeds the corresponding preset threshold, a high priority baseline calibration process is triggered to generate an adaptive compensation amount of the reference potential threshold according to the current touch signal characteristics and historical baseline data; The touch signal, deformation gradient parameter and temperature and humidity parameter after baseline calibration are input into the dielectric constant offset prediction model, the dielectric constant offset trend of the organic material is deduced through the timing prediction algorithm, and the touch signal compensation coefficient is dynamically corrected according to the predicted offset; Based on the corrected and compensated touch signal, an interpolation algorithm is used to calculate the sub-pixel touch coordinates to generate a touch point coordinate residual matrix. When the spatial distribution statistics of the residual matrix exceed a preset tolerance range, the weight parameters of the dielectric constant offset prediction model are reversely optimized, and the threshold boundary of the dynamic trigger function is adjusted synchronously. The optimized touch coordinate data is aligned with the display drive timing for frame synchronization, and combined with the touch pressure sensor data and the contact point movement rate parameters, a touch event priority queue is generated through the event distribution engine and the response result is output.

2. The touch data processing method based on an organic display according to claim 1, characterized in that: After receiving the original touch signal collected by the touch electrode array, the method includes: The touch signal is subjected to adaptive notch filtering to eliminate periodic interference generated by the display driving circuit, the deformation gradient parameter output by the flexible deformation sensor is subjected to Butterworth low-pass filtering to remove high-frequency vibration noise, and the temperature and humidity parameters collected by the environmental sensor are subjected to exponentially weighted sliding average processing to smooth transient fluctuations; The effective amplitude envelope of the touch signal after filtering, the second-order derivative characteristics of the deformation gradient parameter and the change trend slope of the temperature and humidity parameters are extracted, and the characteristic vector is input into the characteristic input port of the dynamic trigger function described in claim 1.

3. The touch data processing method based on an organic display according to claim 1, characterized in that: The trigger baseline calibration comprises: When generating the compensation amount according to the noise spectrum density of the current touch signal and the reference value stored in the historical baseline database, if it is detected that the output value of the deformation acceleration sensor exceeds the critical threshold of material deformation, the compensation strategy dominated by the deformation gradient parameter is preferentially adopted, and the compensation amount weight is allocated to the deformation compensation submodule; When the rate of change collected by the temperature and humidity sensor crosses the environmental gradient threshold, the environmental compensation sub-model is activated and executed in parallel with the baseline calibration main process, and the output of the environmental compensation sub-model is weightedly fused with the compensation amount of the deformation compensation sub-module.

4. The touch data processing method based on an organic display according to claim 1, characterized in that: The dielectric constant shift prediction model processing includes: The input weight of the prediction model is dynamically allocated according to the real-time deformation gradient parameters and temperature and humidity parameters. When the output value of the deformation acceleration sensor exceeds the set threshold, the weight of the deformation factor in the model input layer is increased to above 0.

8. After completing a preset number of valid touch point coordinate analyses, the statistical distribution data of the touch point coordinate residual matrix is ​​used to update the LSTM network hidden layer weights of the timing prediction algorithm or the process noise covariance matrix of the Kalman filter.

5. The touch data processing method based on an organic display according to claim 1, characterized in that: The dynamic correction touch signal compensation coefficient includes: Setting a compensation boundary according to the material deformation interval where the dielectric constant predicted offset is located, wherein the deformation interval is quantified and graded by the curvature change rate collected by the flexible deformation sensor; When the predicted offset enters a nonlinear deformation zone, a saturation compensation mode is enabled to limit the maximum compensation amplitude, and the nonlinear deformation zone corresponds to a state where the deformation gradient parameter exceeds a preset curvature threshold; The compensation boundary value is generated based on the statistical learning results stored in the historical baseline database, and is periodically iterated and updated along with the optimization results of the contact point coordinate residual matrix.

6. The touch data processing method based on an organic display according to claim 1, characterized in that: After generating the contact point coordinate residual matrix, the method includes: A sliding window statistical method is used to analyze the spatial distribution entropy value of the contact coordinate residual matrix, and when the entropy value exceeds a set model optimization threshold, a parameter retraining process of the dielectric constant offset prediction model is triggered; Limit the adjustment step size of the LSTM network hidden layer weights or Kalman filter parameters in a single optimization iteration to prevent model overfitting.

7. The touch data processing method based on an organic display according to claim 1, characterized in that: The touch event queue generation includes: Calculating a touch event confidence score based on the statistical distribution of the touch point coordinate residual matrix, the consistency check result of the pressure sensor data, and the curvature change rate of the sliding track; When the confidence score of continuous touch events is lower than the set environmental gradient threshold, the current event queue is frozen and the fast recalibration process is started to re-parse the pre-processed buffered data and filter low-confidence touch points.

8. The touch data processing method based on an organic display according to claim 7, characterized in that: The rapid recalibration process includes: Synchronously updating the compensation boundary parameters of the dielectric constant offset prediction model within a preset time of the baseline calibration process; During the recalibration process, the touch event distribution thread is paused, and after the calibration is completed, the optimized touch coordinate data processing is resumed.

9. The touch data processing method based on an organic display according to claim 1, characterized in that: The frame synchronization alignment process includes: Adjusting the timing phase of the touch coordinate output thread according to the vertical synchronization signal of the display drive timing; Dynamic delay compensation is inserted between the touch track and the display rendering, and the delay compensation amount is calculated and generated based on the deformation gradient data collected by the strain sensor and the current display frame rate.

10. The touch data processing method based on an organic display according to claim 1, characterized in that: Also includes: A distributed micro-electromechanical system strain sensor is arranged in the flexible substrate, and the sensor and the touch electrode array share the timing signal of the display driving circuit; The X / Y / Z three-axis deformation gradient data collected by the strain sensor is transmitted through a time-division multiplexing mechanism, thereby reducing the crosstalk amplitude with the touch electrode array collection signal.

Images