Eye pattern degradation compensation method and device, storage medium and electronic equipment

Abstract

The invention relates to an eye pattern degradation compensation method and device, a storage medium and electronic equipment, and the method comprises the steps: collecting various parameter data and an eye pattern of a display device, the various parameter data being related to the quality of the eye pattern: carrying out the feature extraction of the various parameter data and the eye pattern, and obtaining a fusion feature vector; the fusion feature vector is input into a degradation traceability model, an eye pattern degradation cause and a compensation strategy are obtained, the compensation strategy is executed to reduce the eye pattern degradation degree, and the degradation traceability model is constructed based on a GBDT model. The eye pattern degradation cause cannot be traced; and the compensation strategy cannot be determined.

Description

Technical Field

[0001] This application relates to the field of display technology, and in particular to an eye diagram degradation compensation method, compensation device, storage medium, and electronic device. Background Technology

[0002] Eye diagrams reflect the display quality of a display device. An eye diagram is a graph displayed on an oscilloscope that accumulates a series of digital signals transmitted from the driver board to the source driver. The eye diagram reflects the signal integrity transmitted from the driver board to the source driver. The signal integrity of the signal transmitted from the driver board to the source driver directly determines whether the source driver can provide accurate driving voltage to the pixel array, and thus determines the display quality of the display device. With the refresh rate of the display screen increasing to 500Hz+ and the transmission rate exceeding 48Gbps, the problem of eye diagram degradation (decrease in the signal integrity transmitted from the driver board to the source driver) has become the core bottleneck restricting the display quality of display devices.

[0003] However, existing technologies can only determine whether the eye diagram has deteriorated, but cannot trace the cause of the eye diagram deterioration or determine the compensation strategy. Summary of the Invention

[0004] This application provides an eye diagram degradation compensation method, an eye diagram degradation compensation device, a computer-readable storage medium, and an electronic device to solve the problems in the prior art that can only determine whether an eye diagram is degraded, but cannot trace the cause of eye diagram degradation or determine a compensation strategy.

[0005] In a first aspect, this application provides an eye diagram degradation compensation method, the method comprising: collecting various parameter data of a display device and an eye diagram, wherein the various parameter data are related to the quality of the eye diagram; extracting features from the various parameter data and the eye diagram to obtain a fused feature vector; inputting the fused feature vector into a degradation tracing model to obtain the causes of eye diagram degradation and a compensation strategy; and executing the compensation strategy to reduce the degree of eye diagram degradation, wherein the degradation tracing model is constructed based on the GBDT model.

[0006] Optionally, the multiple parameter data includes at least: the display panel model, cable specifications, bit error rate of the transmission link, power supply voltage of the chip on the driver board, temperature, ambient temperature, and ambient humidity. Feature extraction is performed on the multiple parameter data and the eye diagram to obtain a fused feature vector, including: encoding the non-numerical parameter data to obtain a static feature vector; standardizing the numerical parameter data to obtain standard parameter data; standardizing the eye diagram to obtain a standard eye diagram; and statistically processing the standard parameter data for a preset duration to obtain a dynamic environmental feature vector. The statistical processing includes at least: taking the average, taking the squared difference, and taking the extreme values. Each element in the static feature vector is then compared with... The weighted static feature vector is obtained by multiplying the corresponding weight coefficients. The weighted dynamic environment feature vector is obtained by multiplying each element in the dynamic environment feature vector with its corresponding weight coefficient. The weighted eye map is obtained by multiplying each element in the standard eye map with its corresponding weight coefficient, where the weight coefficients are determined based on their correlation with the quality of the eye map. The weighted static feature vector and the weighted dynamic environment feature vector are input into a trained first Transformer encoder to obtain a weighted temporal feature vector. The weighted eye map is input into a trained first CNN network to obtain a weighted eye map feature vector. The weighted temporal feature vector and the weighted eye map feature vector are input into a trained fully connected layer to obtain the fused feature vector.

[0007] Optionally, the fused feature vector is input into the degradation tracing model to obtain the causes of eye diagram degradation and the compensation strategy, and the compensation strategy is executed, including: inputting the static feature vector and the dynamic environment feature vector into a trained second Transformer encoder to obtain a temporal feature vector; inputting the standard eye diagram into a trained second CNN network to obtain an eye diagram feature vector; and inputting the temporal feature vector, the eye diagram feature vector, and the fused feature vector into the degradation tracing model to obtain the causes of eye diagram degradation and the compensation strategy.

[0008] Optionally, after acquiring various parameter data and eye diagrams of the display device, and before performing feature extraction on the various parameter data and eye diagrams to obtain a fused feature vector, the method includes: for each type of parameter data, if the sampling frequency of the parameter data is less than a preset frequency, processing the parameter data using a linear interpolation method to make the storage frequency of the parameter data equal to the preset frequency; and processing the eye diagram using the linear interpolation method to make the storage frequency of the eye diagram equal to the preset frequency.

[0009] Optionally, after extracting features from the various parameter data and the eye diagram to obtain a fused feature vector, the method further includes: inputting the fused feature vector from historical time periods into a trained quality prediction model to obtain eye diagram scores for future time periods, wherein the quality prediction model is composed of a multi-layer Bi-LSTM structure; and if the eye diagram scores at multiple consecutive times in the future time period are all less than a preset eye diagram score, inputting the fused feature vector into the degradation tracing model to obtain the causes of eye diagram degradation and the compensation strategy.

[0010] Optionally, after executing the compensation strategy, the method further includes: an acquisition step, acquiring multiple eye diagram quality characterization parameters, the multiple eye diagram quality characterization parameters including at least: eye diagram score and bit error rate of transmission link; an adjustment step, adjusting the adjustment value of the compensation parameter in the compensation strategy based on a preset percentage when the multiple eye diagram quality characterization parameters are all within the corresponding threshold range; and a repeating step, sequentially executing the acquisition step and the adjustment step M times or until the multiple eye diagram quality characterization parameters are all within the corresponding threshold range, where M is a positive integer.

[0011] Optionally, the compensation strategy includes: when there are multiple causes of eye diagram degradation, obtaining a level table for the current scene, the level table including the priority level of the causes of eye diagram degradation; and executing the compensation strategy corresponding to each cause of eye diagram degradation in descending order of priority level of the causes of eye diagram degradation in the level table for the current scene.

[0012] Secondly, this application provides an eye diagram degradation compensation device, the device comprising: an acquisition unit for acquiring various parameter data of a display device and an eye diagram, wherein the various parameter data are related to the quality of the eye diagram; a feature extraction unit for extracting features from the various parameter data and the eye diagram to obtain a fused feature vector; and a source tracing unit for inputting the fused feature vector into a degradation source tracing model to obtain the causes of eye diagram degradation and a compensation strategy, and executing the compensation strategy to reduce the degree of eye diagram degradation, wherein the degradation source tracing model is constructed based on the GBDT model.

[0013] Thirdly, this application provides a computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, it implements the steps of any of the eye diagram degradation compensation methods described above.

[0014] Fourthly, this application also provides an electronic device, the electronic device including a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of any of the eye diagram degradation compensation methods described above.

[0015] In this embodiment, various parameter data of the display device and eye diagrams are collected. These parameter data are all related to the quality of the eye diagrams. Feature extraction is performed on the various parameter data and eye diagrams to obtain a fused feature vector. The fused feature vector is input into the degradation tracing model to obtain the causes of eye diagram degradation and compensation strategies. The compensation strategies are executed to reduce the degree of eye diagram degradation, thereby solving the problem in the prior art that it can only determine whether the eye diagram is degraded, but cannot trace the causes of eye diagram degradation or determine the compensation strategy. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0019] Figure 1 A flowchart illustrating an eye diagram degradation compensation method provided in an embodiment of this application; Figure 2 This is a structural block diagram of an eye diagram degradation compensation device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] The following disclosure provides numerous different embodiments or examples for implementing various structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of the invention. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0022] To address the problem that existing technologies can only determine whether an eye diagram is degraded, but cannot trace the cause of the eye diagram degradation or determine a compensation strategy, this application provides an eye diagram degradation compensation method, an eye diagram degradation compensation device, a computer-readable storage medium, and an electronic device, which can solve the problem that existing technologies can only determine whether an eye diagram is degraded, but cannot trace the cause of the eye diagram degradation or determine a compensation strategy.

[0023] Figure 1 An eye diagram degradation compensation method provided in this application embodiment, such as Figure 1 As shown, the above method includes: Step S101: Collect various parameter data and eye diagrams of the display device, wherein the various parameter data are related to the quality of the eye diagrams. For example, the display device includes: a host, a driver board (the driver board integrates chips such as timing controllers), and a display panel. The display panel integrates a pixel array and a source driver. The host is a signal source that transmits image signals to the driver board. The driver board processes the image signals to generate timing signals and data signals, and transmits the timing signals and data signals to the source driver. The source driver converts the data signals into corresponding driving voltages and provides these driving voltages to the pixel array according to the timing signals, thereby lighting up the pixels and realizing image display.

[0024] For example, the types of the aforementioned parameter data include: non-numerical and numerical types. The non-numerical parameter data includes: static type, which includes at least: display panel model, display panel resolution, transmission protocol of the transmission link, and cable specifications. The numerical parameter data includes: dynamic real-time type and environment-aware type. The dynamic real-time parameter data includes at least: bit error rate of the transmission link, power supply voltage of the chip on the driver board, temperature of the chip on the driver board, and cable impedance. The environment-aware parameter data includes at least: ambient temperature of the display panel, ambient humidity of the display panel, and electromagnetic interference intensity of the display panel. The transmission link in the aforementioned parameter data refers to the signal transmission path related to the quality of the eye diagram, such as the signal transmission path between the driver board and the source driver. The cable in the aforementioned parameter data refers to the cable related to the quality of the eye diagram, such as all cables in the signal transmission path between the driver board and the source driver. The chip on the driver board in the aforementioned parameter data refers to the chip related to the quality of the eye diagram, such as a timing controller.

[0025] For example, the display panel model and resolution are stored in the EEPROM (Electrically Erasable Programmable Read-Only Memory) of the driver board, the transmission protocol of the transmission link is stored in the protocol chip register of the driver board, the cable specifications are stored in the cable ID chip, the bit error rate of the transmission link can be acquired using the MAX32670 bit error detector on the driver board, the power supply voltage of the chips on the driver board is acquired using the LTC2990 voltage monitor on the driver board, the temperature of the chips on the driver board can be acquired using the ADT7420 temperature sensor on the driver board, the cable impedance can be acquired using the impedance detection circuit on the driver board, the ambient temperature of the display panel can be acquired using the SHT35 temperature and humidity sensor (installed near the expansion interface of the driver board), the ambient humidity of the display panel can be acquired using the SHT35 temperature and humidity sensor (installed near the expansion interface of the driver board), and the electromagnetic interference intensity of the display panel can be acquired using the AD8318 power detector (installed near the signal shielding area of ​​the driver board).

[0026] For example, the eye diagram is acquired by the AD9680 eye diagram sensor on the driver board.

[0027] Step S102: Extract features from the above-mentioned multiple parameter data and eye diagram to obtain a fused feature vector; Step S103: Input the above-mentioned fused feature vector into the degradation tracing model to obtain the causes of eye diagram degradation and compensation strategies. Execute the compensation strategies to reduce the degree of eye diagram degradation. The above-mentioned degradation tracing model is constructed based on the GBDT model (Gradient Boosting Decision Tree).

[0028] Through the above embodiments, various parameter data and eye diagrams of the display device are collected. These parameter data are all related to the quality of the eye diagram. Feature extraction is performed on the various parameter data and eye diagrams to obtain a fused feature vector. The fused feature vector is input into the degradation tracing model to obtain the cause of eye diagram degradation and compensation strategy. The compensation strategy is executed to reduce the degree of eye diagram degradation, thereby solving the problem in the prior art that it can only determine whether the eye diagram is degraded, but cannot trace the cause of eye diagram degradation or determine the compensation strategy.

[0029] In an optional embodiment, after step S101 and before step S102, the method further includes: For each of the above parameter data, if the sampling frequency of the above parameter data is less than the preset frequency, a linear interpolation method is used to process the above parameter data so that the storage frequency of the above parameter data is equal to the preset frequency. The above-mentioned eye diagram is processed using the linear interpolation method described above, so that the storage frequency of the eye diagram is equal to the preset frequency described above.

[0030] Specifically, feature extraction is performed on the above-mentioned multiple parameter data and eye diagram to obtain a fused feature vector. Feature extraction is performed on the above-mentioned multiple parameter data and eye diagram according to a preset frequency to obtain the fused feature vector.

[0031] For example, the types of the aforementioned parameter data include: non-numerical and numerical types. The non-numerical parameter data types include: static type. The numerical parameter data types include: dynamic real-time type and environment-aware type. Static parameter data includes at least: display panel model, display panel resolution, transmission link protocol, and cable specifications. Numerical parameter data includes: dynamic real-time type and environment-aware type. Dynamic real-time parameter data includes at least: transmission link bit error rate, power supply voltage of the chip on the driver board, temperature of the chip on the driver board, and cable impedance. Environment-aware parameter data includes at least: display panel ambient temperature, display panel ambient humidity, and display panel electromagnetic interference intensity. For example, the preset frequency is 100Hz (10ms). Since static parameter data is sampled only once, at the preset frequency of 100Hz (10ms), static parameter data is stored every 10ms. The static parameter data stored each time is the same. The sampling frequency of the transmission link bit error rate is... The frequency of 10Hz (100ms) is less than the preset frequency. Therefore, linear interpolation is used to process the parameter data so that the storage frequency of the bit error rate of the transmission link is equal to 100Hz (10ms). The sampling frequency of the chip temperature on the driver board is 1Hz (1000ms), the sampling frequency of the cable impedance is 0.1Hz (10000ms), the sampling frequency of the ambient temperature of the display panel is 0.5Hz (20000ms), the sampling frequency of the ambient humidity of the display panel is 0.5Hz (2000ms), and the sampling frequency of the electromagnetic interference intensity of the display panel is 0.1Hz (10000ms). All of these require linear interpolation to make the storage frequency of these parameter data equal to 100Hz (10ms). The sampling frequency of the power supply voltage of the chip on the driver board is 100Hz (10ms) and does not require linear interpolation. The sampling frequency of the eye diagram is 10Hz (100ms, 10 frames / second) and requires linear interpolation to make the storage frequency of the eye diagram equal to 100Hz (10ms).

[0032] In this embodiment, linear interpolation is used to process the parameter data, ensuring that the storage frequency of the parameter data equals a preset frequency. Similarly, linear interpolation is used to process the eye diagram, ensuring that the storage frequency of the eye diagram equals a preset frequency. Linear interpolation can supplement missing sampling points without disrupting the original data trend, preserving the true change patterns of the parameter data and eye diagram. It also ensures that multiple parameter data and eye diagrams are synchronized over time, providing a standardized and unified data source for subsequent feature extraction and degradation tracing models. This avoids feature extraction deviations caused by frequency differences, ensuring the accuracy of eye diagram degradation cause judgment and the targeted nature of compensation strategies. By uniformly adapting the data storage frequency to the preset frequency, and by uniformly adapting the storage frequency of the parameter data and eye diagram to the preset frequency, and inputting the fused feature vector into the degradation tracing model at the preset frequency, the degradation tracing model can ensure real-time perception of parameter data and eye diagram changes in the display device, accurately capturing instantaneous state fluctuations at the preset frequency level. This meets the low-latency response requirements for eye diagram degradation cause judgment and compensation strategies in high real-time scenarios such as automotive and medical applications.

[0033] In an optional embodiment, step S102 above can be implemented as follows: The non-numerical parameter data mentioned above is encoded to obtain a static feature vector; For example, the types of non-numerical parameter data include: static type. Static parameter data includes at least: display panel model, display panel resolution, transmission protocol of transmission link, and cable specifications. Taking the encoding of display panel model as an example, if the display panel model is a finite enumeration type, for example, the display panel model only reflects the application field of the display panel, and the display panel model includes: TFT-LCD (automotive field), OLED (medical field), LED (data center field), then one-hot encoding is used, with 3 type numbers corresponding to 3 types of one-dimensional vectors. The encoding of TFT-LCD is [1,0,0], OLED... The encoding is [0,1,0], and the LED display panel encoding is [0,0,1]. If the display panel model reflects multi-dimensional attributes, such as size, resolution, and application field, an embedding layer encoding is used to map the multi-dimensional attributes into a low-dimensional dense vector (e.g., a 16-dimensional feature vector). Specifically, the multi-dimensional attributes are first converted into discrete labels: [15.6 inches = 2, 8K = 2, TFT-LCD = 1]. The discrete labels are then input into the pre-trained embedding layer and mapped into a 16-dimensional dense vector [0.23, 0.56, -0.12, 0.89, ... 0.47]. The encoding method for other non-numerical parameter data is similar.

[0034] The numerical parameter data above is standardized to obtain standard parameter data, and the eye diagram above is standardized to obtain a standard eye diagram. For example, the types of numerical parameter data include: dynamic real-time type and environment-aware type. Dynamic real-time type parameter data includes at least: the bit error rate of the transmission link, the power supply voltage of the chip on the driver board, the temperature of the chip on the driver board, and the impedance of the cable. Environment-aware type parameter data includes at least: the ambient temperature of the display panel, the ambient humidity of the display panel, and the electromagnetic interference intensity of the display panel. Taking the standardization of the bit error rate of the transmission link as an example, the bit error rate of the transmission link is sampled with a period of 10 ms. It should be noted that the above storage frequency is the actual sampling frequency, using the formula: The bit error rate (BER) of the transmission link is standardized, where x_norm is the standard BER of the transmission link at the current sampling time, x is the BER of the transmission link at the current sampling time, x_mean is the mean of the BER of the transmission link over the past 24 hours at the current sampling time, and x_std is the standard deviation of the BER of the transmission link over the past 24 hours at the current sampling time. The standardization process for other numerical parameters is similar, as is the standardization process for eye diagrams. It should be noted that in the standardization process of eye diagrams, the average value and standard deviation of all pixels in the eye diagram are used to standardize each pixel in the eye diagram.

[0035] The above standard parameter data with a preset duration are statistically processed to obtain a dynamic environment feature vector. The above statistical processing includes at least: taking the average value, taking the squared difference, and taking the extreme value. For example, the above statistical processing also includes: taking the slope, etc.

[0036] For example, statistical processing is performed on the aforementioned standard parameter data for a preset duration to obtain a dynamic environment feature vector. For instance, for each standard parameter data, statistical processing is performed on the standard parameter data for the past 10 sampling times (preset duration) at the current sampling time to extract the statistical feature values ​​(mean, squared difference, extreme values, etc.) of the standard parameter data at the current sampling time. All standard parameter data at the current sampling time and the statistical feature values ​​of all standard parameter data at the current sampling time are concatenated to obtain the dynamic environment feature vector at the current sampling time. The dynamic environment feature vector can reflect the changes in numerical parameter data (dynamic real-time parameter data and environment-aware parameter data). For example, the dynamic environment feature vector can reflect the temperature rise of the chip on the driver board, the voltage fluctuation of the chip, and the increase in the bit error rate of the transmission link.

[0037] Calculate the product of each element in the static feature vector and its corresponding weight coefficient to obtain the weighted static feature vector. Calculate the product of each element in the dynamic environment feature vector and its corresponding weight coefficient to obtain the weighted dynamic environment feature vector. Calculate the product of each element in the standard eye diagram and its corresponding weight coefficient to obtain the weighted eye diagram. The weight coefficients are determined based on their correlation with the quality of the eye diagram. For example, the weighting coefficients are determined based on the correlation between the parameter data (eye diagram) and the quality of the eye diagram, as well as the actual values ​​of the parameter data. For example, various parameter data include: the impedance deviation of the transmission link (the difference between the impedance of the transmission link and the preset impedance), and the temperature of the chip on the driver board. When the impedance deviation of the transmission link is greater than 3Ω, the weighting coefficient is 0.3 (indicating that it has the greatest impact on the quality of the eye diagram). When the temperature of the chip on the driver board is greater than 65°C, the weighting coefficient of the chip temperature is 0.25 (indicating that it has a relatively large impact on the quality of the eye diagram).

[0038] Specifically, the sum of each element in the static feature vector and its corresponding weight coefficient, the sum of each element in the dynamic environment feature vector and its corresponding weight coefficient, and the sum of each element in the standard eye diagram and its corresponding weight coefficient is 1, and each element in the standard eye diagram has the same weight coefficient.

[0039] Input the above weighted static feature vector and the above weighted dynamic environment feature vector into the trained first Transformer encoder to obtain the weighted temporal feature vector; Specifically, it should be noted that the weighted static feature vector reflects the hardware attributes of the display device. Under different hardware attributes, the allowable normal variation range of numerical parameter data is different. For example, the normal fluctuation range of impedance of thin-diameter cables is different from that of thick-diameter cables. The core function of the weighted static feature vector is to provide a clear hardware attribute reference for the degradation tracing model, help the degradation tracing model accurately define the normal variation range of numerical parameter data, thereby accurately distinguishing between normal fluctuations and abnormal changes in numerical parameter data, and ultimately improving the accuracy of degradation cause judgment and the reliability of compensation strategy.

[0040] For example, the first Transformer encoder is used to capture the causal relationship of changes in numerical (real-time dynamic and environment-aware) parameter data and output a weighted time-series feature vector (dimension 1×128). The weighted time-series feature vector is used to reflect the causal relationship of changes in numerical (real-time dynamic and environment-aware) parameter data in the time dimension. For example, the weighted time-series feature vector reflects the voltage fluctuation of the chip caused by the temperature rise of the chip on the driver board (1.2V→1.18V).

[0041] The above weighted eye diagram is input into the first trained CNN network (Convolutional Neural Network) to obtain the weighted eye diagram feature vector; For example, a two-dimensional eye map contains features such as eye height, eye width, and jitter, which reflect the quality of the eye map (whether it is degraded). The first CNN network can automatically mine these features.

[0042] For example, the weighted eye map (256×256 in dimension) is input into the first trained CNN network to obtain the weighted eye map feature vector (1×256 in dimension).

[0043] The weighted temporal feature vector and the weighted eye map feature vector are input into the trained fully connected layer to obtain the fused feature vector.

[0044] For example, the weighted temporal feature vector (dimension 1×128) and the weighted eye map feature vector (dimension 1×256) are input into the trained fully connected layer to obtain the fused feature vector (dimension 1×64).

[0045] Specifically, the fused feature vector includes: hardware attributes of the display device, causal relationships of changes in numerical parameter data, and eye diagram feature information (reflecting whether the eye diagram has deteriorated). The aforementioned degradation tracing model is built based on the GBDT model. As an ensemble learning model, the GBDT model can effectively capture nonlinear relationships in the data. By training the degradation tracing model, the degradation tracing model learns the causal relationships of changes in numerical parameter data, the feature information of the eye diagram, and the relationship between compensation strategies. This enables the GBDT model to accurately capture the causal relationships of changes in numerical parameter data and the relationship between the feature information of the eye diagram from the fused feature vector, thereby achieving accurate localization of degradation causes and outputting compensation strategies adapted to the hardware attributes of the display device.

[0046] For example, the causal relationship of the changes in numerical parameter data captured by the degradation tracing model is that the temperature of the chip on the driver board increases, causing voltage fluctuations in the chip (1.2V→1.18V). The characteristic information of the eye diagram captured by the degradation tracing model is that the eye diagram edge jitter increases. The degradation tracing model outputs that the cause of eye diagram degradation is that the chip on the driver board is overheated (temperature greater than 65°C), and the compensation strategy is to adjust the cooling fan speed to 1200RPM (rated 1800RPM).

[0047] In this embodiment, a weighting coefficient determined based on the correlation with eye diagram quality is used to weight the static feature vector, dynamic environment feature vector, and standard eye diagram respectively. This can adaptively highlight key features that have a significant impact on eye diagram quality and weaken features with weak correlation or redundancy. This allows the degradation tracing model to focus more on key factors closely related to degradation during feature learning and inference, improving the effectiveness and discriminative power of features, thereby improving the accuracy of eye diagram degradation cause localization and the reliability of compensation strategies.

[0048] In an optional embodiment, after step S102, the method further includes: The above-mentioned fused feature vectors from historical periods are input into the trained quality prediction model to obtain the eye map score for future periods. The quality prediction model consists of a multi-layer Bi-LSTM structure (Bi-directional Long Short-Term Memory network). If the eye diagram score is lower than the preset eye diagram score at multiple consecutive moments in the aforementioned future time period, the aforementioned fused feature vector is input into the aforementioned degradation tracing model to obtain the aforementioned causes of eye diagram degradation and the aforementioned compensation strategy.

[0049] For example, the fused feature vectors accumulated over the past 36 hours (one fused feature vector every 10ms, totaling 1.296 × 10^6) are used. 7 (A fusion feature vector) is used to obtain the eye map score for the next 96 hours.

[0050] For example, the quality prediction model consists of a multi-layer Bi-LSTM structure, which includes both forward LSTM layers and backward LSTM layers. This allows the model to integrate past and future contextual information at the same time step. By training the quality prediction model, it learns the causal relationships between changes in numerical parameter data over historical periods, the feature information of eye diagrams, and the relationship between eye diagram scores in future periods. This enables the quality prediction model to effectively capture the causal relationships between changes in numerical parameter data over long periods and the relationship between the feature information of eye diagrams, accurately captured from the fused feature vectors. Based on the causal relationships between changes in numerical parameter data over long periods and the relationship between the feature information of eye diagrams, it predicts the eye diagram scores for future periods.

[0051] In this embodiment, when the quality prediction model infers the eye diagram score for the future period based on the fused feature vector of the historical period, if the eye diagram score for multiple consecutive moments in the future period is less than the preset eye diagram score, it indicates that the quality of the eye diagram has a serious deterioration trend. At this time, the fused feature vector already contains the causal relationship of the changes in numerical parameter data that cause this trend. Then, the fused feature vector is input into the deterioration tracing model to obtain the causes of the eye diagram deterioration and the compensation strategy. The compensation strategy is then executed to avoid serious eye diagram deterioration.

[0052] In an optional embodiment, step S103 above can be implemented as follows: Input the above static feature vector and the above dynamic environment feature vector into the trained second Transformer encoder to obtain the temporal feature vector; The standard eye diagram described above is input into the trained second CNN network to obtain the eye diagram feature vector; By inputting the aforementioned temporal feature vector, eye diagram feature vector, and fused feature vector into the aforementioned degradation tracing model, the aforementioned causes of eye diagram degradation and the aforementioned compensation strategy are obtained.

[0053] In this embodiment, the aforementioned fused feature vector is obtained by inputting the weighted temporal feature vector and the weighted eye map feature vector into a trained fully connected layer. Due to compression by the fully connected layer, some information is lost. Therefore, the static feature vector and the dynamic environment feature vector are input into a trained second Transformer encoder to obtain the temporal feature vector, and the standard eye map is input into a trained second CNN network to obtain the eye map feature vector. The temporal feature vector, the eye map feature vector, and the fused feature vector are input into the degradation tracing model to compensate for the information lost by the fully connected layer, ensuring the accuracy of eye map degradation cause localization and the reliability of the compensation strategy.

[0054] In an optional embodiment, after step S103, the method further includes: The acquisition step involves acquiring multiple eye diagram quality characterization parameters, which include at least: eye diagram score and bit error rate of the transmission link. The adjustment steps involve adjusting the compensation parameters in the compensation strategy based on a preset percentage, provided that all the above-mentioned eye diagram quality characterization parameters are within the corresponding threshold range. For example, the compensation strategy is to adjust the cooling fan speed to 1200 RPM (rated 1800 RPM). If the eye diagram quality characterization parameters are still not up to standard after the compensation strategy is executed, the cooling fan speed is adjusted from 1200 RPM to 1260 RPM (preset percentage 5%).

[0055] Repeat the steps, performing the above acquisition step and the above adjustment step N times in sequence, or until all the above eye diagram quality characterization parameters are within the corresponding threshold range, where N is a positive integer.

[0056] In this embodiment, after the compensation strategy is executed, several eye diagram quality characterization parameters still do not meet the standards (e.g., eye diagram score is less than 90 points and bit error rate of transmission link is greater than 5 × 10). -14 At this point, the adjustment value of the compensation parameter in the above compensation strategy is adjusted based on a preset percentage (for example, preset percentage ±10%, up to 3 times) to ensure that the eye diagram quality finally meets the standard.

[0057] In an optional embodiment, step S103 above can be implemented as follows: When there are multiple causes of eye diagram degradation, obtain the level table of the current scene. The level table includes the priority level of the causes of eye diagram degradation. According to the priority order of the causes of eye diagram degradation in the current scenario's hierarchy table, from high to low, the compensation strategies corresponding to each of the aforementioned causes of eye diagram degradation are executed.

[0058] For example, the current scenario types include: automotive scenario, medical scenario, and data center scenario. For the automotive scenario, the rating table is: chip overheating > electromagnetic interference > impedance mismatch > cable crosstalk. For the medical scenario, the rating table is: impedance mismatch > excessive bit error rate > cable crosstalk > temperature and humidity. For the data center scenario, the rating table is: electromagnetic interference > cable crosstalk > chip overheating > temperature and humidity.

[0059] For example, Table 1 provides the causes of eye diagram degradation and corresponding compensation strategies for three scenarios: automotive, medical, and data center.

[0060] In this embodiment, considering that there are multiple causes of eye diagram degradation, the execution priority of the compensation strategy corresponding to the cause of eye diagram degradation is different in different scenarios. Therefore, it is necessary to execute the corresponding compensation strategy according to the priority level in the level table of the scenario.

[0061] Table 1

[0062] Figure 2 An eye diagram degradation compensation device provided in this application embodiment, such as Figure 2 As shown, the above-mentioned device includes: Acquisition unit 10 is used to acquire various parameter data and eye diagrams of the display device, wherein the various parameter data are related to the quality of the eye diagrams. Feature extraction unit 20 is used to extract features from the above-mentioned multiple parameter data and the above-mentioned eye diagram to obtain a fused feature vector; The source tracing unit 30 is used to input the above-mentioned fused feature vector into the degradation source tracing model to obtain the causes of eye diagram degradation and compensation strategies, and execute the compensation strategies to reduce the degree of eye diagram degradation. The above-mentioned degradation source tracing model is constructed based on the GBDT model (Gradient Boosting Decision Tree).

[0063] Through the above embodiments, various parameter data and eye diagrams of the display device are collected. These parameter data are all related to the quality of the eye diagram. Feature extraction is performed on the various parameter data and eye diagrams to obtain a fused feature vector. The fused feature vector is input into the degradation tracing model to obtain the cause of eye diagram degradation and compensation strategy. The compensation strategy is executed to reduce the degree of eye diagram degradation, thereby solving the problem in the prior art that it can only determine whether the eye diagram is degraded, but cannot trace the cause of eye diagram degradation or determine the compensation strategy.

[0064] In an optional embodiment, the above-described apparatus further includes: For each of the above parameter data, if the sampling frequency of the above parameter data is less than the preset frequency, a linear interpolation method is used to process the above parameter data so that the storage frequency of the above parameter data is equal to the preset frequency. The above-mentioned eye diagram is processed using the linear interpolation method described above, so that the storage frequency of the eye diagram is equal to the preset frequency described above.

[0065] In an optional embodiment, the feature extraction unit described above is used for: The non-numerical parameter data mentioned above is encoded to obtain a static feature vector; The numerical parameter data above is standardized to obtain standard parameter data, and the eye diagram above is standardized to obtain a standard eye diagram. The above standard parameter data with a preset duration are statistically processed to obtain a dynamic environment feature vector. The above statistical processing includes at least: taking the average value, taking the squared difference, and taking the extreme value. Calculate the product of each element in the static feature vector and its corresponding weight coefficient to obtain the weighted static feature vector. Calculate the product of each element in the dynamic environment feature vector and its corresponding weight coefficient to obtain the weighted dynamic environment feature vector. Calculate the product of each element in the standard eye diagram and its corresponding weight coefficient to obtain the weighted eye diagram. The weight coefficients are determined based on their correlation with the quality of the eye diagram. Input the above weighted static feature vector and the above weighted dynamic environment feature vector into the trained first Transformer encoder to obtain the weighted temporal feature vector; Input the above weighted eye diagram into the trained first CNN network to obtain the weighted eye diagram feature vector; The weighted temporal feature vector and the weighted eye map feature vector are input into the trained fully connected layer to obtain the fused feature vector.

[0066] In an optional embodiment, the above-described apparatus is used for: The above-mentioned fused feature vectors from historical time periods are input into the trained quality prediction model to obtain the eye map score for future time periods. The above-mentioned quality prediction model consists of a multi-layer Bi-LSTM structure. If the eye diagram score is lower than the preset eye diagram score at multiple consecutive moments in the aforementioned future time period, the aforementioned fused feature vector is input into the aforementioned degradation tracing model to obtain the aforementioned causes of eye diagram degradation and the aforementioned compensation strategy.

[0067] In an optional embodiment, the compensation unit is used for: Input the above static feature vector and the above dynamic environment feature vector into the trained second Transformer encoder to obtain the temporal feature vector; The standard eye diagram described above is input into the trained second CNN network to obtain the eye diagram feature vector; By inputting the aforementioned temporal feature vector, eye diagram feature vector, and fused feature vector into the aforementioned degradation tracing model, the aforementioned causes of eye diagram degradation and the aforementioned compensation strategy are obtained.

[0068] In an optional embodiment, the above-described apparatus is used for: The acquisition step involves acquiring multiple eye diagram quality characterization parameters, which include at least: eye diagram score and bit error rate of the transmission link. The adjustment steps involve adjusting the compensation parameters in the compensation strategy based on a preset percentage, provided that all the above-mentioned eye diagram quality characterization parameters are within the corresponding threshold range. Repeat the steps, performing the above acquisition step and the above adjustment step N times in sequence, or until all the above eye diagram quality characterization parameters are within the corresponding threshold range, where N is a positive integer.

[0069] In an optional embodiment, the compensation unit is used for: When there are multiple causes of eye diagram degradation, obtain the level table of the current scene. The level table includes the priority level of the causes of eye diagram degradation. According to the priority order of the causes of eye diagram degradation in the current scenario's hierarchy table, from high to low, the compensation strategies corresponding to each of the aforementioned causes of eye diagram degradation are executed.

[0070] This application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the eye diagram degradation compensation method provided in any of the foregoing method embodiments.

[0071] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the eye diagram degradation compensation method provided in any of the foregoing method embodiments.

[0072] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0074] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0075] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for compensating for eye diagram degradation, characterized in that, The method includes: The system collects various parameter data and eye diagrams from the display device, wherein the various parameter data are related to the quality of the eye diagrams. Feature extraction is performed on the various parameter data and the eye diagram to obtain a fused feature vector; The fused feature vector is input into the degradation tracing model to obtain the causes of eye diagram degradation and compensation strategies. The compensation strategies are then executed to reduce the degree of eye diagram degradation. The degradation tracing model is constructed based on the GBDT model.

2. The method according to claim 1, characterized in that, The various parameter data include at least: the display panel model, cable specifications, bit error rate of the transmission link, power supply voltage of the chip on the driver board, temperature, ambient temperature, and ambient humidity. Feature extraction is performed on the various parameter data and the eye diagram to obtain a fused feature vector, including: The non-numerical parameter data is encoded to obtain a static feature vector; The numerical parameter data is standardized to obtain standard parameter data, and the eye diagram is standardized to obtain a standard eye diagram. Statistical processing is performed on the standard parameter data for a preset duration to obtain a dynamic environment feature vector. The statistical processing includes at least: taking the average value, taking the squared difference, and taking the extreme value. The product of each element in the static feature vector and its corresponding weight coefficient is calculated to obtain a weighted static feature vector. The product of each element in the dynamic environment feature vector and its corresponding weight coefficient is calculated to obtain a weighted dynamic environment feature vector. The product of each element in the standard eye diagram and its corresponding weight coefficient is calculated to obtain a weighted eye diagram. The weight coefficients are determined based on their correlation with the quality of the eye diagram. The weighted static feature vector and the weighted dynamic environment feature vector are input into the trained first Transformer encoder to obtain the weighted temporal feature vector. The weighted eye map is input into the trained first CNN network to obtain the weighted eye map feature vector; The weighted temporal feature vector and the weighted eye map feature vector are input into the trained fully connected layer to obtain the fused feature vector.

3. The method according to claim 2, characterized in that, The fused feature vector is input into the degradation tracing model to obtain the causes of eye diagram degradation and compensation strategies. The compensation strategies are then executed, including: The static feature vector and the dynamic environment feature vector are input into the trained second Transformer encoder to obtain the temporal feature vector; The standard eye diagram is input into the trained second CNN network to obtain the eye diagram feature vector; The temporal feature vector, the eye diagram feature vector, and the fused feature vector are input into the degradation tracing model to obtain the causes of eye diagram degradation and the compensation strategy.

4. The method according to claim 1, characterized in that, After acquiring various parameter data and eye diagrams of the display device, and before extracting features from the various parameter data and eye diagrams to obtain a fused feature vector, the method includes: For each type of parameter data, if the sampling frequency of the parameter data is less than the preset frequency, a linear interpolation method is used to process the parameter data so that the storage frequency of the parameter data is equal to the preset frequency; The eye diagram is processed using the linear interpolation method to make the storage frequency of the eye diagram equal to the preset frequency.

5. The method according to claim 1, characterized in that, After extracting features from the various parameter data and the eye diagram to obtain a fused feature vector, the method further includes: The fused feature vectors from historical time periods are input into the trained quality prediction model to obtain eye map scores for future time periods. The quality prediction model consists of a multi-layer Bi-LSTM structure. If the eye diagram score is less than the preset eye diagram score at multiple consecutive times in the future time period, the fused feature vector is input into the degradation tracing model to obtain the cause of the eye diagram degradation and the compensation strategy.

6. The method according to claim 5, characterized in that, After implementing the compensation strategy, the method further includes: The acquisition step involves acquiring multiple eye diagram quality characterization parameters, which include at least: eye diagram score and bit error rate of the transmission link; The adjustment step involves adjusting the adjustment value of the compensation parameter in the compensation strategy based on a preset percentage when all the multiple eye diagram quality characterization parameters are within the corresponding threshold range. Repeat the steps, performing the acquisition step and the adjustment step M times in sequence, or until all the multiple eye diagram quality characterization parameters are within the corresponding threshold range, where M is a positive integer.

7. The method according to claim 1, characterized in that, Implementing compensation strategies, including: When there are multiple causes of eye diagram degradation, obtain the level table of the current scene, which includes the priority level of the causes of eye diagram degradation. According to the priority of the causes of eye diagram degradation in the current scene's hierarchy table, from high to low, the compensation strategy corresponding to each cause of eye diagram degradation is executed.

8. An eye diagram degradation compensation device, characterized in that, The device includes: The acquisition unit is used to acquire various parameter data and eye diagrams of the display device, wherein the various parameter data are related to the quality of the eye diagram: The feature extraction unit is used to extract features from the multiple parameter data and the eye diagram to obtain a fused feature vector. The source tracing unit is used to input the fused feature vector into the degradation source tracing model to obtain the causes of eye diagram degradation and compensation strategies, and execute the compensation strategies to reduce the degree of eye diagram degradation. The degradation source tracing model is constructed based on the GBDT model.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the eye diagram degradation compensation method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the eye diagram degradation compensation method as described in any one of claims 1 to 7.

Images