LED display matrix dynamic control system

Abstract

The invention relates to the technical field of control and adjustment systems, in particular to an LED display matrix dynamic control system, which comprises a time sequence feature extraction module, a self-adaptive threshold calibration module, a self-adaptive threshold calibration module, a self-adaptive threshold calibration module, a self-adaptive threshold calibration module, a self-adaptive threshold calibration module and a self-adaptive threshold calibration module, wherein the time sequence feature extraction module is used for collecting time domain waveforms of scanning signals and data signals; a display scene is recognized based on a k-nearest neighbor classifier, a composite reference value is generated in combination with historical optimization parameters, a hierarchical compensation execution module is used for executing a three-level compensation strategy according to the composite reference value to correct time sequence deviation, a dynamic model updating module is used for optimizing scene recognition and threshold generation parameters through a genetic algorithm, and a scene library is periodically updated. All the modules cooperate to form a closed-loop link, static, dynamic, high-speed motion and temperature change scenes are dynamically adapted, and the time sequence synchronous control precision and the system robustness of the row driving circuit and the column driving circuit are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical signal timing synchronization control for dynamic control of an LED display matrix, and particularly to an LED display matrix dynamic control system. Background Art

[0002] An LED display matrix is based on an active matrix display structure, and its core is composed of pixel units arranged in rows and columns. Each pixel unit integrates a driving element. The TFT driving element serves as a switching device, and controls the on-off of the source-drain current through a gate voltage signal, thereby regulating the injection current of the LED diode. The row driving circuit sequentially applies a scanning signal to the gates of the TFTs of each row of pixel units, and the column driving circuit synchronously transmits data voltages to the sources of the TFTs of each column of pixel units. The two cooperate to achieve row-by-row addressing and data writing of the pixel units. This control logic based on array driving converts electrical signals into changes in the luminous intensity of the LED diodes through the timing cooperation of row scanning and column data, and finally forms an image display on a two-dimensional plane. The driving control process of this structure involves links such as signal timing generation, switching element regulation, and conversion between electrical signals and optical signals.

[0003] The technical pain point of the LED display matrix in the dynamic control system lies in the insufficient timing synchronization control accuracy between the row driving circuit and the column driving circuit. During the dynamic display process, image data needs to be written row by row into the pixel units through the timing cooperation of the row scanning signal and the column data signal. If there are small synchronization errors in the rising edge, falling edge, or setup time of the scanning signal output by the row driving circuit and the data signal output by the column driving circuit, it will cause the gates of the thin-film transistors (TFTs) of some pixel units to receive data voltages before reaching the stable on state, thereby causing instability in the injection current of the LED diodes. For example, when playing a high-speed moving picture, such timing synchronization errors may cause data writing delays or advances in the pixel units, and finally manifest as display anomalies such as local brightness flickering or blurred motion trajectories in the picture. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides an LED display matrix dynamic control system to solve the problem of insufficient timing synchronization control accuracy between the row driving circuit and the column driving circuit in the LED display matrix dynamic control system.

[0005] To solve the above technical problems, the specific technical solution of the present invention is as follows: The LED display matrix dynamic control system provided by the present invention includes a timing feature extraction module, an adaptive threshold calibration module, a hierarchical compensation execution module, and a dynamic model update module; The timing feature extraction module collects the scanning signal output by the row driving circuit and the data signal output by the column driving circuit through a dual-channel independent sampling circuit. After extracting the basic timing parameters (such as the rising edge time and falling edge time of the scanning signal) and environment-related features (such as the jitter standard deviation of the rising edge of the scanning signal) based on the collected signal timing data, it packages them into structured data and transmits them to the adaptive threshold calibration module; The adaptive threshold calibration module receives the structured data, identifies the display scene (such as static, high-speed motion scene) based on the environment-related features (such as the signal-to-noise power ratio) and the change rate of the basic timing parameters (such as the fluctuation frequency of the setup time of the data signal) therein, and generates a composite reference value in combination with the historical optimization parameters (such as the mean value of the deviation distribution of the same type of scene in the past 20 frames) fed back by the dynamic model update module and transmits it to the hierarchical compensation execution module; The hierarchical compensation execution module executes a hierarchical compensation strategy (such as first-level, second-level, third-level compensation) according to the composite reference value, and outputs adjustment parameters including compensation adjustment amounts (such as digital delay line adjustment amounts, capacitor array switching states) to the dynamic model update module; The dynamic model update module optimizes the scene recognition classification rules and threshold generation calculation rules of the adaptive threshold calibration module based on the adjustment parameters and historical timing data, and feeds back the optimized parameters to the adaptive threshold calibration module to form a closed-loop control link for improving the timing synchronization control accuracy of the row driving circuit and the column driving circuit.

[0006] Further, in the LED display matrix dynamic control system of the present invention, the timing feature extraction module is respectively connected to the signal output ends of the row driving circuit and the column driving circuit through a dual-channel independent sampling circuit, and synchronously collects the time-domain waveforms of the scanning signal and the data signal at a sampling rate 10 times the highest frequency of the signal; The timing feature extraction module is internally provided with a feature extraction unit, which analyzes and processes the collected time-domain waveforms based on signal processing technology to extract basic timing parameters and environment-related features, where the basic timing parameters include the rising edge time and falling edge time of the scanning signal and the setup time and hold time of the data signal, and the environment-related features include the jitter standard deviation of the rising edge of the scanning signal, the temperature drift rate of the setup time of the data signal, and the signal-to-noise power ratio; After the extracted basic timing parameters and environment-related features are marked with time stamps (recording the time information of data acquisition), they are packaged into structured data in the structure of frame number, parameter type, value, and environment label, and transmitted to the adaptive threshold calibration module through the internal bus.

[0007] Further, in the LED display matrix dynamic control system of the present invention, after receiving the structured data, the adaptive threshold calibration module extracts the environmental correlation features (including the jitter standard deviation of the rising edge of the scan signal, the temperature drift rate of the data signal setup time, and the signal-to-noise power ratio) and the change rates of the basic timing parameters (including the fluctuation frequency of the rising edge time of the scan signal and the change slope of the data signal setup time) from the structured data, inputs the above features and rates into the k-nearest neighbor classifier built in the adaptive threshold calibration module, matches the scene library pre-stored including static, dynamic, high-speed motion, and temperature change scene labels, and outputs the current scene recognition result.

[0008] Further, in the LED display matrix dynamic control system of the present invention, based on the scene recognition result, the adaptive threshold calibration module calls the reference threshold matching the current scene (such as static and high-speed motion scenes) from the preset threshold library built in the adaptive threshold calibration module; meanwhile, obtains the historical optimization parameters fed back by the dynamic model update module (the historical optimization parameters include the mean and variance of the deviation distribution of the same type of scenes in the past 20 frames); the module performs weighted fusion of the reference threshold and the historical optimization parameters according to the weight coefficient determined by the scene recognition result (such as the weight coefficient of the high-speed motion scene is 0.7 and that of the temperature change scene is 0.5), generates a composite reference value including the allowable maximum deviation range (such as ±2 ns) and the predicted deviation direction (such as the scan signal lagging), and transmits it to the hierarchical compensation execution module.

[0009] Further, in the LED display matrix dynamic control system of the present invention, the hierarchical compensation execution module executes a primary compensation strategy based on the composite reference value: When the timing deviation between the scan signal and the data signal collected by the timing feature extraction module (i.e., the timing synchronization error between the scan signal and the data signal) ≤ 1 / 3 of the composite reference value, adjust the phase of the scan signal through the digital delay line of the row driving circuit, and the adjustment amount is 0.8 times the measured timing deviation, so as to correct the small synchronization error between the scan signal and the data signal.

[0010] Further, in the LED display matrix dynamic control system of the present invention, the hierarchical compensation execution module executes a secondary compensation strategy based on the composite reference value: When the timing deviation between the scan signal and the data signal collected by the timing feature extraction module is greater than 1 / 3 of the composite reference value and less than or equal to 2 / 3 of the composite reference value, synchronously adjust the phase of the scan signal of the digital delay line of the row driving circuit (the adjustment amount is 0.6 times the measured timing deviation) and the source output load capacitance of the column driving circuit (change the data signal slope by switching the parallel capacitor array, and the slope adjustment amount is 0.4 times the measured timing deviation), and jointly correct the medium synchronization error between the scan signal and the data signal through dual-channel adjustment.

[0011] Furthermore, for the LED display matrix dynamic control system of the present invention, the hierarchical compensation execution module executes a three-level compensation strategy based on the composite reference value: When the timing deviation between the scan signal and the data signal collected by the timing feature extraction module (i.e., the timing synchronization error between the scan signal and the data signal) is greater than 2 / 3 of the composite reference value, the clock synchronization unit built in the timing feature extraction module is triggered to adjust the control voltage of the voltage-controlled oscillator, and the clock sources of the row drive and the column drive are relocked to align the phases of the clock sources of the row drive and the column drive, so as to eliminate the serious synchronization error between the scan signal and the data signal at the clock source level.

[0012] Furthermore, for the LED display matrix dynamic control system of the present invention, the dynamic model update module receives the adjustment parameters including the compensation adjustment amount output by the hierarchical compensation execution module (including the digital delay line adjustment amount, the capacitor array switching state, and the voltage-controlled oscillator control voltage), and the historical data of the most recent 100 frames collected by the timing feature extraction module (including the structured data of the basic timing parameters and the environmental correlation features). Based on the above adjustment parameters and historical data, the classification weights of the k-nearest neighbor classifier and the weighting coefficients generated by the threshold in the adaptive threshold calibration module are encoded into chromosomes. The chromosomes include the parameter combinations of the classification weights and the weighting coefficients, which are used to optimize the scene recognition and threshold generation parameters of the adaptive threshold calibration module.

[0013] Furthermore, for the LED display matrix dynamic control system of the present invention, the dynamic model update module optimizes the chromosome parameters through a genetic algorithm, and uses "the proportion of the number of frames in which the timing deviation between the scan signal and the data signal after compensation by the hierarchical compensation execution module is ≤ the reference threshold (the reference threshold from the built-in preset threshold library of the module)" as the fitness function; The top 30% of the individuals with high fitness are selected as parents through roulette wheel selection, and offspring are generated through single-point crossover. The classification weights and weighting coefficients in the chromosomes are randomly perturbed with a probability of 5%, and the chromosome parameters are iteratively optimized to improve the accuracy of scene recognition and the adaptability of threshold generation.

[0014] Furthermore, for the LED display matrix dynamic control system of the present invention, after every 5 generations of genetic algorithm evolution are completed by the dynamic model update module, the optimal chromosome after evolution is extracted based on the evaluation result of the fitness function, and the corresponding k-nearest neighbor classifier classification weights and threshold generation weighting coefficients are obtained. The above weights and coefficients are written into the scene library of the adaptive threshold calibration module, and the classification weights and threshold generation strategies corresponding to the static, dynamic, high-speed motion, and temperature change scene labels in the scene library are updated, so that the adaptive threshold calibration module can more accurately match the synchronization control requirements of the current display scene.

[0015] Advantages of the present invention; The present invention obtains multi-dimensional timing parameters and environment-related features of the row driving scan signal and the column driving data signal through a timing feature extraction module, providing an accurate data basis for scene recognition and threshold generation; the adaptive threshold calibration module realizes the dynamic recognition of the display scene based on a k-nearest neighbor classifier, and combines historical optimization parameters to generate a composite reference value adapted to the current scene, solving the problem that a fixed threshold cannot meet the requirements of different display scenes; the hierarchical compensation execution module covers the full range of error correction from minute timing deviation to severe timing deviation through a three-level strategy (row driving phase adjustment, row-column dual-path collaborative adjustment, clock source phase locking), improving the flexibility and pertinence of control; the dynamic model update module uses a genetic algorithm to optimize the scene recognition classification weight and threshold generation coefficient, and periodically updates the scene library parameters, enabling the system to have the ability of self-learning and self-optimization. The collaborative work of the above modules realizes a closed-loop control from data acquisition, scene adaptation, error correction to parameter evolution, effectively improving the timing synchronization control accuracy of the row driving circuit and the column driving circuit, and enhancing the robustness of the system to complex display scenes such as static, dynamic, high-speed motion and temperature change. BRIEF DESCRIPTION OF THE DRAWINGS

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

[0017] Figure 1 FIG. is a system architecture diagram of the LED display matrix dynamic control system provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[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 the specific embodiments and corresponding drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are 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 LED display matrix dynamic control system provided by the present invention includes a timing feature extraction module, an adaptive threshold calibration module, a hierarchical compensation execution module, and a dynamic model update module; The timing feature extraction module collects the scan signal output by the row driving circuit and the data signal output by the column driving circuit through a dual-channel independent sampling circuit. After extracting the basic timing parameters (such as the rising edge time and falling edge time of the scan signal) and environment-related features (such as the jitter standard deviation of the rising edge of the scan signal) based on the collected signal timing data, it packages them into structured data and transmits them to the adaptive threshold calibration module; The adaptive threshold calibration module receives the structured data, identifies the display scenario (such as static, high-speed motion scenario) based on the environment-related features (such as signal-to-noise power ratio) and the change rate of basic timing parameters (such as the fluctuation frequency of the setup time of the data signal) therein, and generates a composite reference value in combination with the historical optimization parameters (such as the mean value of the deviation distribution of the same type of scenario in the past 20 frames) fed back by the dynamic model update module and transmits it to the hierarchical compensation execution module; The hierarchical compensation execution module executes a hierarchical compensation strategy (such as first-level, second-level, third-level compensation) according to the composite reference value, and outputs adjustment parameters including compensation adjustment amounts (such as digital delay line adjustment amounts, capacitor array switching states) to the dynamic model update module; The dynamic model update module optimizes the scenario recognition classification rules and threshold generation calculation rules of the adaptive threshold calibration module based on the adjustment parameters and historical timing data, and feeds back the optimized parameters to the adaptive threshold calibration module to form a closed-loop control link for improving the timing synchronization control accuracy of the row driving circuit and the column driving circuit.

[0020] The LED display matrix dynamic control system provided by the present invention realizes the improvement of the timing synchronization control accuracy of the row driving circuit and the column driving circuit through the collaborative work of four functional modules. The specific implementation and logical relationship of each module are as follows: As the data input link of the system, the timing feature extraction module is respectively connected to the signal output ends of the row driving circuit and the column driving circuit through a dual-channel independent sampling circuit, and synchronously collects the time-domain waveforms of the scan signal and the data signal at a sampling rate 10 times the highest frequency of the signal. This module has a built-in feature extraction unit, which analyzes and processes the collected time-domain waveforms based on signal processing technology, and extracts two types of key parameters: one is the basic timing parameters, including the rising edge time and falling edge time of the scan signal and the setup time and hold time of the data signal; the other is the environment-related features, including the jitter standard deviation of the rising edge of the scan signal (reflecting signal stability), the temperature drift rate of the setup time of the data signal (reflecting the influence of temperature on timing), and the signal-to-noise power ratio (reflecting signal quality). All the extracted parameters are marked with time stamps (recording the specific time information of data acquisition), and are packaged into structured data in the structure of frame number, parameter type, value, and environment label, and are transmitted to the adaptive threshold calibration module through the internal bus to provide a data basis for subsequent scenario recognition and threshold generation.

[0021] After receiving the structured data output by the timing feature extraction module, the adaptive threshold calibration module first extracts the environmental correlation features and the change rates of the basic timing parameters (such as the fluctuation frequency of the rising edge time of the scan signal and the change slope of the setup time of the data signal) from this data, and inputs the above features and rates into the built-in k-nearest neighbor classifier of the module. By matching the pre-stored scene library containing static, dynamic, high-speed motion, and temperature change scene labels, this classifier identifies the current display scene (for example, when the fluctuation frequency of the setup time of the data signal increases significantly, it is determined as a high-speed motion scene). Based on the scene recognition result, the module calls the reference threshold matching the current scene from the built-in preset threshold library (such as a more stringent reference threshold for the high-speed motion scene), simultaneously obtains the historical optimization parameters fed back by the dynamic model update module (including the mean and variance of the deviation distribution of the same type of scenes in the past 20 frames), and according to the weight coefficient determined by the scene recognition result (such as the weight coefficient of the high-speed motion scene being 0.7), performs weighted fusion of the reference threshold and the historical optimization parameters to generate a composite reference value containing the allowable maximum deviation range (such as ±2ns) and the predicted deviation direction (such as the scan signal lagging), and transmits it to the hierarchical compensation execution module for guiding the selection of subsequent compensation strategies.

[0022] According to the received composite reference value, the hierarchical compensation execution module executes a three-level compensation strategy for different degrees of timing deviation: when the timing deviation between the scan signal and the data signal (i.e., the timing synchronization error) ≤ 1 / 3 of the composite reference value, it executes the first-level compensation, adjusts the phase of the scan signal through the digital delay line of the row driving circuit, and the adjustment amount is 0.8 times the measured timing deviation to correct the small synchronization error; when the deviation is greater than 1 / 3 of the composite reference value and less than or equal to 2 / 3, it executes the second-level compensation, synchronously adjusts the phase of the scan signal of the digital delay line of the row driving circuit (the adjustment amount is 0.6 times the measured deviation) and the source output load capacitance of the column driving circuit (changes the data signal slope by switching the parallel capacitance array, and the slope adjustment amount is 0.4 times the measured deviation), and corrects the medium synchronization error through dual-channel collaborative adjustment; when the deviation is greater than 2 / 3 of the composite reference value, it executes the third-level compensation, triggers the built-in clock synchronization unit of the module to adjust the control voltage of the voltage-controlled oscillator, re-lock the clock sources of the row driving and column driving, and achieve the phase alignment of the clock sources to eliminate the serious synchronization error from the source. After the compensation is completed, the module records the adjustment parameters (including the adjustment amount of the digital delay line, the switching state of the capacitance array, and the control voltage of the voltage-controlled oscillator) and transmits them to the dynamic model update module for subsequent model optimization.

[0023] The dynamic model update module receives the adjustment parameters output by the hierarchical compensation execution module and the historical data of the most recent 100 frames collected by the timing feature extraction module (including the structured data of the basic timing parameters and the environment-related features), and encodes the classification weights of the k-nearest neighbor classifier and the weighting coefficients generated by the threshold in the adaptive threshold calibration module into chromosomes (parameter combinations) based on the above data. The module optimizes the chromosome parameters through a genetic algorithm: taking "the proportion of the number of frames where the timing deviation between the scan signal and the data signal compensated by the hierarchical compensation execution module is ≤ the reference threshold (from the preset threshold library)" as the fitness function, selecting the top 30% of the individuals with high fitness as parents through roulette wheel selection, generating offspring through single-point crossover, and randomly perturbing the classification weights and weighting coefficients in the chromosome with a 5% probability, and iteratively optimizing the parameters to improve the accuracy of scene recognition and the adaptability of threshold generation. After every 5 generations of evolution, the module extracts the optimal chromosome based on the evaluation results of the fitness function, obtains the corresponding classification weights and weighting coefficients, writes them into the scene library of the adaptive threshold calibration module, and updates the classification weights and threshold generation strategies corresponding to the static, dynamic, high-speed motion, and temperature change scene labels, so that the adaptive threshold calibration module can more accurately match the synchronous control requirements of the current display scene.

[0024] Specifically, in the LED display matrix dynamic control system of the present invention, the timing feature extraction module is respectively connected to the signal output ends of the row driving circuit and the column driving circuit through a dual-channel independent sampling circuit, and synchronously collects the time-domain waveforms of the scan signal and the data signal at a sampling rate 10 times the highest frequency of the signal; The timing feature extraction module is built with a feature extraction unit, which analyzes and processes the collected time-domain waveforms based on signal processing technology to extract basic timing parameters and environment-related features, where the basic timing parameters include the rising edge time and falling edge time of the scan signal and the setup time and hold time of the data signal, and the environment-related features include the jitter standard deviation of the rising edge of the scan signal, the temperature drift rate of the setup time of the data signal, and the signal noise power ratio; After the extracted basic timing parameters and environment-related features are marked with time stamps (recording the time information of data acquisition), they are packed into structured data in the structure of frame number, parameter type, value, and environment label, and transmitted to the adaptive threshold calibration module through the internal bus.

[0025] The timing feature extraction module is the signal acquisition and preprocessing core unit of the LED display matrix dynamic control system, and the realization of its function includes the following closely related technical steps: The module is physically connected to the signal output terminals of the row driving circuit and the column driving circuit respectively through a dual-channel independent sampling circuit. The sampling circuit consists of a differential amplifier and an analog-to-digital converter (ADC). The differential amplifier is used to suppress common-mode noise. The analog-to-digital converter synchronously samples the scanning signal and the data signal at a sampling rate 10 times the highest frequency of the signal. The collected time-domain waveform can completely reproduce key features such as the rising edge and falling edge of the original signal (According to the Nyquist sampling theorem, a 10-fold sampling rate can effectively avoid aliasing distortion).

[0026] The collected time-domain waveform (in the form of a digital signal) is input into the feature extraction unit built into the module. This unit preprocesses and extracts features from the waveform based on digital signal processing technology: First, high-frequency noise is filtered out through a low-pass filter. Then, the rising edge and falling edge moments of the scanning signal are identified using an edge detection algorithm (for example, setting the voltage threshold to 90% of the signal amplitude, and recording the rising edge time when the signal exceeds this threshold). At the same time, the time window from the signal reaching stability to maintaining stability (i.e., the setup time and hold time) of the data signal is calculated. The extraction of environment-related features is achieved through statistical analysis. For example, the standard deviation of the jitter of the rising edge of the scanning signal is obtained by calculating the variance of the rising edge times of 100 consecutive cycles. The temperature drift rate of the setup time of the data signal is determined by comparing the ratio of the change in the setup time to the change in temperature at different temperatures (such as 25°C, 50°C, -10°C). The signal-to-noise power ratio is obtained by calculating the ratio of the effective signal power to the noise power.

[0027] The extracted basic timing parameters (such as the rising edge time t_r and falling edge time t_f of the scanning signal, the setup time t_su and hold time t_ho of the data signal) and environment-related features (such as the standard deviation of jitter σ_r, the temperature drift rate d_t_su / dT, the signal-to-noise power ratio SNR) need to be timestamped. This timestamp is achieved through the clock counter built into the module. The counter is synchronized with the system main clock, so that each parameter corresponds to a unique timestamp (such as a time value accurate to the nanosecond level), which is used to align the processing times of different modules during subsequent analysis.

[0028] All the timestamped parameters are packed in a preset structured format: The data frame header includes a frame sequence number (used to distinguish the sampling data at different time points, incrementing from 0001), a parameter type field (01 represents basic timing parameters, 02 represents environment-related features), a numerical field (stores the specific parameter value, such as t_r = 15ns), and an environment label field (records the environmental conditions during sampling, such as temperature T = 30°C, humidity H = 50%). The packed data is transmitted to the adaptive threshold calibration module through the I2C bus inside the system, providing time-aligned and clearly classified input data for subsequent scenario recognition and threshold generation.

[0029] Specifically, in the LED display matrix dynamic control system of the present invention, after receiving the structured data, the adaptive threshold calibration module extracts the environment-related features (including the jitter standard deviation of the rising edge of the scan signal, the temperature drift rate of the data signal setup time, and the signal-to-noise power ratio) and the change rates of the basic timing parameters (including the fluctuation frequency of the rising edge time of the scan signal and the change slope of the data signal setup time) from the structured data, and inputs the above features and rates into the k-nearest neighbor classifier built in the adaptive threshold calibration module to match the pre-stored scene library including static, dynamic, high-speed motion, and temperature change scene labels, and outputs the current scene recognition result.

[0030] The adaptive threshold calibration module is the core unit for scene recognition and threshold generation in the LED display matrix dynamic control system, and the realization of its function includes the following closely related technical steps: The module receives the structured data transmitted by the timing feature extraction module through the I2C bus inside the system, and this data is stored in a fixed frame format (the frame header is a 4-byte frame sequence number, the parameter type field is 1 byte, the value field is an 8-byte floating-point number, and the environment label field is a 4-byte temperature / humidity code). The module first calls the built-in parsing subroutine to check the data integrity according to the frame sequence number of the frame header (such as no jump in consecutive frame sequence numbers), and then separates the basic timing parameters and the environment-related features according to the parameter type field (for example, parameter type 01 corresponds to the rising edge time of the scan signal, 02 corresponds to the data signal setup time, and 03 corresponds to the jitter standard deviation of the rising edge of the scan signal), and finally extracts the environment-related features (the jitter standard deviation of the rising edge of the scan signal, the temperature drift rate of the data signal setup time, the signal-to-noise power ratio) and the change rates of the basic timing parameters (the fluctuation frequency of the rising edge time of the scan signal, the change slope of the data signal setup time).

[0031] The extraction of the environment-related features is achieved through data statistics and analysis: the jitter standard deviation of the rising edge of the scan signal is calculated based on the rising edge time series of 100 consecutive frames (after taking the mean value, calculate the average value of the sum of the squared differences between each time point and the mean value); the temperature drift rate of the data signal setup time is calculated by comparing the setup time values under different temperature labels (such as 25°C, 50°C, -10°C) to calculate the change amount per degree Celsius (for example, when the temperature rises from 25°C to 50°C, the setup time increases from 10 ns to 12 ns, then the drift rate is 0.08 ns / °C); the signal-to-noise power ratio is obtained by calculating the ratio of the effective signal power (the average value of the squared signal amplitude) to the noise power (the power of the high-frequency component of the signal) (such as SNR = 15 dB means that the signal power is 31.6 times the noise power).

[0032] The change rate of the basic timing parameters is extracted through time series analysis: the fluctuation frequency of the rising edge time of the scan signal is obtained by performing a fast Fourier transform (FFT) on the rising edge times of 50 consecutive frames to identify the main frequency components (for example, a fluctuation frequency of 120 Hz means that the fluctuation of the rising edge time repeats every 8.3 ms); the change slope of the setup time of the data signal is calculated by linearly regressing the relationship between the setup times of 20 consecutive frames and the frame numbers to obtain the slope value (for example, a slope of 0.5 ns / frame means that the setup time increases by 0.5 ns for each additional frame).

[0033] The extracted environmental correlation features and the change rate of the basic timing parameters are integrated into a 5-dimensional input vector ([σ_r, d_t_su / dT, SNR, f_r, k_su], where σ_r is the jitter standard deviation, d_t_su / dT is the temperature drift rate, SNR is the noise power ratio, f_r is the fluctuation frequency, and k_su is the change slope), and input to the k-nearest neighbor (k-NN) classifier built into the module. The classifier is pre-configured with k = 5 (to avoid ties) and uses the Euclidean distance to measure the similarity between the input vector and the samples in the scene library (the distance calculation formula is the square root of the sum of the squares of the differences in each dimension). The scene library is stored in the non-volatile memory of the module (such as EEPROM) and contains typical samples of 4 types of scene labels: static scene (f_r ≤ 50 Hz, k_su ≤ 0.2 ns / frame), dynamic scene (50 Hz < f_r ≤ 100 Hz, 0.2 ns / frame < k_su ≤ 0.5 ns / frame), high-speed motion scene (f_r > 100 Hz, k_su > 0.5 ns / frame), and temperature change scene (d_t_su / dT ≥ 0.1 ns / °C or ≤ -0.1 ns / °C).

[0034] The classifier calculates the Euclidean distances between the input vector and all samples in the scene library, selects the 5 nearest samples, counts the occurrences of each scene label (for example, 3 samples are of the high-speed motion scene and 2 are of the dynamic scene), and finally outputs the current scene recognition result (such as the high-speed motion scene label) through the majority voting rule.

[0035] Specifically, for the LED display matrix dynamic control system of the present invention, the adaptive threshold calibration module, based on the scene recognition result, calls a reference threshold that matches the current scene (such as a static or high-speed motion scene) from the preset threshold library built into the adaptive threshold calibration module; meanwhile, it obtains the historical optimization parameters fed back by the dynamic model update module (the historical optimization parameters include the mean and variance of the deviation distribution of the same type of scene in the past 20 frames); the module weights and fuses the reference threshold and the historical optimization parameters according to the weight coefficient determined by the scene recognition result (such as the weight coefficient for the high-speed motion scene is 0.7, and for the temperature change scene is 0.5), generates a composite reference value including the allowable maximum deviation range (such as ±2 ns) and the predicted deviation direction (such as the scanning signal lagging), and transmits it to the hierarchical compensation execution module.

[0036] After completing the scene recognition, the adaptive threshold calibration module generates a composite reference value through the following technical steps to provide a control reference adapted to the current scene for the hierarchical compensation execution module: The module first calls the preset threshold library from the built-in non-volatile memory (such as EEPROM). The threshold library is stored classified by scene tags, and each scene tag corresponds to a reference threshold. For example, the reference threshold for the static scene is ±3 ns (allowing a timing deviation of ±3 ns between the scanning signal and the data signal), the reference threshold for the high-speed motion scene is ±1.5 ns (since high-speed motion is more sensitive to timing errors and requires a stricter threshold), and the reference threshold for the temperature change scene is ±2.5 ns (considering the impact of temperature fluctuations on timing). The calling process is implemented through the scene tag as an index. For example, if the scene recognition result is "high-speed motion scene", the module locates the reference threshold ±1.5 ns corresponding to this tag through the lookup table.

[0037] Meanwhile, the module receives the historical optimization parameters fed back by the dynamic model update module through the SPI bus inside the system. These parameters are stored in the RAM register of the module and include the mean (μ) and variance (σ²) of the timing deviation distribution of the same type of scene in the past 20 frames. For example, if the current scene is "high-speed motion scene", the historical optimization parameters are the mean μ = 0.8 ns (indicating that the scanning signal lags the data signal by an average of 0.8 ns) and the variance σ² = 0.25 ns² (indicating the degree of dispersion of the deviation) of the timing deviation in the past 20 high-speed motion scenes.

[0038] The module determines the weight coefficient according to the scene recognition result, which reflects the dependence of the current scene on the reference threshold and historical optimization parameters. The weight coefficient is determined by the mapping relationship between the scene label and the preset weight table. For example, the weight coefficient for the static scene is 0.4 (more dependent on historical data), 0.6 for the dynamic scene, 0.7 for the high-speed motion scene (more dependent on the reference threshold to ensure control accuracy), and 0.5 for the temperature change scene (balancing the reference and historical data). The setting of the weight coefficient is based on the sensitivity of the scene to the timing error. The higher the sensitivity (such as high-speed motion), the higher the weight of the reference threshold.

[0039] Based on the above parameters, the module performs weighted fusion calculation: First, linearly weight the reference threshold (such as ±1.5 ns for the high-speed motion scene) and the mean of the historical optimization parameters (such as μ = 0.8 ns). The formula is the composite reference value = reference threshold × weight coefficient + historical mean × (1 - weight coefficient). For example, in the high-speed motion scene, the allowable maximum deviation range of the composite reference value is ±(1.5×0.7 + 0.8×0.3) = ±(1.05 + 0.24) = ±1.29 ns (rounded to ±1.3 ns). The prediction deviation direction is determined by the sign of the historical mean. If μ is positive (such as 0.8 ns), the predicted scan signal lags the data signal; if μ is negative (such as -0.5 ns), the predicted scan signal leads the data signal.

[0040] The finally generated composite reference value (including the allowable maximum deviation range of ±1.3 ns and the prediction deviation direction "scan signal lags") is transmitted to the hierarchical compensation execution module through the internal bus, providing a scene-adapted control reference for the selection of the subsequent three-level compensation strategy.

[0041] Specifically, in the LED display matrix dynamic control system of the present invention, the hierarchical compensation execution module executes the first-level compensation strategy based on the composite reference value: When the timing deviation between the scan signal and the data signal collected by the timing feature extraction module (i.e., the timing synchronization error between the scan signal and the data signal) ≤ 1 / 3 of the composite reference value, adjust the phase of the scan signal through the digital delay line of the row driving circuit, and the adjustment amount is 0.8 times the measured timing deviation to correct the small synchronization error between the scan signal and the data signal.

[0042] The first-level compensation strategy of the hierarchical compensation execution module is a refined correction mechanism for small timing deviations, and its technical implementation includes the following closely related technical steps: The module first receives the composite reference value (including the allowable maximum deviation range and the predicted deviation direction) transmitted by the adaptive threshold calibration module through the internal bus of the system, and at the same time obtains the timing deviation value between the scan signal and the data signal of the current frame from the timing feature extraction module. This deviation value is calculated by comparing the rising edge time of the scan signal with the setup time of the data signal (for example, if the rising edge time of the scan signal is t_r = 15 ns and the setup time of the data signal is t_su = 13.5 ns, then the timing deviation is t_r - t_su = 1.5 ns).

[0043] When it is detected that the timing deviation ≤ 1 / 3 of the composite reference value (for example, if the allowable maximum deviation of the composite reference value is ±3 ns, then the trigger condition is that the deviation ≤ ±1 ns), the module triggers the first-level compensation strategy. At this time, the module calls the digital delay line (DLL) component built into the row driving circuit. This component consists of multiple cascadable delay units (each unit provides a fixed delay amount, such as 0.1 ns / unit), a phase detector (PD), and a charge pump (CP), and precisely adjusts the phase of the scan signal by adjusting the number of delay units.

[0044] The adjustment amount is calculated based on 0.8 times the measured timing deviation (for example, if the measured deviation is +0.8 ns, then the adjustment amount is +0.8 ns × 0.8 = +0.64 ns). The proportional coefficient of 0.8 is selected to avoid overcompensation (retaining a 20% adjustment margin) and prevent the adjusted phase from exceeding the target range due to signal jitter. The module realizes phase adjustment by controlling the control register of the DLL (such as writing an instruction for the number of delay units): if the scan signal lags (the deviation is positive), then increase the number of delay units to delay the phase of the scan signal; if the scan signal leads (the deviation is negative), then reduce the number of delay units to advance the phase of the scan signal (for example, when the adjustment amount is +0.64 ns, 6 delay units need to be added, and each unit provides 0.1 ns of delay).

[0045] After the adjustment is completed, the module verifies the compensation effect through the timing deviation between the corrected scan signal and the data signal fed back by the timing feature extraction module (for example, the corrected deviation drops from +0.8 ns to +0.1 ns), and confirms that the tiny synchronization error has been corrected. This process is only achieved through the single-path adjustment of the DLL in the row driving circuit, avoiding interference with the column driving circuit, improving the simplicity and fast responsiveness of the control strategy, and meeting the requirements for correcting tiny timing errors in the LED display matrix in static or low-speed motion scenarios.

[0046] Specifically, for the LED display matrix dynamic control system described in the present invention, the hierarchical compensation execution module executes a second-level compensation strategy based on the composite reference value: When the timing deviation between the scan signal and the data signal collected by the timing feature extraction module is greater than 1 / 3 of the composite reference value and less than or equal to 2 / 3 of the composite reference value, synchronously adjust the phase of the scan signal of the row drive circuit digital delay line (the adjustment amount is 0.6 times the measured timing deviation) and the source output load capacitance of the column drive circuit (change the data signal slope by switching the parallel capacitor array, and the slope adjustment amount is 0.4 times the measured timing deviation), and jointly correct the medium synchronization error between the scan signal and the data signal through dual-channel adjustment.

[0047] The secondary compensation strategy of the hierarchical compensation execution module is a collaborative correction mechanism for medium timing deviation, and its technical implementation includes the following closely related technical steps: The module obtains the composite reference value (such as the allowable maximum deviation range of ±3 ns) transmitted by the adaptive threshold calibration module in real time through the system internal bus, and reads the timing deviation values of the scan signal and the data signal of the current frame from the timing feature extraction module (for example, the rising edge time of the scan signal is t_r = 16 ns, the setup time of the data signal is t_su = 13.5 ns, and the timing deviation is t_r - t_su = 2.5 ns). When it is detected that the deviation value is greater than 1 / 3 of the composite reference value (i.e., >±1 ns) and less than or equal to 2 / 3 (i.e., ≤±2 ns), the module triggers the secondary compensation strategy, and realizes error correction through the dual-channel collaborative adjustment of the row drive circuit and the column drive circuit.

[0048] The adjustment of the row drive circuit is completed by the digital delay line (DLL). The DLL consists of 16 cascaded delay units (each level provides a fixed delay of 0.1 ns), a phase comparator, and control logic. The adjustment amount is 0.6 times the measured timing deviation (for example, when the measured deviation is +2.5 ns, the row adjustment amount is +2.5 ns×0.6 = +1.5 ns). The module writes an adjustment instruction to the control register of the DLL to increase or decrease the number of delay units: if the scan signal lags (the deviation is positive), then add 15 delay units (15×0.1 ns = 1.5 ns) to delay the phase of the scan signal; if the scan signal leads (the deviation is negative), then reduce the corresponding number of delay units to advance the phase of the scan signal.

[0049] The adjustment of the column driver circuit is achieved by switching the source output load capacitor. The output stage of the column driver circuit integrates a parallel capacitor array, which includes 4 capacitors of different capacitance values (C1=1pF, C2=2pF, C3=4pF, C4=8pF), each of which is connected to the output terminal through a MOSFET switch. The data signal slope (dV / dt) is inversely proportional to the load capacitance (slope = current / capacitance), so switching the capacitor can change the rise / fall rate of the data signal. The adjustment amount is 0.4 times the measured timing deviation (for example, when the measured deviation is +2.5ns, the column adjustment amount is +2.5ns×0.4=+1.0ns), and the corresponding capacitance value that needs to be increased is derived through the slope calculation formula (for example, if the original slope is 5V / ns, the target slope needs to be reduced to extend the data signal setup time, and the switch C2=2pF is selected to make the slope become 2.5V / ns). The module switches the capacitor array and adjusts the slope of the data signal by controlling the on / off state of the MOSFET switch (such as closing the switch of C2).

[0050] The synchronization of the dual-channel adjustment is guaranteed by the built-in synchronous clock signal of the module: the adjustment instruction of the row driver DLL and the switching instruction of the column driver capacitor array are issued in the same clock cycle (such as the rising edge of the system master clock), so that the time of the scanning signal phase adjustment and the data signal slope adjustment is aligned. After the adjustment is completed, the module verifies the compensation effect through the corrected timing deviation feedback from the timing feature extraction module (for example, the corrected deviation is reduced from +2.5ns to +0.3ns), confirming that the medium synchronization error has been corrected collaboratively. This process takes into account the adjustment accuracy and response speed through the dual-channel adjustment of the row and column drive circuits, and is suitable for the medium timing error correction needs of the LED display matrix in dynamic or medium-speed motion scenes.

[0051] Specifically, in the LED display matrix dynamic control system of the present invention, the hierarchical compensation execution module executes a three-level compensation strategy based on the composite reference value: When the timing deviation of the scanning signal and the data signal collected by the timing feature extraction module (i.e., the timing synchronization error between the scanning signal and the data signal) is greater than 2 / 3 of the composite reference value, the built-in clock synchronization unit of the timing feature extraction module is triggered to adjust the control voltage of the voltage-controlled oscillator, re-lock the clock sources of the row drive and the column drive, and align the phases of the row drive and the column drive clock sources, thereby eliminating the serious synchronization errors between the scanning signal and the data signal from the clock source level.

[0052] The three-level compensation strategy of the hierarchical compensation execution module is a root cause correction mechanism for severe timing deviations. Its technical implementation includes the following closely related technical steps: The module obtains in real time the composite reference value (such as the allowable maximum deviation range of ±3 ns) transmitted by the adaptive threshold calibration module through the internal bus of the system, and reads the timing deviation value between the scan signal and the data signal of the current frame from the timing feature extraction module (for example, the rising edge time of the scan signal is t_r = 18 ns, the setup time of the data signal is t_su = 13.5 ns, and the timing deviation is t_r - t_su = 4.5 ns). When it is detected that the deviation value is greater than 2 / 3 of the composite reference value (i.e., >±2 ns), the module triggers a three-level compensation strategy, and adjusts the clock sources of the row drive and the column drive through the clock synchronization unit built in the timing feature extraction module to eliminate serious synchronization errors at the source.

[0053] The clock synchronization unit consists of a phase frequency detector (PFD), a charge pump (CP), a loop filter (LF), and a voltage controlled oscillator (VCO), forming a phase locked loop (PLL) structure. The PFD monitors in real time the phase difference between the row drive clock source (CLK_row) and the column drive clock source (CLK_col), and outputs an error signal (Up / Down pulse) proportional to the phase difference. This error signal is input to the charge pump, converted into a current signal, and then filtered by the loop filter (to filter out high-frequency noise) to generate the control voltage (Vctrl) of the VCO.

[0054] The output frequency of the VCO has a linear relationship with the control voltage (for example, for every 0.1 V increase in Vctrl, the output frequency increases by 10 MHz). The module adjusts the phase of the row drive and column drive clock sources by adjusting Vctrl to change the output frequency of the VCO: if the row clock source lags behind the column clock source (the phase difference is positive), the PFD outputs an Up pulse, the charge pump increases Vctrl, and the VCO frequency is increased to make the phase of the row clock source catch up; if the row clock source is ahead (the phase difference is negative), a Down pulse is output, Vctrl is decreased, and the VCO frequency is decreased to make the phase of the row clock source lag. This process continues until the PFD detects that the phase difference between the row and column clock sources is less than one clock cycle (such as 1 ns), completing the phase locking.

[0055] After the phase locking is completed, the clock sources (CLK_row, CLK_col) of the row drive and the column drive are generated based on the same VCO output frequency, making their phases strictly aligned (for example, the rising edge time difference between CLK_row and CLK_col is reduced from 4.5 ns to 0.2 ns). Since the timing generation of the scan signal and the data signal directly depends on their respective clock sources (the scan signal is generated by dividing the frequency of the row clock source, and the data signal is generated by dividing the frequency of the column clock source), the phase alignment of the clock sources fundamentally reduces the timing deviation between the scan signal and the data signal.

[0056] After the adjustment is completed, the module verifies the compensation effect through the corrected timing deviation feedback by the timing feature extraction module (for example, the corrected deviation drops from 4.5 ns to 0.3 ns), and confirms that the serious synchronization error has been eliminated. This process solves the long-term and cumulative timing deviation of the row and column drive circuits caused by independent clock sources through the global adjustment of the clock source by the phase-locked loop, and is applicable to the correction requirements of serious timing errors in the LED display matrix under high-speed movement or temperature change scenarios.

[0057] Specifically, for the LED display matrix dynamic control system of the present invention, the dynamic model update module receives the adjustment parameters including compensation adjustment amounts (including digital delay line adjustment amounts, capacitor array switching states, voltage-controlled oscillator control voltages) output by the hierarchical compensation execution module, and the most recent 100-frame historical data (including structured data of basic timing parameters and environment-related features) collected by the timing feature extraction module. Based on the above adjustment parameters and historical data, the classification weights of the k-nearest neighbor classifier and the weighting coefficients generated by the threshold in the adaptive threshold calibration module are encoded as chromosomes. The chromosomes include parameter combinations of classification weights and weighting coefficients, and are used to optimize the scene recognition and threshold generation parameters of the adaptive threshold calibration module.

[0058] The dynamic model update module is the self-optimizing core unit of the LED display matrix dynamic control system, and the implementation of its function includes the following closely related technical steps: The module obtains input information through the dual-bus architecture inside the system (one path receives data from the hierarchical compensation execution module, and the other path receives data from the timing feature extraction module). The adjustment parameters output by the hierarchical compensation execution module are transmitted in a fixed data frame format (the frame header contains a compensation strategy identifier, such as 01 indicating first-level compensation, 02 indicating second-level compensation, and the parameter field stores the digital delay line adjustment amount <such as +15 levels>, the capacitor array switching state <such as the binary code 0010 indicating closing the C2 capacitor>, the voltage-controlled oscillator control voltage <such as 1.2 V>); the most recent 100-frame historical data transmitted by the timing feature extraction module is a structured data packet (each frame contains a frame sequence number, basic timing parameters <such as the rising edge time of the scan signal t_r = 15 ns>, and environment-related features <such as the jitter standard deviation σ_r = 0.3 ns>). The module first verifies the data integrity through cyclic redundancy check (CRC) (such as calculating whether the check code of the data frame is consistent with the received code), and then stores the adjustment parameters in the "compensation parameter table" of the RAM (indexed by compensation strategy), and stores the historical data in the "historical database" of the FLASH (arranged in reverse order of the frame sequence number time).

[0059] Based on the above - stored adjustment parameters and historical data, the module extracts the optimization target parameters related to the adaptive threshold calibration module: the classification weights (w1, w2, w3, w4, w5) of the k - nearest neighbor (k - NN) classifier and the weighting coefficients (α, β) for threshold generation. Among them, the classification weights correspond to the importance of each dimension of the environmental correlation features and the change rate of the basic timing parameters during k - NN classification (for example, w1 is the weight of the standard deviation of the rising - edge jitter of the scanning signal, w2 is the weight of the temperature drift rate of the data signal setup time, and the value range is 0 - 1); the weighting coefficients correspond to the fusion ratio of the reference threshold and the historical optimization parameters (α is the weight of the reference threshold, β is the weight of the historical mean, and α+β = 1).

[0060] The module encodes the above parameters into chromosomes (parameter combinations), which are represented in the form of floating - point vectors (such as [w1 = 0.4, w2 = 0.3, w3 = 0.2, w4 = 0.1, w5 = 0.0, α = 0.7, β = 0.3]), and each parameter retains 3 decimal places to ensure accuracy. The encoding process is implemented through the built - in parameter mapping subroutine of the module: First, linearly map the value range (0 - 1) of the classification weights and the weighting coefficients to 16 - bit unsigned integers (0 - 65535), then convert the integers into binary strings, and finally splice them into chromosomes of a fixed length (such as 112 bits, corresponding to 7 parameters×16 bits).

[0061] The encoded chromosomes are used for subsequent genetic - algorithm optimization. Its core goal is to improve the scene recognition accuracy (such as reducing the probability of misjudging a high - speed motion scene as a static scene) and the threshold - generation adaptability (such as making the composite reference value more conform to the actual error distribution of the current scene) of the adaptive threshold calibration module by adjusting the classification weights and the weighting coefficients. This process combines the historical compensation effect of the module (reflected by the adjusted parameters) with the signal features collected in real - time (reflected by the historical data), provides a data - driven input basis for the parameter optimization of the adaptive threshold calibration module, and finally realizes the dynamic evolution of the system control strategy.

[0062] Specifically, for the LED display matrix dynamic control system described in the present invention, the dynamic model update module optimizes the chromosome parameters through a genetic algorithm, with the "proportion of frames where the timing deviation between the scanning signal and the data signal after compensation by the hierarchical compensation execution module ≤ the reference threshold (the reference threshold from the built - in preset threshold library of the module)" as the fitness function; Select the top 30% of the individuals with high fitness as parents through roulette - wheel selection, generate offspring through single - point crossover, and randomly perturb the classification weights and the weighting coefficients in the chromosome with a 5% probability, and iteratively optimize the chromosome parameters to improve the accuracy of scene recognition and the adaptability of threshold generation.

[0063] The genetic algorithm optimization process of the dynamic model update module is the core technical link for the system to achieve self-optimization, and its function implementation includes the following closely related technical steps: The module first initializes the population parameters of the genetic algorithm: the population size is set to 50 chromosomes (to avoid insufficient diversity due to too small a population). The initial population is composed of historical optimization parameters (such as the optimal chromosome of the previous evolution) and randomly generated chromosomes (the classification weights and weighting coefficients are randomly selected within the range of 0-1) in a ratio of 7:3. Each chromosome corresponds to a set of parameter combinations to be optimized (such as [w1 = 0.4, w2 = 0.3, w3 = 0.2, w4 = 0.1, w5 = 0.0, α = 0.7, β = 0.3]), where w1-w5 are the classification weights of the k-NN classifier, and α and β are the weighting coefficients for threshold generation.

[0064] The calculation of the fitness function is based on the historical compensation effect of the hierarchical compensation execution module: the module extracts the time series deviation data after compensation for the last 1000 frames from the historical database (each frame contains the time series deviation values of the compensated scan signal and data signal, and the reference threshold), counts the number of frames where "the deviation after compensation ≤ the reference threshold", and calculates the proportion (such as 850 frames meet the conditions, accounting for 85%). This proportion is used as the fitness value of the current chromosome, directly reflecting the scene recognition and threshold generation performance of the adaptive threshold calibration module under this parameter combination (the higher the proportion, the better the performance).

[0065] The roulette wheel selection method is used for the selection operation: first, calculate the sum of the fitness values of all chromosomes in the population (such as the total is 4250%), and then assign a selection probability to each chromosome (probability = individual fitness value / total, for example, if a chromosome has a fitness of 85%, the probability is 85 / 4250 = 2%). The module generates 50 random numbers (corresponding to 50 parent selection positions), and selects the top 30% (15) high-fitness individuals as parents (such as the chromosomes with the top 15 fitness values), and retains the current optimal parameter combination.

[0066] The single-point crossover method is used for the crossover operation: randomly select pairs of two from the 15 parents (such as parent A and parent B), set a crossover point at the 40th position of the chromosome (the middle position of the binary encoding), and exchange the gene segments after the crossover point (such as the first 40 bits of parent A and the last 72 bits of parent B are combined to generate offspring C). The crossover operation generates 15 offspring chromosomes, retains the advantageous features of the parents and introduces new parameter combinations, increasing the population diversity.

[0067] The mutation operation perturbs the classification weights and weighting coefficients in the chromosome with a probability of 5%. For each parameter (such as w1) of each chromosome, a random number between 0 and 1 is generated. If the random number < 5%, a random perturbation of ±0.05 is applied to the parameter (e.g., w1 = 0.4 becomes 0.45 or 0.35). The perturbation range is restricted within 0 - 1 (e.g., only positive perturbation is allowed when w1 = 0.0) to avoid the parameters exceeding the valid value range and to reflect the physical meaning of the classification weights and weighting coefficients (e.g., the weight cannot be negative).

[0068] The iterative optimization process continues until the termination condition is met (e.g., the optimal fitness does not improve for 5 consecutive generations, or the maximum number of iterations reaches 100). After each generation of evolution is completed, the module updates the population (a total of 30 chromosomes for the parent and offspring generations, and 20 individuals with low fitness are eliminated), and records the current optimal chromosome (e.g., the chromosome with a fitness of 92%). Finally, the module outputs the classification weights and weighting coefficients corresponding to the optimal chromosome, which are used to update the scene recognition and threshold generation parameters of the adaptive threshold calibration module, improving its adaptability to different display scenarios (e.g., the recognition accuracy for high-speed motion scenarios is increased from 80% to 95%).

[0069] Specifically, in the LED display matrix dynamic control system described in the present invention, after every 5 generations of genetic algorithm evolution are completed by the dynamic model update module, the optimal chromosome after evolution is extracted based on the evaluation result of the fitness function, and the corresponding classification weights of the k-nearest neighbor classifier and the threshold generation weighting coefficients are obtained. The above weights and coefficients are written into the scene library of the adaptive threshold calibration module, updating the classification weights and threshold generation strategies corresponding to the static, dynamic, high-speed motion, and temperature change scene labels in the scene library, enabling the adaptive threshold calibration module to more accurately match the synchronous control requirements of the current display scene.

[0070] The parameter update process of the dynamic model update module is a key technical link for the system to achieve adaptive evolution, and its functional implementation includes the following closely related technical steps: The module has an in-built evolution generation counter with an initial value of 0. After each generation of genetic algorithm evolution is completed (i.e., after selection, crossover, and mutation operations are completed and a new population is generated), the counter is incremented by 1. When the counter value reaches 5 (i.e., 5 generations of evolution are completed), the module triggers the parameter update process to achieve a balance between sufficient optimization (to avoid premature update) and timely feedback (to avoid excessive delay).

[0071] The module extracts the evolved optimal chromosome based on the evaluation result of the fitness function: traverse the 50 chromosomes of the current population, compare their fitness values one by one (i.e., "the proportion of frames with compensated deviation ≤ the reference threshold"), and select the chromosome with the highest fitness value as the optimal chromosome (for example, the fitness of the chromosome with the highest fitness in the 3rd generation of evolution is 92%). The binary encoding of this chromosome (such as 112 bits) is converted into actual parameter values through the built-in decoding subroutine of the module: convert every 16-bit binary number into an unsigned integer (0 - 65535), and then restore it to a floating-point number in the range of 0 - 1 through the linear mapping formula (actual value = integer / 65535), and finally obtain the classification weights (w1 - w5) of the k-nearest neighbor classifier and the weighting coefficients (α, β) generated by the threshold.

[0072] The module writes the decoded parameters into the scenario library of the adaptive threshold calibration module. The scenario library is stored in the non-volatile memory (such as EEPROM) of the adaptive threshold calibration module, and is stored in a table structure. Each row corresponds to a scenario label (static, dynamic, high-speed movement, temperature change), and contains a classification weight field (w1 - w5) and a threshold generation coefficient field (α, β). For example, the storage row format of the high-speed movement scenario is: [scenario label: high-speed movement, w1: 0.5, w2: 0.3, w3: 0.1, w4: 0.05, w5: 0.05, α: 0.7, β: 0.3]. The module sends a write instruction to the EEPROM through the I2C bus to overwrite and write the parameters corresponding to the optimal chromosome to the storage location of the current scenario label (such as located through the address offset of the scenario label, the address of the static scenario is 0x00, and the high-speed movement scenario is 0x0C).

[0073] After the parameter update is completed, the module triggers the parameter verification process of the adaptive threshold calibration module: re-run the scenario recognition and threshold generation process using the most recent 100 frames of historical data, and statistically calculate the scenario recognition accuracy (such as the correct recognition rate of the high-speed movement scenario increases from 85% before the update to 93%) and the threshold generation adaptability (such as the matching degree between the composite reference value and the actual error distribution increases from 78% to 89%). If the verification passes (such as accuracy ≥ 90%, matching degree ≥ 85%), it is confirmed that the parameter update is effective; if not, roll back to the parameters before the update and record an exception log (such as "The parameter verification of the 5th generation of evolution fails, and the historical optimal parameters are retained").

[0074] This process enables the scenario recognition and threshold generation strategies of the adaptive threshold calibration module to dynamically adapt to the actual operating environment of the LED display matrix (such as temperature changes, display content switching) through periodic parameter updates and verifications, and ultimately improves the synchronous control accuracy of the system for different display scenarios (such as the timing deviation in the high-speed movement scenario decreases from ±2.5ns to ±1.2ns).

[0075] The present invention constructs a closed-loop control link including timing feature extraction, scene adaptation threshold generation, hierarchical compensation execution, and self-optimization model update, and the system solves the problem of insufficient timing synchronization control accuracy between the row driving circuit and the column driving circuit. First, the timing feature extraction module synchronously collects the time-domain waveforms of the row driving scan signal and the column driving data signal through a dual-channel independent sampling circuit, and uses signal processing technology to extract basic timing parameters such as the rising edge time, falling edge time of the scan signal, and the setup time, hold time of the data signal. At the same time, it extracts environmental correlation features such as the standard deviation of the rising edge jitter of the scan signal and the temperature drift rate of the data signal setup time, and after timestamp marking, it is packaged into structured data. This data provides multi-dimensional information reflecting the signal timing characteristics and environmental impacts for subsequent analysis, and is the basis for achieving precise control.

[0076] Secondly, based on the environmental correlation features and the change rate of the basic timing parameters in the structured data, the adaptive threshold calibration module matches the preset scene library (including static, dynamic, high-speed motion, and temperature change scene labels) through the built-in k-nearest neighbor classifier to identify the current display scene; combined with the historical optimization parameters (such as the mean and variance of the deviation distribution of the same type of scene in the past 20 frames) fed back by the dynamic model update module, it calls the reference threshold for scene matching and generates a composite reference value through weighted fusion. The composite reference value includes the allowable maximum deviation range and the predicted deviation direction, providing a control benchmark adapted to the current scene for the hierarchical compensation strategy, and avoiding the defect that a fixed threshold cannot meet the requirements of different display scenes.

[0077] Finally, the hierarchical compensation execution module executes a three-level compensation strategy according to the composite reference value: when the deviation is small, it adjusts the phase of the scan signal through the row driving digital delay line; when the deviation is medium, it synchronously adjusts the row driving phase and the column driving load capacitance; when the deviation is severe, it triggers the phase-locked loop to adjust the phases of the row and column driving clock sources. At the same time, the dynamic model update module uses historical compensation data and feature data to optimize the classification weights and threshold generation coefficients of the adaptive threshold calibration module through genetic algorithms, and periodically updates the parameter strategies in the scene library, enabling the system to have the ability of self-learning and self-optimization. Through the collaborative work of the above modules, the system can dynamically adapt to different display scenes, forming a closed-loop optimization from signal acquisition, threshold generation to error correction, effectively improving the timing synchronization control accuracy between the row driving circuit and the column driving circuit.

Claims

1. The dynamic control system of the LED display matrix, characterized in that: It includes a timing feature extraction module, an adaptive threshold calibration module, a hierarchical compensation execution module, and a dynamic model update module; The timing feature extraction module collects the scanning signal output by the row driving circuit and the data signal output by the column driving circuit through a dual-channel independent sampling circuit. After extracting the basic timing parameters and environment-related features based on the collected signal timing data, it packages them into structured data and transmits them to the adaptive threshold calibration module; The adaptive threshold calibration module receives the structured data, identifies the display scenario based on the environment-related features and the change rate of the basic timing parameters therein, and generates a composite reference value in combination with the historical optimization parameters fed back by the dynamic model update module and transmits it to the hierarchical compensation execution module; The hierarchical compensation execution module executes a hierarchical compensation strategy according to the composite reference value, and outputs adjustment parameters including compensation adjustment amounts to the dynamic model update module; The dynamic model update module optimizes the scenario recognition classification rules and threshold generation calculation rules of the adaptive threshold calibration module based on the adjustment parameters and historical timing data, and feeds back the optimized parameters to the adaptive threshold calibration module to improve the timing synchronization control accuracy of the row driving circuit and the column driving circuit.

2. The LED display matrix dynamic control system according to claim 1, wherein The timing feature extraction module is respectively connected to the signal output terminals of the row driving circuit and the column driving circuit through a dual-channel independent sampling circuit, and synchronously collects the time-domain waveforms of the scanning signal and the data signal at a sampling rate 10 times the highest frequency of the signal; The timing feature extraction module has a built-in feature extraction unit, which analyzes and processes the collected time-domain waveforms based on signal processing technology, and extracts basic timing parameters and environment-related features. The basic timing parameters include the rising edge time, falling edge time of the scanning signal, and the setup time and hold time of the data signal, and the environment-related features include the jitter standard deviation of the rising edge of the scanning signal, the temperature drift rate of the setup time of the data signal, and the signal noise power ratio; The extracted basic timing parameters and environment-related features are marked with time stamps, and are packaged into structured data according to the structure of frame number, parameter type, value, and environment label, and are transmitted to the adaptive threshold calibration module through the internal bus.

3. The LED display matrix dynamic control system according to claim 2, characterized in that, After receiving the structured data, the adaptive threshold calibration module extracts the change rates of the environment-related features and the basic timing parameters from the structured data, inputs the change rates of the environment-related features and the basic timing parameters into the k-nearest neighbor classifier built in the adaptive threshold calibration module, matches the pre-stored scenario library including static, dynamic, high-speed motion, and temperature change scenario labels, and outputs the current scenario recognition result.

4. The LED display matrix dynamic control system according to claim 3, wherein Based on the scenario recognition result, the adaptive threshold calibration module calls the reference threshold matching the current scenario from the preset threshold library built in the adaptive threshold calibration module; At the same time, obtain the historical optimization parameters fed back by the dynamic model update module; According to the weight coefficient determined by the scenario recognition result, the reference threshold and the historical optimization parameters are weighted and fused to generate a composite reference value including the allowable maximum deviation range and the predicted deviation direction, and transmitted to the hierarchical compensation execution module.

5. The LED display matrix dynamic control system according to claim 4, characterized in that, The hierarchical compensation execution module executes a primary compensation strategy based on the composite reference value: When the timing deviation between the scan signal and the data signal collected by the timing feature extraction module is ≤ 1 / 3 of the composite reference value, the phase of the scan signal is adjusted by the digital delay line of the row driving circuit, and the adjustment amount is 0.8 times the measured timing deviation, so as to correct the small synchronization error between the scan signal and the data signal.

6. The LED display matrix dynamic control system according to claim 5, characterized in that The hierarchical compensation execution module executes a two-level compensation strategy based on the composite reference value: When the timing deviation between the scan signal and the data signal collected by the timing feature extraction module is greater than 1 / 3 of the composite reference value and less than or equal to 2 / 3 of the composite reference value, the phase of the scan signal of the digital delay line of the row driving circuit and the source output load capacitance of the column driving circuit are synchronously adjusted, and the medium synchronization error between the scan signal and the data signal is corrected through the dual-path adjustment.

7. The LED display matrix dynamic control system according to claim 6, characterized in that, The hierarchical compensation execution module executes a three-level compensation strategy based on the composite reference value: When the timing deviation between the scan signal and the data signal collected by the timing feature extraction module is greater than 2 / 3 of the composite reference value, the clock synchronization unit built in the timing feature extraction module is triggered to adjust the control voltage of the voltage-controlled oscillator, and the clock sources of the row driving and the column driving are relocked, which is used for the phase alignment of the clock sources of the row driving and the column driving, so as to eliminate the serious synchronization error between the scan signal and the data signal from the clock source level.

8. The LED display matrix dynamic control system according to claim 7, characterized in that, The dynamic model update module receives the adjustment parameters including the compensation adjustment amount output by the hierarchical compensation execution module, and the recent 100-frame historical data collected by the timing feature extraction module. Based on the above adjustment parameters and historical data, the classification weights of the k-nearest neighbor classifier and the weighting coefficients generated by the threshold in the adaptive threshold calibration module are encoded into chromosomes. The chromosomes include the parameter combinations of the classification weights and the weighting coefficients, which are used to optimize the scene recognition and threshold generation parameters of the adaptive threshold calibration module.

9. The LED display matrix dynamic control system according to claim 8, characterized in that, The dynamic model update module optimizes the chromosome parameters through the genetic algorithm, and takes the proportion of the number of frames in which the timing deviation between the scan signal and the data signal after being compensated by the hierarchical compensation execution module is ≤ the reference threshold as the fitness function; The top 30% of the individuals with high fitness are selected as the parents through roulette wheel selection, the offspring are generated by single-point crossover, and the classification weights and weighting coefficients in the chromosomes are randomly perturbed with a probability of 5%, and the chromosome parameters are iteratively optimized to improve the accuracy of scene recognition and the adaptability of threshold generation.

10. The LED display matrix dynamic control system according to claim 9, characterized in that, After the dynamic model update module completes 5 generations of genetic algorithm evolution each time, it extracts the optimal chromosome after evolution based on the evaluation result of the fitness function, obtains the corresponding classification weights of the k-nearest neighbor classifier and the weighting coefficients for threshold generation, and writes the above weights and coefficients into the scene library of the adaptive threshold calibration module, and updates the classification weights and threshold generation strategies corresponding to the static, dynamic, high-speed motion, and temperature change scene labels in the scene library, so that the adaptive threshold calibration module can more accurately match the synchronization control requirements of the current display scene.

Images