Dynamic compensation method for driving signal of organic display panel

Abstract

The invention relates to the technical field of data processing of organic display panels, in particular to a dynamic compensation method for driving signals of an organic display panel, which comprises the following steps of: acquiring brightness data in parallel by a photoelectric sensor array in different regions, and synchronously acquiring threshold voltage offset and environment temperature parameters of a thin film transistor; and after noise filtering, inputting into an aging feature extraction module and storing in a historical database. In the feature extraction stage, convolution kernel aggregation threshold voltage data is utilized to generate a regional aging coefficient matrix, the weight of a convolution kernel is dynamically adjusted based on a lightweight neural network, and an aging feature model is optimized in combination with real-time temperature parameters. The real-time problem of pixel-level compensation in a high-resolution scene is solved, and the calculation load of an embedded system is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of organic display panel data processing, and in particular to a method for dynamic compensation of a driving signal of an organic display panel. Background Art

[0002] Dynamic compensation of the driving signal of an organic display panel is a technology that adjusts the input voltage or current parameters in real time to maintain stable display performance. Since organic light-emitting materials may experience electroluminescent efficiency decay or threshold voltage drift when working for a long time or when the ambient temperature changes, causing the brightness and chromaticity to deviate from the design target, a feedback mechanism needs to be introduced in the driving circuit. The compensation system usually collects real-time brightness data based on the built-in sensor of the panel or the external optical detection unit, and analyzes the pixel aging degree and temperature-related parameters through an embedded algorithm in combination with the electrical characteristic model of the driving thin-film transistor (TFT). The compensation process dynamically corrects the driving pulse width or amplitude of each pixel based on the nonlinear characteristics of the voltage-brightness transfer function, offsets the carrier mobility changes caused by material degradation, and enables the light-emitting unit to maintain a constant light output under the same grayscale instruction.

[0003] The dynamic compensation method for driving signals of organic display panels has a contradiction between the computational load of real-time multi-dimensional data fusion and the response delay in data processing. The compensation system needs to synchronously process high-frequency brightness sampling data from optical sensors, thin-film transistor threshold voltage drift monitoring data, and ambient temperature parameters, and calculate pixel-level compensation amounts in combination with voltage and brightness nonlinear transfer functions. As the resolution of display panels increases, the amount of pixel-level data grows exponentially, and embedded systems are limited by hardware resources and algorithm parallelization capabilities. It is difficult to complete full-pixel aging feature extraction and compensation parameter iterative calculations within the frame refresh cycle, resulting in the compensation response lagging behind the drift rate of material properties, which may cause instantaneous brightness fluctuations or afterimages. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method for dynamic compensation of driving signals of organic display panels, which solves the problem of real-time pixel-level aging feature extraction and iterative calculation of compensation parameters in high-resolution scenarios. The core contradiction lies in the imbalance between the exponentially growing amount of pixel data and the limited parallel computing capability of the embedded system, which causes the compensation response to lag behind the drift of material properties and cause display abnormalities.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: The present invention provides a method for dynamic compensation of a driving signal of an organic display panel, comprising: The photoelectric sensor array collects brightness data in parallel in different regions, synchronously obtains the threshold voltage offset data of the thin film transistor voltage monitoring module and the temperature parameters of the ambient temperature sensor, and inputs the brightness data, threshold voltage offset data and temperature parameters into the aging feature extraction module after noise filtering, and stores the filtered data into the historical database; In the aging feature extraction module, the threshold voltage offset data is aggregated using a convolution kernel to generate a regional aging coefficient matrix, and based on the aging feature data and real-time temperature data stored in the historical database, the weight parameters of the convolution kernel are dynamically adjusted through a lightweight neural network, and the regional aging coefficient matrix is ​​output to a pixel-level calibration module; In the pixel-level calibration module, the driving pulse duty cycle is adjusted according to the aging rate parameter in the regional aging coefficient matrix, the voltage amplitude adjustment gradient is generated based on the piecewise linearization result of the voltage-brightness transfer function, the compensation weight is dynamically allocated in combination with the correlation constraint of the compensation amount between pixels stored in the historical database, and the compensation parameter is output to the driving signal generation module; In the distributed iteration module, the screen is divided into multiple logic units and the compensation parameters are updated frame by frame. The sliding window division module is used to dynamically adjust the division boundary of the next frame logic unit. The sliding window mechanism is combined with the current frame residual data output by the driving signal generation module to dynamically adjust the division boundary and iteration priority of the next frame logic unit, and output the updated compensation parameters to the driving module. In the closed-loop feedback module, the compensation weight coefficient is dynamically corrected according to the real-time residual distribution data fed back by the driving module, and the aging feature model parameters in the aging feature extraction module are optimized after accumulating multiple frames of residual data. A global optimization instruction is generated to synchronously update the convolution kernel weight of the aging feature extraction module, the gradient parameters of the compensation calculation engine and the data acquisition strategy. At the same time, a lightweight compensation knowledge base is constructed based on the key features in the historical database to initialize the compensation parameters of the new panel.

[0006] Furthermore, in the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the parallel collection of brightness data in different regions includes: dynamically adjusting the regional division weight based on the aging data stored in the historical database, increasing the sampling point density for the high-frequency aging area, establishing a temperature gradient prediction model according to the temperature history data in the historical database, predicting the trend of ambient temperature changes and optimizing the activation mode of the sensor array in real time, and after adjusting the integration time and gain parameters of the sensor, synchronously transmitting the noise-filtered brightness data, threshold voltage offset data and temperature parameters to the aging feature extraction module and the historical database.

[0007] Furthermore, in the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the regional-level aging feature extraction includes: embedding an online learning module in the embedded system of the aging feature extraction module, back-propagating fine-tuning the hyperparameters of the feature extraction model based on the compensated error data in the historical database, synchronizing the updated convolution kernel weights to the regional division module in real time to optimize the regional weight allocation for subsequent data acquisition, outputting the generated regional aging coefficient matrix to a pixel-level calibration module, and feeding back the error signal to the online learning module to update the parameters of the feature extraction model.

[0008] Furthermore, in the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the generation of pixel-level compensation parameters comprises: dynamically switching the segmented granularity of compensation calculation according to the real-time load state of the processor fed back by the driving signal generating module; when a high load state is detected, coarse-grained compensation is used to generate a driving pulse duty cycle adjustment instruction; when a low load state is detected, switching to fine-grained compensation to generate a voltage amplitude adjustment gradient; the output compensation weight distribution result is transmitted to the driving module; and the load state data is fed back to the compensation strategy controller to adjust the subsequent compensation calculation strategy.

[0009] Furthermore, in the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the execution of the distributed iteration module includes: embedding a dynamic priority algorithm in a task scheduler, adjusting the update order of the logic units in combination with the aging rate parameters in the regional aging coefficient matrix output by the pixel-level calibration module, generating a priority adjustment signal through a pre-verification module based on the current frame residual data fed back by the driving module and returning the signal to the sliding window division module to dynamically adjust the next frame logic unit division boundary, and synchronously inputting the compensation parameters after iteration into the driving module and the residual analysis module.

[0010] Furthermore, in the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the closed-loop feedback module executes the following steps: constructing a cross-layer feedback path to transmit the residual data output by the residual analysis module back to the data acquisition module, dynamically adjusting the coverage range and sensor activation frequency of the high-frequency sampling area, and inputting the optimized acquisition strategy data and the parameters in the lightweight compensation knowledge base into the new panel initialization module to reduce the cold start time of the new panel compensation model.

[0011] Furthermore, the method for dynamic compensation of driving signals of an organic display panel described in the present invention also includes: extracting key features from a historical database in the data compression and storage stage to construct a lightweight compensation knowledge base, combining a shared memory pool to realize real-time synchronization of the convolution kernel weight parameters of the aging feature extraction module, the regional weight strategy of the regional division module, and the weight coefficient of the compensation strategy controller, and completing the adaptive optimization of complex aging scenarios through a collaborative mechanism of real-time updating of compensation parameters by an online fine-tuning module and updating of a global aging feature model by an offline training module.

[0012] Beneficial effects of the present invention: The beneficial effect of the present invention lies in that the data processing load of the high-resolution panel is effectively reduced through the regional parallel acquisition and dynamic weight allocation mechanism. The photoelectric sensor array optimizes the sampling point density based on the historical aging characteristics, and dynamically adjusts the sensor activation mode in combination with the temperature gradient prediction model, thereby reducing redundant data acquisition while improving the monitoring accuracy of key areas; the lightweight neural network updates the convolution kernel weights in real time through the online learning module, and uses the generation of the regional aging coefficient matrix to achieve rapid aggregation of local aging characteristics, thereby reducing the computational complexity of the embedded system; the distributed iteration module adopts the sliding window mechanism to update the compensation parameters in frames, and dynamically adjusts the logical unit division boundaries in combination with the residual data, thereby achieving a dynamic balance between computing resources and compensation accuracy; the closed-loop feedback system continuously corrects the compensation weight coefficient through the cross-layer optimization path, and shortens the initialization time of the new panel in combination with the preloading mechanism of the lightweight compensation knowledge base, thereby forming a multi-dimensional collaborative optimization system for adaptive aging scenarios, and significantly improving the real-time performance of the drive signal compensation and the stability of the display effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0014] Figure 1 The present invention provides a flowchart of a method for dynamic compensation of a driving signal of an organic display panel according to an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described in detail below.

[0016] See also Figure 1 The present invention provides a method for dynamic compensation of a driving signal of an organic display panel, comprising: Step S101, collecting brightness data in parallel by using a photoelectric sensor array in different regions, synchronously acquiring threshold voltage offset data of a thin film transistor voltage monitoring module and temperature parameters of an ambient temperature sensor, inputting the brightness data, threshold voltage offset data and temperature parameters into an aging feature extraction module after noise filtering, and storing the filtered data into a historical database; Step S102, in the aging feature extraction module, the threshold voltage offset data is aggregated using a convolution kernel to generate a regional aging coefficient matrix, based on the aging feature data and real-time temperature data stored in the historical database, the weight parameters of the convolution kernel are dynamically adjusted through a lightweight neural network, and the regional aging coefficient matrix is ​​output to a pixel-level calibration module; Step S103, in the pixel-level calibration module, the driving pulse duty cycle is adjusted according to the aging rate parameter in the regional aging coefficient matrix, a voltage amplitude adjustment gradient is generated based on the piecewise linearization result of the voltage-brightness transfer function, compensation weights are dynamically allocated in combination with the correlation constraint of the compensation amount between pixels stored in the historical database, and compensation parameters are output to the driving signal generation module; Step S104, in the distributed iteration module, the screen is divided into multiple logic units and the compensation parameters are updated frame by frame, the division boundary of the next frame logic unit is dynamically adjusted by the sliding window division module, the division boundary and iteration priority of the next frame logic unit are dynamically adjusted by the sliding window mechanism combined with the current frame residual data output by the driving signal generation module, and the updated compensation parameters are output to the driving module; Step S105, in the closed-loop feedback module, the compensation weight coefficient is dynamically corrected according to the real-time residual distribution data fed back by the driving module, the aging feature model parameters in the aging feature extraction module are optimized after accumulating multiple frames of residual data, and a global optimization instruction is generated to synchronously update the convolution kernel weight of the aging feature extraction module, the gradient parameters of the compensation calculation engine and the data acquisition strategy, and at the same time, a lightweight compensation knowledge base is constructed based on the key features in the historical database to initialize the compensation parameters of the new panel.

[0017] In the method for dynamic compensation of driving signals of an organic display panel provided by the present invention, the parallel collection of brightness data in different regions specifically includes the following implementation methods: based on the aging area distribution data in the historical aging feature database, the screen area is dynamically divided and different weight values ​​are assigned, and the sampling point density of the photoelectric sensor is increased preferentially in areas with high brightness attenuation or significant threshold voltage drift. Real-time temperature parameters are obtained through an ambient temperature sensor, and a temperature gradient prediction model is constructed in combination with the temperature change trend in the historical database to predict the temperature fluctuation range in the future time window. The activation mode of the sensor array is optimized according to the prediction results, and the sensor integration time and gain parameters of different areas are adjusted to reduce the impact of environmental noise on the brightness data. The brightness data, threshold voltage offset data and temperature parameters after noise filtering are standardized and synchronously transmitted to the aging feature extraction module and the historical database to provide multi-dimensional input for subsequent feature extraction.

[0018] In the aging feature extraction module, the convolution kernel is used to aggregate the threshold voltage offset data, and the weight of the convolution kernel is dynamically adjusted in combination with the real-time temperature parameters. Specifically, the correlation between the aging features and temperature in the historical database is analyzed through a lightweight neural network, and a convolution kernel weight configuration matching the current temperature is generated, and the threshold voltage offset data is mapped to a regional aging coefficient matrix. This matrix characterizes the electroluminescent efficiency decay rate and material degradation degree in different regions. The online learning module receives the compensation error data in real time, fine-tunes the hyperparameters of the feature extraction model through the back-propagation algorithm, and synchronizes the updated convolution kernel weights to the regional division module, optimizing the regional weight allocation strategy for subsequent data collection, forming a dynamic closed-loop optimization mechanism.

[0019] The pixel-level calibration module adjusts the duty cycle of the driving pulse of the corresponding area according to the aging rate parameters in the regional aging coefficient matrix to compensate for the change in carrier mobility caused by the degradation of organic materials. Based on the piecewise linearization processing results of the voltage-brightness transfer function, the voltage amplitude adjustment gradient is generated, and the compensation weights of different pixels are dynamically allocated in combination with the correlation constraints of the compensation amounts between pixels in the historical database. The compensation strategy controller dynamically switches the granularity level of the compensation calculation according to the processor load state fed back by the driving signal generation module: a coarse-grained compensation mode is used under high load conditions to achieve rapid response by adjusting the duty cycle of the driving pulse; it switches to a fine-grained compensation mode under low load conditions, and performs fine calibration in combination with the voltage amplitude gradient to balance computing resources and compensation accuracy requirements.

[0020] The distributed iteration module divides the screen into multiple logical units and uses a sliding window mechanism to update the compensation parameters frame by frame. The task scheduler dynamically adjusts the update priority of the logical unit based on the aging rate parameters in the regional aging coefficient matrix, giving priority to processing areas where the aging rate exceeds the threshold. The pre-verification module generates an adjustment signal for the logical unit division boundary based on the current frame residual data fed back by the driver module, and optimizes the compensation parameter iteration range for the next frame through the sliding window division module. The iterated compensation parameters are synchronously input into the driver module and the residual analysis module to form a cross-frame compensation parameter update mechanism.

[0021] The closed-loop feedback module builds a cross-layer feedback path, and transmits the residual distribution data output by the residual analysis module back to the data acquisition module, dynamically adjusting the coverage of the high-frequency sampling area and the sensor activation frequency. After accumulating multiple frames of residual data, the convolution kernel weights of the aging feature extraction module and the gradient parameters of the compensation calculation engine are updated through the global optimization algorithm. After the key feature data in the historical database is compressed, a lightweight compensation knowledge base is constructed, and the shared memory pool is combined to achieve real-time synchronization of the region division strategy and compensation parameters. The initialization parameters in the knowledge base are input into the new panel compensation model to reduce the cold start time of the compensation parameters of the new panel and improve the consistency of display performance.

[0022] Specifically, in the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the parallel collection of brightness data in different regions includes: dynamically adjusting the regional division weight based on the aging data stored in the historical database, increasing the sampling point density for the high-frequency aging area, establishing a temperature gradient prediction model according to the temperature history data in the historical database, predicting the trend of ambient temperature changes and optimizing the activation mode of the sensor array in real time, and after adjusting the integration time and gain parameters of the sensor, synchronously transmitting the noise-filtered brightness data, threshold voltage offset data and temperature parameters to the aging feature extraction module and the historical database.

[0023] In the process of parallel collection of brightness data in different regions, the pixel aging index stored in the historical aging feature database is quantified as the basis for regional weight allocation. By analyzing the spatial distribution characteristics of the brightness decay rate and threshold voltage drift in the historical data, a regional aging degree quantitative model is established to calculate the aging factor score of each sub-region. The weight allocation algorithm dynamically adjusts the sampling weight of the sensor array according to the scoring results. For areas where the aging factor score exceeds the set threshold, the sampling point density is increased exponentially, and a data interpolation relationship is established between adjacent sampling points to improve the spatial resolution. The regional division module adopts a sliding window mechanism to dynamically shrink or expand the coverage of the high-frequency aging area according to the correlation coefficient between the real-time temperature parameters and the historical temperature data.

[0024] The temperature gradient prediction model is based on the temperature time series data in the historical database, and uses the autoregressive moving average algorithm to analyze the periodic characteristics and mutation trends of temperature changes. The model output includes the predicted value of the temperature change rate within the preset time window in the future, combined with the real-time reading of the current ambient temperature sensor, to generate the sensor array activation strategy optimization instruction. The sensor control module adjusts the activation frequency and sampling interval of sensors in different areas according to the predicted value of the temperature change rate: when the predicted temperature gradient exceeds the critical value, the high-frequency sampling mode is started and the overlapping coverage area of ​​adjacent sensors is reduced; when the temperature tends to stabilize, it switches to the low-frequency sampling mode to expand the sensor coverage range to reduce power consumption.

[0025] The sensor integration time and gain parameters are adjusted based on the noise spectrum analysis results. The optimal integration time window is dynamically matched by real-time monitoring of the background noise intensity of each sensor channel. For high-frequency aging areas, a short integration time combined with a high gain configuration is used to capture rapid brightness changes; for low aging risk areas, a long integration time and a low gain configuration are used to suppress random noise. The noise filtering module uses a multi-stage sliding average filtering algorithm to suppress the impulse noise in the brightness data in the time domain, and at the same time, the spatial correlation check of adjacent sensor data is used to eliminate abnormal data points caused by local temperature fluctuations.

[0026] After normalization, the filtered brightness data, threshold voltage offset data and temperature parameters are packaged into multi-dimensional data packets according to the preset protocol. The data synchronization transmission module adopts a timestamp alignment mechanism to ensure the timing consistency of the data received by the aging feature extraction module and the historical database. The historical database adopts a hierarchical storage architecture to store the real-time collected data and historical feature data by regional index classification, and provides an on-demand data interface for the online learning module. The data compression engine extracts key feature parameters before storage and removes redundant information to reduce storage load.

[0027] The dynamic optimization mechanism updates the regional division weights and sensor activation strategies in real time through closed-loop feedback. The residual analysis module compares the compensated display effect data with the target value to generate a regional residual distribution map. The feedback control loop triggers the dynamic redistribution of regional weight coefficients based on the coordinates of the abnormal area in the residual distribution map, and synchronously updates the input parameters of the temperature gradient prediction model. The activation mode of the sensor array is adaptively adjusted according to the updated weight coefficients, forming a collaborative optimization link between the data acquisition strategy and the compensation effect.

[0028] Specifically, the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the regional-level aging feature extraction includes: embedding an online learning module in the embedded system of the aging feature extraction module, back-propagating the compensated error data in the historical database to fine-tune the hyperparameters of the feature extraction model, synchronizing the updated convolution kernel weights to the regional division module in real time to optimize the regional weight allocation for subsequent data acquisition, outputting the generated regional aging coefficient matrix to the pixel-level calibration module, and feeding back the error signal to the online learning module to update the parameters of the feature extraction model.

[0029] During the regional aging feature extraction process, the online learning module built into the embedded system calls the compensation error data through the historical database to dynamically optimize the hyperparameters of the feature extraction model. The compensation error data is converted into an error gradient tensor after preprocessing. The back propagation algorithm calculates the update amount of the convolution kernel weight according to the gradient tensor, and uses the momentum optimizer to suppress weight oscillation and accelerate convergence. The updated convolution kernel weight is synchronized to the regional division module through the shared memory pool, triggering the dynamic adjustment of the regional weight allocation strategy, and giving priority to allocating computing resources to data collection tasks in areas with abnormal aging rates.

[0030] The feature extraction model adopts a lightweight convolutional neural network architecture, and its input layer receives the fusion features of the normalized threshold voltage offset data and the real-time temperature parameters. The convolution kernel aggregates the electrical characteristic parameters of the local area during the sliding process, and generates a spatial correlation feature map through the activation function. The regional aging coefficient matrix is ​​generated by downsampling the feature map through the pooling layer. Each element in the matrix corresponds to the quantized value of the aging degree of the screen sub-area, including the weighted comprehensive index of the electroluminescent efficiency attenuation rate and the carrier mobility change rate. After verification, the matrix data is transmitted to the pixel-level calibration module as the reference parameter for adjusting the drive signal.

[0031] The error signal feedback mechanism extracts the deviation between the compensated brightness data and the target value through the residual analysis module to generate a regional error distribution map. The online learning module correlates the error distribution map with the current aging coefficient matrix to identify the direction of parameter deviation in the feature extraction model. In the back-propagation process, a layered gradient clipping strategy is adopted to set update thresholds for the gradients of the convolutional layer and the fully connected layer respectively to prevent model instability caused by gradient explosion. The optimized hyperparameters are written into the feature extraction model through the real-time task scheduler of the embedded system to complete the online iteration of the model parameters.

[0032] The dynamic weight synchronization mechanism adopts a double buffer storage structure to maintain the integrity of the original weight data when updating the convolution kernel weight. After receiving the updated weight parameters, the region division module reconstructs the region division priority queue and recalculates the sampling point density coefficient of each sub-region. The updated region division strategy is sent to the photoelectric sensor array through the sensor control interface to form a coordinated optimization of feature extraction accuracy and data acquisition efficiency. The generation frequency of the aging coefficient matrix is ​​synchronized with the screen refresh rate to avoid compensation parameter lag caused by cross-frame data misalignment.

[0033] The closed-loop optimization process achieves adaptive evolution of the feature extraction model through real-time interaction between error signals and weight parameters. The model verification unit built into the online learning module regularly performs forward reasoning verification on the updated feature extraction model and compares the matching degree between the output matrix and the measured aging data. When the matching degree is lower than the preset threshold, the model rollback mechanism is triggered to restore to the stable weight version, and the offline training module is started to optimize the global parameters to maintain the robustness and real-time performance of the compensation system.

[0034] Specifically, the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the generation of pixel-level compensation parameters includes: dynamically switching the segmented granularity of compensation calculation according to the real-time load state of the processor fed back by the driving signal generation module, using coarse-grained compensation to generate a driving pulse duty cycle adjustment instruction when a high load state is detected, and switching to fine-grained compensation to generate a voltage amplitude adjustment gradient when a low load state is detected, the output compensation weight distribution result is transmitted to the driving module, and the load state data is fed back to the compensation strategy controller to adjust the subsequent compensation calculation strategy.

[0035] During the generation of pixel-level compensation parameters, the real-time load status of the processor is monitored by obtaining the current thread queue length and computing resource occupancy rate through the task scheduler. The compensation strategy controller establishes a load status classification model, quantifies the processor load into three threshold intervals of high, medium, and low, and configures the corresponding compensation calculation mode switching rules. When it is detected that the thread queue length exceeds the preset upper limit or the computing resource occupancy rate is continuously higher than the critical value, the coarse-grained compensation mode is triggered, and the screen is divided into macroblock units for batch processing to reduce the data throughput of the correlation calculation between pixels.

[0036] In the coarse-grained compensation mode, the generation of the drive pulse duty cycle adjustment instruction adopts the regional average aging coefficient algorithm. According to the average aging rate of each macroblock in the regional aging coefficient matrix, combined with the linear approximation segment of the voltage-brightness transfer function, the duty cycle compensation increment of the pixels in the macroblock is calculated. The compensation weight allocation module adopts the regional interpolation algorithm to establish a transition zone at the macroblock boundary to smooth the compensation parameter differences between adjacent macroblocks. The generated duty cycle adjustment instruction is transmitted in parallel to the pixel drive circuit of the target area through the drive signal bus, shortening the instruction transmission delay under high load conditions.

[0037] The activation of the fine-grained compensation mode is determined based on the duration of the processor load falling back to the low threshold interval. Under low load conditions, the voltage amplitude adjustment gradient is calculated using the pixel-level voltage-brightness transfer function differentiation method, and the slope of the local transfer curve is fitted according to the historical data of the driving voltage of adjacent pixels. The gradient generation module combines the constraints of the correlation of the compensation amount between pixels and solves the optimal compensation weight allocation scheme through the Lagrange multiplier method to balance the compensation accuracy and computational complexity. The compensation weight allocation result is written to the dual-port memory after verification, and the driver module reads it in real time according to the pixel coordinate index.

[0038] The compensation strategy controller has a built-in adaptive learning mechanism to analyze the correlation between load state switching events and compensation effect data. By statistically analyzing the distribution characteristics of residual data caused by coarse-grained compensation under high load conditions, the macroblock partition size and transition band width parameters are dynamically adjusted. After the load state data and compensation mode switching records are aligned by timestamp, they are input into the strategy optimization engine to generate an iterative version of the compensation calculation mode switching rules. The updated rule parameters are broadcast to each functional module through the system bus to achieve dynamic optimization of the compensation strategy.

[0039] The data feedback path adopts a priority queue management mechanism to classify and transmit the execution efficiency indicators of the driver module and the compensation parameter verification results. The compensation strategy controller allocates different processing threads and storage buffers according to the type and urgency of the feedback data. High-priority data directly triggers the compensation parameter recalculation process, and low-priority data starts batch processing after accumulating to the preset batch. The feedback control loop and the load monitoring module form a collaborative mechanism to maintain a dynamic balance between compensation computing resource allocation and display quality requirements.

[0040] Specifically, the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the execution of the distributed iteration module includes: embedding a dynamic priority algorithm in a task scheduler, adjusting the update order of the logic unit in combination with the aging rate parameters in the regional aging coefficient matrix output by the pixel-level calibration module, based on the current frame residual data fed back by the driving module, generating a priority adjustment signal through a pre-verification module and returning it to the sliding window division module to dynamically adjust the next frame logic unit division boundary, and synchronously inputting the compensation parameters after iteration into the driving module and the residual analysis module.

[0041] During the execution of the distributed iteration module, the task scheduler receives the regional aging coefficient matrix output by the pixel-level calibration module and extracts the aging rate parameters of each logical unit as the priority calculation basis. The dynamic priority algorithm constructs a weighted evaluation model of aging rate and residual amplitude, divides the logical units into emergency update queues and regular update queues, and preferentially allocates computing resources to areas where the aging rate exceeds the critical threshold or the residual accumulation increases. The queue division results are sent to each computing node through the system bus to drive the parallel compensation parameter calculation of the multi-core processor.

[0042] The pre-verification module analyzes the residual data of the current frame fed back by the driving module, constructs a residual space distribution heat map and performs superposition analysis with the logical unit division boundary. When it is detected that the residual peak area crosses multiple logical unit boundaries, a logical unit merge instruction is generated; for areas where the residual distribution is concentrated and does not cross the boundary, a logical unit split instruction is generated. After verification, the priority adjustment signal is input into the sliding window division module to trigger the dynamic reconstruction of the division boundary and optimize the calculation unit layout for the next frame compensation parameter iteration.

[0043] The sliding window partitioning module uses an adaptive grid generation algorithm to adjust the unit merging or splitting instructions in the signal according to the priority, and recalculate the geometric center and coverage of the logic unit. For the merged logic unit, the coverage area of ​​the sliding window is expanded and the calculation frequency of the local compensation parameters is reduced; for the split logic unit, the window size is reduced and the number of compensation iterations is increased. The partition boundary data is written into the shared memory pool after coordinate mapping, which is synchronously called by the compensation calculation engine and the residual analysis module.

[0044] The iterated compensation parameters are output in parallel through a dual-channel transmission mechanism. The main channel writes the parameters into the instruction buffer of the driver module in real time, and the auxiliary channel transmits the parameter copy to the residual analysis module for pre-verification. The residual analysis module compares the change trend of the current frame compensation parameters with the historical data, detects the parameter mutation area and generates an abnormal marking signal. The marking signal is input into the task scheduler through the feedback loop, triggering the priority re-evaluation and compensation parameter recalculation process of the corresponding logic unit.

[0045] The closed-loop iteration mechanism realizes dynamic optimization of compensation parameter partitions through the synergy of sliding window partitioning and priority adjustment. After each iteration, the compensation calculation engine updates the state identification register of the logic unit and records the number of parameter adjustments and residual improvement rate data. After statistical analysis, the state data is input into the dynamic priority algorithm to optimize the queue partition strategy of subsequent iteration cycles, forming an adaptive balance system between compensation parameter accuracy and system resource occupancy.

[0046] Specifically, the method for dynamic compensation of driving signals of an organic display panel described in the present invention, the closed-loop feedback module executes the following steps: constructing a cross-layer feedback path to transmit the residual data output by the residual analysis module back to the data acquisition module, dynamically adjusting the coverage range and sensor activation frequency of the high-frequency sampling area, and inputting the optimized acquisition strategy data and the parameters in the lightweight compensation knowledge base into the new panel initialization module to reduce the cold start time of the new panel compensation model.

[0047] During the execution of the closed-loop feedback module, the construction of the cross-layer feedback path adopts a multi-level data routing mechanism to convert the residual distribution data output by the residual analysis module into a standardized protocol format. After spatial encoding, the residual data is transmitted back to the buffer of the data acquisition module through the system bus. During the transmission process, a cyclic redundancy check code is embedded to maintain data integrity. The dynamic adjustment of the high-frequency sampling area is based on the hot spot distribution map of the residual amplitude. The adaptive grid subdivision algorithm is started for the area where the residual exceeds the threshold for multiple consecutive frames, increasing the spatial density of the sampling points, while reducing the sensor activation frequency in non-critical areas to reduce system power consumption.

[0048] The optimization of sensor activation frequency adopts a joint evaluation strategy of historical residual data and real-time aging rate. By analyzing the spatiotemporal correlation characteristics of residual data, a sensor activation frequency prediction model is established, and a burst sampling mode is started for areas where the residual volatility is higher than the preset value. The sensor control module dynamically configures the sampling interval and working cycle of sensors in each area based on the output results of the prediction model, balancing resource consumption while maintaining data collection accuracy.

[0049] The fusion of the optimized acquisition strategy data and the lightweight compensation knowledge base adopts a feature matching mechanism to extract the benchmark parameter set corresponding to the new panel model from the knowledge base. The data interface module performs a weighted fusion of the key parameters in the real-time acquisition strategy and the benchmark parameters in the knowledge base to generate an initialization parameter matrix. After the matrix data is normalized, it is written into the parameter register of the new panel initialization module through a dual-channel verification mechanism to eliminate data deviation during the transmission process.

[0050] The new panel initialization module uses parameter preheating loading technology to inject the baseline aging characteristic data in the compensation knowledge base in advance during the panel startup phase. During the initialization process, a shadow buffer of the compensation model parameters is established to pre-calculate the compensation weight coefficients before the drive signal is generated. Cold start time optimization is achieved by executing parameter loading and hardware self-test processes in parallel, reducing the waiting delay of the traditional serial initialization process.

[0051] The synergy between the cross-layer feedback mechanism and the knowledge base forms a dynamic evolution system for compensation parameters. The long-term accumulation of residual data triggers the incremental update process of the knowledge base, and the feature extraction engine is used to screen compensation parameter combinations with generalization capabilities. The updated knowledge base parameters use the version management module to implement the backtracking and recovery functions of historical data, maintaining the robustness of the compensation system under abnormal conditions. The parameter preheating mechanism of the initialization module and the dynamic acquisition strategy form a closed loop, improving the compensation parameter adaptation efficiency of different batches of panels.

[0052] Specifically, the method for dynamic compensation of driving signals of an organic display panel described in the present invention also includes: extracting key features from a historical database in the data compression and storage stage to construct a lightweight compensation knowledge base, combining a shared memory pool to realize real-time synchronization of the convolution kernel weight parameters of the aging feature extraction module, the regional weight strategy of the regional division module, and the weight coefficient of the compensation strategy controller, and completing the adaptive optimization of complex aging scenarios through a collaborative mechanism of real-time updating of compensation parameters by an online fine-tuning module and updating of a global aging feature model by an offline training module.

[0053] In the data compression and storage stage, the historical database uses a feature importance evaluation algorithm to screen key aging feature parameters, and constructs a lightweight compensation knowledge base by combining principal component analysis dimensionality reduction processing with feature hash coding. The selection of key feature parameters is based on the variance contribution rate and temperature correlation index of the aging coefficient of each region, and the feature dimensions that have a significant impact on the compensation weight allocation are retained first. The compressed knowledge base is stored in a hierarchical index structure, and the convolution kernel weights, regional division strategies, and compensation control parameters are classified and encoded according to functional modules to improve data retrieval efficiency.

[0054] The real-time synchronization mechanism of the shared memory pool realizes parameter interaction between multiple modules through the memory address mapping table, and establishes data association channels between the convolution kernel weight parameters of the aging feature extraction module, the weight strategy of the region division module, and the coefficients of the compensation strategy controller. The memory pool manager adopts a double buffer switching strategy to maintain the integrity of the original data copy when writing new parameters, and controls the read and write timing between modules through the semaphore mechanism to avoid parameter inconsistency caused by data competition. The parameter verification algorithm is embedded in the synchronization process, and the cyclic redundancy check and threshold range comparison are performed on the data before and after transmission to eliminate parameter distortion caused by transmission noise.

[0055] The collaboration between the online fine-tuning module and the offline training module is achieved through the task priority scheduling mechanism. The online fine-tuning module receives the residual data fed back by the driving module in real time, and uses the stochastic gradient descent algorithm to locally optimize the compensation parameters. The update results are synchronized to the compensation calculation engine in real time through the shared memory pool. The offline training module periodically calls the accumulated data in the historical database, updates the network structure and hyperparameters of the global aging feature model through batch training, and imports the trained model parameters into the online system after version management. The two modules coordinate the parameter update process through the mutex mechanism, suspend the online fine-tuning operation during the offline model import, and maintain system stability.

[0056] The adaptive optimization mechanism dynamically adjusts the feature matching strategy based on the real-time collected ambient temperature and panel usage time data. When a temperature change or continuous working timeout scenario is detected, the dynamic loading process of the compensation knowledge base is triggered, and the historical parameter combination with the highest similarity to the current working condition is matched from the lightweight knowledge base. During the matching process, the cosine similarity algorithm is used to calculate the correlation between the real-time feature vector and the knowledge base sample, and the optimal parameter set is selected for incremental loading. The loaded parameters are locally calibrated through the online fine-tuning module to form a rapid response capability to complex aging scenarios.

[0057] The incremental update process of the lightweight knowledge base is implemented through the feature fusion engine, which performs weighted fusion of the new parameters generated by online fine-tuning with historical data. The fusion weight is dynamically allocated according to the signal-to-noise ratio index of the parameter, and a higher weight is given to the parameters with significant residual improvement effect. The updated knowledge base retains a copy of the historical parameters through the version iteration mechanism, and starts the version rollback function when parameter degradation is detected to restore to a stable state. The interface between the knowledge base data and the new panel initialization module adopts a dynamic loading protocol. During the panel startup phase, the compensation parameters are pre-loaded in batches by region, and are injected synchronously with the drive signal generation timing to shorten the parameter configuration time in the initialization phase.

[0058] In a specific embodiment of the present invention, the photoelectric sensor array dynamically configures the sampling strategy based on the aging distribution data of the historical database, adopts dense grid division for high-frequency aging areas and increases the sampling frequency to 1.5 times the baseline value, and reduces the sampling frequency of non-critical areas to 0.8 times. When the ambient temperature fluctuates by more than ±3°C, the temperature gradient prediction model is started, and the sensor integration time is optimized to an adjustable range of 5-15ms. The aging feature extraction module uses a 3×3 deformable convolution kernel to spatially aggregate the threshold voltage offset data, and the lightweight neural network updates the convolution kernel weight coefficient for each frame. The regional aging coefficient matrix is ​​generated in combination with the temperature compensation factor, and the matrix element value range is set to [0,1] to characterize the aging degree of each sub-area. The pixel-level calibration module divides the screen into 8×8 macroblock units according to the aging coefficient matrix, and adopts the macroblock-level duty cycle compensation mode under high load conditions, with the duty cycle adjustment step set to 100ns; when the load is low, it switches to 4×4 pixel units for voltage gradient compensation, and the voltage amplitude adjustment accuracy reaches 10mV level. The sliding window size of the distributed iteration module is dynamically adjusted according to the residual distribution. The window size of the residual peak area is reduced to 32×32 pixels and the iteration frequency is increased to twice per frame. The window size of the low residual area is expanded to 128×128 pixels and the iteration interval is extended to three frames. The closed-loop feedback module builds an association model between residual data and sensor activation strategy. When the residual standard deviation exceeds the set threshold, the high-frequency sampling area expansion instruction is triggered to increase the sensor activation frequency in the affected area to 200Hz. The optimized compensation parameters are compressed and stored in the knowledge base to form a 32-bit feature vector. When the new panel is initialized, the feature vector is preloaded and the parameters are calibrated within the range of ±15% through the online fine-tuning module, so that the cold start time is shortened compared with the traditional solution.

[0059] The present invention addresses the data volume challenge in high-resolution scenarios through a mechanism of regional parallel acquisition and dynamic feature extraction. The photoelectric sensor array dynamically divides high-frequency aging areas based on historical aging data, and improves the acquisition efficiency of key areas by increasing the sampling density and optimizing the sensor activation mode. The threshold voltage offset data and ambient temperature parameters collected synchronously are filtered after noise, and then input into a lightweight neural network to generate a regional aging coefficient matrix. The matrix dynamically adjusts and aggregates local aging features through convolution kernel weights to reduce the amount of redundant data for full-pixel calculations. The synergy of regional division weights and temperature gradient prediction models concentrates computing resources on real-time aging-significant areas, reducing the parallel processing load of embedded systems.

[0060] The distributed iteration module uses a sliding window mechanism and a dynamic priority algorithm to divide the screen into multiple logical units and update the compensation parameters frame by frame. The task scheduler dynamically adjusts the update order and boundary division of the logical units according to the aging rate parameters and the residual data distribution, giving priority to areas where the aging rate exceeds the limit or the residual accumulation increases. The sliding window division module optimizes the iteration range of the next frame in combination with the pre-verification results, and updates the compensation parameters incrementally by frame, avoiding the computational pressure of full-pixel iteration in a single frame. The parallel generation of compensation parameters and the real-time verification of the residual analysis module form a cross-frame coordination mechanism to maintain the timeliness of compensation response under limited hardware resources.

[0061] The closed-loop feedback module builds a cross-layer optimization path and transmits the residual data back to the data acquisition and feature extraction module to form an adaptive adjustment link for the compensation parameters. The lightweight compensation knowledge base extracts historical key features and compresses them for storage, and combines them with a shared memory pool to achieve real-time synchronization of multi-module parameters and reduce data transmission delays. During the initialization phase of the new panel, the benchmark parameters of the knowledge base are preloaded, and the local calibration of the online fine-tuning module and the global model update of the offline training module are combined to dynamically balance the compensation accuracy and computing resource consumption. The multi-module collaborative mechanism achieves rapid tracking and compensation of material property drift by continuously optimizing the acquisition strategy, compensation weights, and logic unit division.

Claims

1. A method for dynamic compensation of a driving signal of an organic display panel, characterized in that: include: The photoelectric sensor array collects brightness data in parallel in different regions, synchronously obtains the threshold voltage offset data of the thin film transistor voltage monitoring module and the temperature parameters of the ambient temperature sensor, and inputs the brightness data, threshold voltage offset data and temperature parameters into the aging feature extraction module after noise filtering, and stores the filtered data into the historical database; In the aging feature extraction module, the threshold voltage offset data is aggregated using a convolution kernel to generate a regional aging coefficient matrix, and based on the aging feature data and real-time temperature data stored in the historical database, the weight parameters of the convolution kernel are dynamically adjusted through a lightweight neural network, and the regional aging coefficient matrix is ​​output to a pixel-level calibration module; In the pixel-level calibration module, the driving pulse duty cycle is adjusted according to the aging rate parameter in the regional aging coefficient matrix, the voltage amplitude adjustment gradient is generated based on the piecewise linearization result of the voltage-brightness transfer function, the compensation weight is dynamically allocated in combination with the correlation constraint of the compensation amount between pixels stored in the historical database, and the compensation parameter is output to the driving signal generation module; In the distributed iteration module, the screen is divided into multiple logic units and the compensation parameters are updated frame by frame. The sliding window division module is used to dynamically adjust the division boundary of the next frame logic unit. The sliding window mechanism is combined with the current frame residual data output by the driving signal generation module to dynamically adjust the division boundary and iteration priority of the next frame logic unit, and output the updated compensation parameters to the driving module. In the closed-loop feedback module, the compensation weight coefficient is dynamically corrected according to the real-time residual distribution data fed back by the driving module, and the aging feature model parameters in the aging feature extraction module are optimized after accumulating multiple frames of residual data. A global optimization instruction is generated to synchronously update the convolution kernel weight of the aging feature extraction module, the gradient parameters of the compensation calculation engine and the data acquisition strategy. At the same time, a lightweight compensation knowledge base is constructed based on the key features in the historical database to initialize the compensation parameters of the new panel.

2. The method for dynamic compensation of driving signals of an organic display panel according to claim 1, characterized in that: The regional parallel collection of brightness data includes: dynamically adjusting the regional division weight based on the aging data stored in the historical database, increasing the sampling point density for the high-frequency aging area, establishing a temperature gradient prediction model according to the temperature history data in the historical database, predicting the trend of ambient temperature changes and optimizing the activation mode of the sensor array in real time, and after adjusting the integration time and gain parameters of the sensor, synchronously transmitting the noise-filtered brightness data, threshold voltage offset data and temperature parameters to the aging feature extraction module and the historical database.

3. The method for dynamic compensation of driving signals of an organic display panel according to claim 2, characterized in that: The regional-level aging feature extraction includes: embedding an online learning module in the embedded system of the aging feature extraction module, fine-tuning the hyperparameters of the feature extraction model based on the back-propagation of the compensated error data in the historical database, synchronizing the updated convolution kernel weights to the regional division module in real time to optimize the regional weight allocation for subsequent data acquisition, outputting the generated regional aging coefficient matrix to the pixel-level calibration module, and feeding back the error signal to the online learning module to update the parameters of the feature extraction model.

4. The method for dynamic compensation of driving signals of an organic display panel according to claim 3, characterized in that: The pixel-level compensation parameter generation includes: dynamically switching the segmented granularity of the compensation calculation according to the real-time load state of the processor fed back by the drive signal generation module; when a high load state is detected, coarse-grained compensation is used to generate a drive pulse duty cycle adjustment instruction; when a low load state is detected, switching to fine-grained compensation to generate a voltage amplitude adjustment gradient; the output compensation weight distribution result is transmitted to the drive module; and the load state data is fed back to the compensation strategy controller to adjust the subsequent compensation calculation strategy.

5. The method for dynamic compensation of driving signals of an organic display panel according to claim 4, characterized in that: The distributed iteration module execution includes: embedding a dynamic priority algorithm in a task scheduler, adjusting the update order of the logic unit in combination with the aging rate parameters in the regional aging coefficient matrix output by the pixel-level calibration module, generating a priority adjustment signal through a pre-verification module based on the current frame residual data fed back by the driving module and returning it to the sliding window division module to dynamically adjust the next frame logic unit division boundary, and synchronously inputting the iterated compensation parameters into the driving module and the residual analysis module.

6. The method for dynamic compensation of driving signals of an organic display panel according to claim 5, characterized in that: The execution of the closed-loop feedback module includes: constructing a cross-layer feedback path to transmit the residual data output by the residual analysis module back to the data acquisition module, dynamically adjusting the coverage range and sensor activation frequency of the high-frequency sampling area, and inputting the optimized acquisition strategy data and the parameters in the lightweight compensation knowledge base into the new panel initialization module to reduce the cold start time of the new panel compensation model.

7. The method for dynamic compensation of driving signals of an organic display panel according to claim 6, characterized in that: Also includes: In the data compression and storage stage, key features are extracted from the historical database to build a lightweight compensation knowledge base. The convolution kernel weight parameters of the aging feature extraction module, the regional weight strategy of the regional division module and the weight coefficient of the compensation strategy controller are synchronized in real time in combination with the shared memory pool. The online fine-tuning module updates the compensation parameters in real time and the offline training module updates the global aging feature model through a collaborative mechanism to complete the adaptive optimization of complex aging scenarios.

Images