Method for establishing corrected current setpoint, training method, and electronic terminal

Abstract

A method for establishing a corrected current setpoint, a training method, and an electronic terminal (110). The method for establishing a corrected current setpoint is used for a backlight module (140), the backlight module (140) comprising multiple light-emitting units (141), and the multiple light-emitting units (141) being driven by current to emit light. The establishment method comprises: driving and measuring a measured current value set corresponding to the light-emitting units (141) at multiple dimming levels (301); correcting the light-emitting units (141) on the basis of the measured current value set, to generate a corrected current preliminary setpoint set (302); on the basis of a portion of the light-emitting units (141) among the light-emitting units (141), obtaining multiple dimming levels, the measured current value set, and the corrected current preliminary setpoint set, to establish an input data matrix (x) (303); and inputting the input data matrix (x) into a training completion model (M) to generate an output data matrix (G(x)) (304), the output data matrix (G(x)) comprising a corrected current optimal setpoint set corresponding to each light-emitting unit (141) at multiple dimming levels.

Description

Methods for establishing calibration current setpoints, training methods, and electronic terminals

[0001] This application claims priority to Chinese Patent Application No. 202411514120.3, filed on October 28, 2024, entitled "Method for Establishing Correction Current Setting Value and Electronic Terminal Using the Method", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This disclosure relates to a current correction method for establishing a correction current setting value for the light-emitting unit of a backlight module. Background Technology

[0003] Monitors are among the most common electronic devices in modern life, used in various scenarios and situations. Some monitors include a backlight module to provide the light source. This backlight module has multiple light-emitting units, and with local dimming technology, the brightness of these light-emitting units can be independently controlled, thereby improving the monitor's contrast. The brightness of the light-emitting units is controlled and adjusted through current amplitude and pulse width modulation (PWM). However, each light-emitting unit differs in production, manufacturing, and use. If each light-emitting unit is driven with the same current value, the brightness of each unit will be different. Furthermore, the position or density of each light-emitting unit in the backlight module is not entirely the same. Therefore, the current value required to drive each light-emitting unit will not be the same if a uniform brightness is desired within the backlight module.

[0004] To ensure uniform brightness across all light-emitting units within a backlight module, current measurement and calibration are typically performed for each unit at various brightness levels to obtain a calibration current setting. For example, consider a backlight module with 2400 light-emitting units in one area, representing 256 brightness levels. A single current measurement of each unit takes approximately 0.4 seconds. Therefore, measuring the current at each brightness level for all units in that area would take approximately 245760 seconds. Adding the time required for calibration, the overall operation is extremely time-consuming, making it difficult to reliably perform current measurement and calibration for all light-emitting units in every area of ​​the backlight module in a practical production line.

[0005] Therefore, the problem that must be overcome is how to quickly provide a more accurate calibration current setting value for each light-emitting unit to drive each light-emitting unit to emit uniform brightness, thereby significantly reducing the overall detection and calibration time, and how to use the light-emitting units on the backlight module or on the irregularly shaped light panel to achieve uniform brightness of the light-emitting units in the backlight module or the irregularly shaped light panel by using different calibration current settings for each light-emitting unit. Summary of the Invention

[0006] The embodiments disclosed herein propose a method for establishing calibration current settings, applicable to backlight modules. This backlight module includes multiple light-emitting units that are driven by current to emit light. The method includes: driving these light-emitting units; measuring the current of each light-emitting unit at various dimming levels to obtain a set of measured current values ​​for each light-emitting unit at each dimming level; calibrating each light-emitting unit based on this set of measured current values ​​to generate a preliminary calibration current setting set for each light-emitting unit; establishing an input data matrix based on a subset of these light-emitting units, obtaining the various dimming levels, the set of measured current values, and the preliminary calibration current setting set; and inputting this input data matrix into a trained model to generate an output data matrix. This output data matrix contains a set of optimal calibration current settings for each light-emitting unit at various dimming levels.

[0007] The method for establishing the calibration current setting value disclosed herein can improve upon the conventional method which requires measuring the current value of each luminance level of each light-emitting unit located at different positions. The method disclosed herein only requires measuring the current of each luminance level of a subset (rather than all) of the light-emitting units as input for training the current calibration model. Then, the method disclosed herein can quickly infer the calibration current setting value of all light-emitting units at each luminance level, which can greatly save the time spent measuring the current value of each light-emitting unit at different luminance levels and calibrating the luminance of each light-emitting unit.

[0008] Meanwhile, the electronic terminal provided in this disclosure can quickly infer the correction current setting value of other different brightness levels through current measurement values, so that each light-emitting unit can provide a uniform light source on irregularly shaped light panels or backlight modules. Attached Figure Description

[0009] To make the above-described features and advantages of this disclosure more apparent and understandable, specific embodiments are described below in conjunction with the accompanying drawings for detailed explanation.

[0010] Figure 1 is a schematic diagram of a current correction system according to one embodiment.

[0011] Figure 2 is a schematic diagram illustrating multiple areas on a display panel and their corresponding light-emitting units according to an embodiment.

[0012] Figure 3 is a flowchart of a method for establishing a correction current setting value according to a first embodiment of the present disclosure.

[0013] Figure 4 is a schematic diagram of the steps of a method for establishing a correction current setting value according to a first embodiment of the present disclosure.

[0014] Figure 5 is an illustrative schematic diagram of a specific light-emitting unit according to a second embodiment of the present disclosure.

[0015] Figure 6 is a flowchart of the training phase of the method for establishing the correction current setting value according to the first embodiment of the present disclosure, and may also be a flowchart of the method for training the correction current setting value according to the second embodiment.

[0016] Figure 7 is a schematic diagram of the steps of a training method for correcting current setting values ​​according to a second embodiment of the present disclosure. Detailed Implementation

[0017] Figure 1 is a schematic diagram of a current correction system according to an embodiment. Referring to Figure 1, the current correction system 100 includes an electronic terminal 110 and a display device 120. The electronic terminal 110 may be a personal computer, a server, or various electronic devices with computing capabilities. The display device 120 includes a circuit 130, a backlight module 140, and a display panel 150. The circuit 130 includes a timing controller 131 and a microcontroller unit (MCU) 132. The microcontroller 132 may also be a programmable gate array (FPGA), and therefore should not be limited to the microcontroller 132 disclosed in this embodiment. The backlight module 140 includes a plurality of light-emitting units, such as light-emitting diodes, which are driven by the current of the backlight module 140 to provide a backlight source. The display panel 150 is, for example, a liquid crystal display panel. For example, the display device 120 is a liquid crystal display device employing full array local dimming (FALD) technology.

[0018] Figure 2 is a schematic diagram illustrating multiple regions and corresponding light-emitting units on a display panel 150 according to an embodiment. Referring to Figures 1 and 2, in the embodiment of Figure 2, the display panel 150 includes 15 regions 151, each region 151 corresponding to multiple light-emitting units 141 included in the backlight module 140. Here, the brightness of each light-emitting unit 141 can be controlled by providing different currents to increase screen contrast. For example, when the image to be displayed in a certain region 151 is darker, the brightness of the corresponding light-emitting unit 141 can be reduced; conversely, when the image to be displayed in a certain region 151 is brighter, the brightness of the corresponding light-emitting unit 141 can be increased. Figure 2 is merely an example; this disclosure does not limit the number of regions 151 included in the display panel 150, nor does it limit the number of light-emitting units 141 corresponding to each region 151. Furthermore, each region 151 can also correspond to a different number of light-emitting units 141, which is also an extension of this disclosure.

[0019] It is worth noting that, due to differences in production, manufacturing, and use, the luminance of different light-emitting units 141 will vary when driven by the same current value. Furthermore, the luminance of the light-emitting unit 141 can be divided into 256 levels, and the current value for each level of luminance of the same light-emitting unit 141 will be different. Different light-emitting units 141 will also have different current values ​​at the same level of luminance. Additionally, if Figure 2 shows different numbers of light-emitting units 141 corresponding to different regions 151, the density of light-emitting units 141 within each region 151 will differ. Therefore, it can be understood that if each region 151 is to present the same level of luminance, the current value of each light-emitting unit 141 at the same level of luminance will vary slightly due to the different densities of light-emitting units 141 within each region 151.

[0020] Referring to Figure 1, when displaying an image, the timing controller 131 calculates the dimming level of each area 151 on the display panel 150. This dimming level indicates the required backlight brightness. Based on the calculated dimming level, the timing controller 131 calculates a current preset value, which is used to drive the light-emitting unit 141 to generate a specific current. In some embodiments, the current preset value is directly related to the dimming level. For example, the dimming level range is represented by 0 to 255 levels, while the current preset value range is represented by 0 to 1023 levels. The mapping relationship between the dimming level and the current preset value can be linear or non-linear. The order of the dimming level and current preset value ranges described above is merely an example, and this disclosure is not limited thereto. The order of the dimming level range can also be the same as the order of the current preset value range.

[0021] Due to variations in the manufacturing process, the current preset setting may not be sufficient to drive the light-emitting unit 141 to provide the required brightness; therefore, the current preset setting needs to be calibrated. The microcontroller 132 contains multiple output data matrices, each corresponding to a light-emitting unit 141. The output data matrices record multiple calibrated current optimal settings corresponding to multiple dimming levels. The timing controller 131 can access the output data matrices in the microcontroller 132 based on the calculated dimming level to obtain the corresponding calibrated current optimal setting. Then, it drives the corresponding light-emitting unit 141 based on this calibrated current optimal setting, thus enabling the light-emitting unit 141 to provide the expected brightness. A method for establishing the calibrated current setting is proposed, executed by the electronic terminal 110, and utilizes a generative adversarial network (GAN) to generate the aforementioned output data matrices.

[0022] Figure 3 is a flowchart of a method 300 for establishing a calibration current setting value according to a first embodiment of the present disclosure. In step 301, the plurality of light-emitting units 141 included in the backlight module 140 are driven, and then the current of these light-emitting units 141 at various dimming levels (0 to 255) is measured to obtain a set of measured current values ​​corresponding to each light-emitting unit 141 at various dimming levels. In other words, the set of measured current values ​​consists of 256 measured current values ​​corresponding to the light-emitting units 141 at dimming levels from 0 to 255. The specific method of step 301 is as follows: the electronic terminal 110 sends out a command to cause the timing controller 131 to drive each light-emitting unit 141 to emit light at multiple dimming levels according to multiple default current settings corresponding to multiple dimming levels (0 to 255 levels), and to measure the current of each light-emitting unit 141 at multiple dimming levels (0 to 255 levels) by means of an ammeter or a power meter, so as to obtain the set of measured current values ​​corresponding to each light-emitting unit 141 at multiple dimming levels.

[0023] In step 302, each light-emitting unit 141 is calibrated based on this set of measured current values ​​to generate a preliminary set of calibration current values ​​for each light-emitting unit 141. In other words, the preliminary set of calibration current values ​​consists of 256 preliminary calibration current values ​​corresponding to dimming levels 0 to 255 for each light-emitting unit 141. The specific method of step 302 is as follows: if the measured current value corresponding to a certain dimming level is less than the target current value corresponding to that dimming level, the preliminary calibration current value corresponding to that dimming level is obtained by slightly increasing the current preset value; if the measured current value corresponding to that dimming level is greater than the target current value corresponding to that dimming level, the preliminary calibration current value corresponding to that dimming level is obtained by slightly decreasing the current preset value. The target current value can be an objective target value obtained by the electronic terminal 110 after program calculation, a subjective setting value required by the user, or a specification setting value required by the manufacturer of the electronic terminal 110. For example, the preset current setting value corresponding to dimming level 255 can be "995". Under this preset current setting value, the target current value is 64 mA. That is, 64 mA can make the light-emitting unit 141 provide the brightness corresponding to the default current setting value of "995". Step 302 is to determine whether the actual measured current value is greater than or less than 64 mA. If the measured current value corresponding to dimming level 255 is 65.3 mA (greater than the target current value of 64 mA), then the preset current setting value of "995" is slightly reduced to obtain the preliminary adjusted setting value of program corresponding to dimming level 255, for example, "994".

[0024] It is worth noting that the aforementioned current values ​​are concepts corresponding to, for example, the current values ​​of the 1024th order (0 to 1023). The current level of the current value can be in the ampere (A) level, milliampere (mA) level, microampere (μA) level, nanoampere (nA) level, but is not limited thereto. In addition, the difference between each order is not the same. For example, the difference between the current values ​​of the 801st order and the 802nd order can be 1 microampere, and the difference between the current values ​​of the 802nd order and the 803rd order can be 2 microamperes, but is not limited thereto. Furthermore, a specific order can be a specific current value. For example, the current value of the 994th order is a certain value such as 64 milliamperes, but is not limited thereto.

[0025] In step 303, an input data matrix is ​​established by obtaining multiple dimming levels, measured current value sets, and preliminary calibration current setting sets based on a portion of the light-emitting units 141. In other words, the input data matrix includes multiple dimming levels (0 to 255 levels) and 256 measured current values ​​and 256 preliminary calibration current setting values ​​corresponding to dimming levels 0 to 255 for a portion of the light-emitting units 141. It is worth noting that the dimming levels are not limited to 0 to 255 levels. The number of dimming levels can be increased or decreased according to actual needs. In addition, the relationship between the dimming level and the corresponding current value level is not limited to one-to-one. In certain cases, it can be one-to-many. For example, each dimming level can correspond to three current value levels (such as normal strong, normal, and normal weak). In throttling mode, it corresponds to the normal weak current value level, and in special development effect mode, it corresponds to the normal strong mode. The above description of the three current value levels is just an example and is not a limitation.

[0026] It is worth mentioning that the input data matrix also includes the index, position (LED position x & y), model (LED model), driver IC model, driver IC serial number, target current value, and default current setting value for each of the light-emitting units 141. The target current value corresponds to the dimming level, and the default current setting value corresponds to the target current value. In other words, the input data matrix includes multiple dimming levels (0 to 255) and 256 measured current values, preliminary calibration current settings, index, position, model, driver IC model, driver IC serial number, target current value, and default current setting value for each of the light-emitting units 141 at dimming levels from 0 to 255.

[0027] Figure 4 is a schematic diagram of step 304 of the method for establishing the correction current setting value according to the first embodiment of this disclosure. Referring to Figures 3 and 4 together, in step 304, this input data matrix x is input into the trained model M of the generative adversarial network to generate an output data matrix G(x). This output data matrix contains a set of optimal correction current settings corresponding to all light-emitting units 141 at various dimming levels. In other words, the output data matrix contains various dimming levels (0 to 255) and 256 optimal correction current settings corresponding to all light-emitting units 141 at dimming levels from 0 to 255. The trained model M in step 304 can be used to predict the optimal correction current setting value.

[0028] Specifically, the flowcharts in Figures 3 and 4 use the measurement results of a subset of the light-emitting units 141 to generate the optimal setting value of the correction current for all light-emitting units 141. This means that the generative adversarial network (GAN) is trained to complete the model M, using the data of a subset of the light-emitting units 141 (e.g., 20% of the total number of light-emitting units 141, but this disclosure is not limited to this) (i.e., the input data matrix) to predict the data of all light-emitting units 141 (i.e., the output data matrix, which contains a set of optimal setting values ​​of the correction current for each light-emitting unit 141 at various dimming levels). In the flowchart of Figure 3, the optimal setting value of the correction current for all light-emitting units 141 is obtained using the trained model M of the GAN. Therefore, the method described in the above embodiment can quickly set the current value of the light-emitting units 141, saving a significant amount of brightness calibration time.

[0029] Besides the above-described embodiment (first embodiment), a second embodiment of the method for establishing the calibration current setting value can also follow a similar process. Referring to Figure 5, the main difference between the second and first embodiments is that the second embodiment only drives specific light-emitting units 143 (i.e., a portion of the light-emitting units 144), instead of driving all light-emitting units 144. Details are as follows: multiple specific light-emitting units 143 among the multiple light-emitting units 144 included in the backlight module 142 are driven, and then the current of these specific light-emitting units 143 is measured at various dimming levels (0 to 255) to obtain a set of measured current values ​​corresponding to each specific light-emitting unit 143 at various dimming levels. In other words, the set of measured current values ​​consists of 256 measured current values ​​corresponding to the specific light-emitting units 143 at dimming levels from 0 to 255.

[0030] It should also be noted that the specific light-emitting units 143 circled in Figure 5 are examples of light-emitting units 144 located at the corners of the backlight module 142, but this is not a limitation; light-emitting units 144 located in the middle area or at the edge of the backlight module 142 can also be selected. It is worth noting that because specific light-emitting units 143 differ in production, manufacturing, and use, the luminance of different specific light-emitting units 143 will be different when driven by the same current value. In addition, the luminance of specific light-emitting units 143 can be divided into 256 levels, and the current value of each level of luminance of the same specific light-emitting unit 143 will be different, and the current value of different specific light-emitting units 143 at the same level of luminance will also be different.

[0031] After obtaining the set of measured current values ​​corresponding to each specific light-emitting unit 143 at various dimming levels, each specific light-emitting unit 143 is calibrated according to this set of measured current values ​​to generate a preliminary set of calibration current values ​​for each specific light-emitting unit 143.

[0032] Next, based on these specific light-emitting units 143, multiple dimming levels, sets of measured current values, and sets of preliminary calibration current settings are obtained to establish an input data matrix. In other words, the input data matrix contains multiple dimming levels (0 to 255 levels) and 256 measured current values ​​and 256 preliminary calibration current settings corresponding to these specific light-emitting units 143 at dimming levels from 0 to 255. It is worth noting that the number of dimming levels is not limited to 0 to 255 levels; the number of dimming levels can be increased or decreased according to actual needs. Furthermore, the number of dimming levels and the corresponding current value levels are not limited to a one-to-one relationship; in certain cases, a one-to-many relationship can be established. For example, each dimming level can correspond to three current value levels (such as generally strong, generally, and generally weak). In throttling mode, it corresponds to the generally weak current value level, and in special development effect mode, it corresponds to the generally strong mode. The above description of the three current value levels is merely an example and is not a limitation.

[0033] Finally (after establishing the input data matrix), please refer to Figure 5, and input the input data matrix x into the trained model M to generate the output data matrix G(x). The output data matrix contains the optimal set of correction current values ​​for each of these light-emitting units 144 at various dimming levels.

[0034] The most distinctive feature of the second embodiment disclosed herein is that the specific light-emitting unit 143 of the backlight module 142 includes light-emitting units 144 located at the edges, corners, arcs, etc. of the backlight module 142 and light-emitting units 144 located at the center of the backlight module 142. To ensure uniform luminance in the backlight module 142, the luminance provided by the light-emitting unit 144 at the center of the backlight module 142 can be slightly dimmer (i.e., the required current can be slightly smaller) than that provided by the specific light-emitting units 143 located at the edges, corners, and arcs of the backlight module 142. This is because the light-emitting unit 144 at the center of the backlight module 142 is surrounded by other light-emitting units 144 that can maintain the luminance level, while the specific light-emitting units 143 located at the edges, corners, and arcs of the backlight module 142 are not surrounded by enough other light-emitting units 144 to maintain the luminance level. Therefore, the luminance provided by the specific light-emitting units 143 located at the edges, corners, and arcs of the backlight module 142 must be brighter (i.e., the required current must be larger) than that provided by the light-emitting unit 144 at the center of the backlight module 142. Only in this way can the overall luminance of the backlight module 142 be uniform.

[0035] However, the selection of a specific light-emitting unit 143 can be any of the light-emitting units 144, the entire boundary, the entire arc, the entire periphery, any polygon within a specific range, any shape within a specific range, any arc within a specific range, or a selection in a specified manner, but is not limited thereto.

[0036] After the selected specific light-emitting unit 143 is established using the correction current setting method provided in this disclosure, the current values ​​that need to be set for all light-emitting units 144 (including the specific light-emitting unit 143) distributed at various positions on the backlight module 142 can be deduced. This allows the correction current setting value to be provided quickly, so that the backlight module 142 emitting light has uniform brightness.

[0037] It is worth noting that the specific light-emitting unit 143 disclosed herein may also refer to the light-emitting unit 144 that has not yet completed training or is currently being trained. The light-emitting unit 144 may also refer to the light-emitting unit 144 that can bring out the data of the light-emitting unit 144 that has been trained and completed training when it encounters similar environmental conditions in the future (such as being arranged in similar polygons, similar shapes, similar arcs or combinations thereof, or in similar arrangement densities), so that the data of the entire light-emitting unit 144 can be predicted by using only the specific light-emitting unit 143.

[0038] Figure 6 is a flowchart of the training phase of the method for establishing the correction current setting value according to the first embodiment of this disclosure. Alternatively, it can be a flowchart of the training phase of the correction current setting value according to the second embodiment. In step 501, multiple training backlight modules are obtained, each training backlight module including multiple training light-emitting units. Step 501 involves training the generator of the generative adversarial network using multiple (e.g., 5, but this disclosure is not limited to) different training backlight modules to obtain a trained model, thereby improving the prediction accuracy of the generative adversarial network. Furthermore, another advantage of this disclosure is that only a small amount of data is required for training (e.g., only 5 training backlight modules), saving time spent collecting large amounts of data for training.

[0039] In step 502, the multiple training light-emitting units included in each of these training backlight modules are driven, and then the current of each training light-emitting unit is measured at various dimming levels (0 to 255) to obtain a set of training measurement current values ​​corresponding to each training light-emitting unit at various dimming levels. In other words, the set of training measurement current values ​​consists of 256 training measurement current values ​​corresponding to the training light-emitting units at dimming levels from 0 to 255. The specific method of step 502 is similar to that of step 301, and will not be described again here.

[0040] In step 503, each training light-emitting unit is calibrated based on this set of training measurement current values ​​to generate a preliminary set of training calibration current values ​​for each training light-emitting unit. In other words, the preliminary set of training calibration current values ​​consists of 256 preliminary training calibration current values ​​corresponding to dimming levels from 0 to 255 for each training light-emitting unit. The specific implementation of step 503 is similar to that of step 302, and will not be described again here.

[0041] In step 504, based on these training light-emitting units, a real data matrix is ​​established by obtaining various dimming levels, training measurement current value sets, and preliminary training correction current setting sets. In other words, the real data matrix contains various dimming levels (0 to 255 levels) and 256 training measurement current values ​​and 256 preliminary training correction current setting values ​​corresponding to these training light-emitting units at dimming levels from 0 to 255. It is worth noting that the dimming level is not limited to 0 to 255 levels; the dimming level can be increased or decreased according to actual needs. Furthermore, the dimming level and the corresponding current value level are not limited to a one-to-one relationship; in specific cases, a one-to-many relationship can be established. For example, each dimming level can correspond to three current value levels (such as generally strong, generally, and generally weak). In throttling mode, it corresponds to the generally weak current value level, and in special development effect mode, it corresponds to the generally strong mode. The above description of the three current value levels is merely an example and is not a limitation.

[0042] Figure 7 is a schematic diagram based on steps 505 and 506. In step 505, the training input data matrix z is obtained based on the real data matrix R, and the training input data matrix z is input into the generator G to generate the training output data matrix G(z). In this embodiment, a portion of the real data matrix R is sampled as the training input data matrix z, or random sampling data is used to establish the training input data matrix z. For example, the real data matrix R is an 11×2400 data matrix (the above 2400 corresponds to 2400 training light-emitting units, and the above 11 corresponds to the following 11 parameters: light-emitting unit number, light-emitting unit X position, light-emitting unit Y position, light-emitting unit model, driving circuit model, driving circuit serial number, dimming level, target current value, current default setting value, measured current value and correction). The initial current setting is used to randomly sample 128 data points from the aforementioned 2400 data points to create an 11×128 data matrix, which serves as the training input data matrix z. This training input data matrix z is then input into the generator G of the generative adversarial network to generate the training output data matrix G(z). The number of data points in the training input data matrix z is less than the number of data points in the training output data matrix G(z), and the number of data points in the training output data matrix G(z) is equal to the number of data points in the real data matrix R. Both are 11×2400 data matrices. This training output data matrix G(z) contains a set of optimal training correction current settings for each training emitting unit at various dimming levels. In other words, the training output data matrix G(z) contains various dimming levels (0-255) and 256 optimal training correction current settings corresponding to dimming levels 0-255 for each training emitting unit. The generator G in step 505 is an untrained generator G of the generative adversarial network. Therefore, step 506 is required to train the generator G of the generative adversarial network. After training is completed, the trained model M of the generative adversarial network is obtained.

[0043] In step 506, the real data matrix R and the training output data matrix G(z) are input into the discriminator D of the generative adversarial network (GAN). Then, the loss function Loss is obtained based on the discrimination result of the discriminator D, and the generator G and discriminator D of the GAN are trained based on the loss function Loss. The loss function for optimizing the generator and discriminator is less than a loss threshold. The specific implementation of step 506 is detailed below.

[0044] In step 506, the real data matrix R and the training output data matrix G(z) are input into the discriminator D. The discriminator D compares the real data matrix R with the training output data matrix G(z) to generate a discrimination result. Then, based on the discrimination result (real or fake) of the discriminator D, the loss function Loss is obtained, and the generator G and discriminator D are trained based on the loss function Loss. In other words, the generator G and discriminator D are optimized based on the loss function Loss, and the neural network weights with the minimum loss value in the training cycle are saved as the optimal network model weight settings. In addition, the generator's initial weights are generated using a truncated normal distribution method, with the mean set to 0 and the standard deviation set to 0.02. The above values ​​are only illustrative and this disclosure is not limited to these.

[0045] The loss function, Loss, is the binary cross-entropy. The loss functions of the generator and discriminator are optimized to be less than the critical loss values. It's worth noting that the critical loss values ​​vary depending on the number of training iterations. The optimal (minimum) loss value during training must be less than a set value within the training iteration range. In other words, the critical loss values ​​are selected based on the number of training iterations. For example, with 10,000 training iterations, the critical loss value for the generator can be set to 1.5 and the critical loss value for the discriminator can be set to 0.8. This means that the minimum value of the generator's loss function should be less than 1.5 and the minimum value of the discriminator's loss function should be less than 0.8 for the results to be acceptable. These values ​​are merely illustrative and this disclosure is not limited to them.

[0046] The method of training the generator and discriminator based on the loss function is called error backpropagation. The concept of error backpropagation is to feed back the error of the loss function to the generator and discriminator, so that the generator and discriminator can use the information to perform gradient descent to update the weights of the generator and discriminator, thereby further reducing the error. When Nash equilibrium is reached or the discrimination result of all training input data matrix z is equal to 0.5, it means that the data generated by the generator is close to the real data, so training is terminated.

[0047] In short, there are two ways to obtain a fully trained model M. First, determine whether the discrimination result generated by the discriminator D falls within the acceptable range. If it does not fall within the acceptable range, repeat the process to generate another training output data matrix G(z) and generate another discrimination result until the discrimination result falls within the acceptable range. The generator G then completes the training based on the discrimination result to obtain the fully trained model M.

[0048] Secondly, the process of generating multiple training output data matrices G(z) is repeated to produce multiple discrimination results. When any of these discrimination results falls outside the acceptable range, the generator G completes training based on the smallest value among the discrimination results to obtain the trained model M. In other words, the trained model M is the generator that has completed training.

[0049] In addition, the present invention also proposes an electronic terminal to train a generator of a generative adversarial network, thereby obtaining a trained model of the generative adversarial network, and then using the trained model to generate the above-mentioned output data matrix, which is used to generate multiple correction current settings for multiple light-emitting units of a backlight module.

[0050] When the backlight module 142 (as shown in Figure 5, the backlight module 142 is, for example, an irregularly shaped light panel, but this disclosure is not limited to this) needs to be trained, for the specific light-emitting units 143 arranged at the edges, corners, arcs, etc. of the backlight module 142, due to the complex characteristics of the slope, angle, curvature, etc. at the edges, corners, arcs, etc., more training times and deeper learning are required. As for the non-specific light-emitting units (the light-emitting units 144 arranged at the center of the backlight module 142), it is not necessary to consider the complex characteristics of the slope, angle, curvature, etc. at the edges, corners, arcs, etc., and the light-emitting units 144 that are more linear and have lower simulation difficulty can be used to reduce the number of training times. In summary, during the training phase, these training luminescent units are divided into multiple training-specific luminescent units and multiple training-non-specific luminescent units. The training-specific luminescent units are deployed around the training-non-specific luminescent units. During the training phase, the proportion of the training-specific luminescent units in the training input data matrix is ​​greater than that of the training-non-specific luminescent units, so as to carry out deeper learning and training for edges, corners, arcs, etc., and finally generate a trained model.

[0051] When the backlight module 142 (as shown in Figure 5, the backlight module 142 is, for example, an irregularly shaped light panel, but this disclosure is not limited to this) is used to infer using the trained generative adversarial network model M, that is, to predict the optimal setting value of the correction current, since the proportion of the total area occupied by the edges, corners, arcs, etc. of the backlight module 142 is smaller than the proportion of the total area occupied by the center of the backlight module 142, in the method for establishing the setting value of the correction current in the second embodiment of this disclosure, an input data matrix can be established based on the various dimming levels of these non-specific light-emitting units and these specific light-emitting units 143, the measured current value group, and the preliminary setting value group of the correction current. When the input data matrix is ​​input into the trained model, the proportion of the specific light-emitting units 143 arranged at the edges, corners, arcs, etc. in the input data matrix is ​​designed to be smaller than the proportion of the non-specific light-emitting units arranged at the center, thereby conforming to the actual light panel shape. The output data matrix generated by the trained model can obtain the optimal setting value group of the correction current to achieve uniform brightness overall.

[0052] Therefore, in the deduction process, this disclosure can quickly predict the appropriate correction current setting value for each light-emitting unit at different positions on the light board. This allows the light board containing the light-emitting units to save a significant amount of time when measuring the current value of each light-emitting unit at different brightness levels and correcting the brightness of each light-emitting unit when producing different shapes or irregular shapes.

[0053] In summary, conventional methods require a lengthy time to measure the different current values ​​corresponding to each light-emitting unit at brightness levels 0-255, and then perform current correction before the entire backlight module can achieve uniform brightness. In contrast, the method for establishing the correction current setting value disclosed in this paper establishes a predictive model and trains it with a small but sufficient amount of current and brightness data. This method is not limited by the fundamental differences in the production, manufacturing, and use of light-emitting units, nor by the different positions and densities of the light-emitting units. Regardless of the shape, arc, or specific position of the light-emitting unit, the method disclosed in this paper can predict the appropriate correction current setting value for each light-emitting unit at different positions. This significantly improves the time efficiency of measuring current and correcting brightness for the light-emitting units, and also provides the benefit of achieving uniform brightness on backlight modules with different arrangement shapes or unit densities.

[0054] In other words, the application scenarios disclosed herein are not limited to situations where the arrangement of the light-emitting units is any polygon, arbitrary shape, random arc, or combination thereof, or when the arrangement density of the light-emitting units on the same lamp board is inconsistent. To achieve consistent luminous brightness on the lamp board, the current value driving each light-emitting unit will differ, and each light-emitting unit will have differences in production, manufacturing, and use. The method for establishing the calibration current setting value disclosed herein can quickly predict the appropriate calibration current setting value for each light-emitting unit at different positions on the lamp board. This significantly reduces the time spent measuring the current value of each light-emitting unit at different brightness levels and calibrating the luminous brightness of each unit when producing lamp boards with different shapes or irregular forms.

[0055] The foregoing has outlined the features of several embodiments, thus enabling those skilled in the art to better understand the nature of this disclosure. Those skilled in the art should recognize that this disclosure can be readily used as a basis to design or modify other processes and structures, thereby achieving the same objectives and / or advantages as the embodiments described herein. Those skilled in the art should also understand that these equivalent constructions do not depart from the spirit and scope of this disclosure, and that various changes, substitutions, and modifications can be made without departing from the spirit and scope of this disclosure.

[0056] The embodiments disclosed above are merely illustrative of the principles, features, and effects of the present invention and are not intended to limit the scope of implementation of the present invention. Any person skilled in the art can modify and alter the above embodiments without departing from the spirit and scope of the present invention. Any equivalent changes and modifications made using the content disclosed in this invention are still covered by the appended claims.

[0057]

List of reference numerals

Claims

1. A method for establishing a calibration current setting value, applicable to a backlight module, the backlight module comprising a plurality of light-emitting units, the plurality of light-emitting units being driven by current to emit light, the method comprising: Drive the plurality of light-emitting units and measure the current of the plurality of light-emitting units at various dimming levels to obtain a set of measured current values ​​for each of the plurality of light-emitting units at the various dimming levels; Each of the plurality of light-emitting units is calibrated according to the measured current value set to generate a preliminary set of calibration current values ​​for each of the plurality of light-emitting units; Based on a subset of the multiple light-emitting units, an input data matrix is ​​established by obtaining the various dimming levels, the measured current value set, and the preliminary setting value set of the correction current. as well as The input data matrix is ​​input into the trained model to generate an output data matrix, and the output data matrix contains a set of optimal correction current settings for each of the plurality of light-emitting units at the plurality of dimming levels.

2. The method according to claim 1 further includes a training phase for training the generator to obtain the trained model, the training phase including: A plurality of training backlight modules are obtained, wherein each of the plurality of training backlight modules includes a plurality of training light-emitting units; Drive the plurality of training light-emitting units and measure the current of the plurality of training light-emitting units at the various dimming levels to obtain a group of training measurement current values ​​corresponding to each of the plurality of training light-emitting units at the various dimming levels. Each of the plurality of training light-emitting units is calibrated according to the training measurement current value set to generate a preliminary set of training calibration current values ​​for each of the plurality of training light-emitting units. A real data matrix is ​​established based on the various dimming levels, the training measurement current value group, and the training correction current preliminary setting value group. A portion of the real data matrix is ​​sampled as the training input data matrix, and the training input data matrix is ​​input into the generator to generate the training output data matrix. The number of data entries in the training input data matrix is ​​less than the number of data entries in the training output data matrix, and the number of data entries in the training output data matrix is ​​equal to the number of data entries in the real data matrix. as well as The real data matrix and the training output data matrix are input into a discriminator, which compares the training output data matrix with the real data matrix to generate a discrimination result. The generator completes training based on the discrimination result to obtain the trained model.

3. The method for establishing according to claim 2, wherein, During the training phase, it is determined whether the discrimination result generated by the discriminator falls within the acceptable range. If it does not fall within the acceptable range, the process of sampling a portion of the real data matrix as the training input data matrix and inputting it into the generator to generate the training output data matrix is ​​repeated until the discrimination result falls within the acceptable range. The generator then completes training based on the discrimination result to obtain the trained model.

4. The method for establishing according to claim 2, wherein, During the training phase, a portion of the real data matrix is ​​sampled as the training input data matrix and input to the generator to produce the training output data matrix. The real data matrix and the training output data matrix are then input to the discriminator, which compares the training output data matrix with the real data matrix to generate the discrimination result. The above steps are performed multiple times to generate multiple discrimination results. When any of the multiple discrimination results does not fall within the acceptable range, the generator completes training based on the smallest value among the discrimination results to obtain the trained model.

5. The method for establishing according to claim 3 or 4 further includes: The loss function is obtained based on the discrimination result of the discriminator. as well as The generator and the discriminator are trained based on the loss function.

6. The method for establishing according to claim 5, wherein, During the training phase, the loss function of the generator and the discriminator is optimized to be less than a loss threshold.

7. The method for establishing according to claim 1, wherein, The input data matrix also includes the number, position, model, driving circuit model, driving circuit serial number, target current value and current preset setting value of each of the plurality of light-emitting units, wherein the target current value corresponds to the dimming level and the current preset setting value corresponds to the target current value.

8. An electronic terminal for obtaining the optimal set of correction current settings for driving the plurality of light-emitting units of the backlight module, according to the method for establishing correction current settings as described in any one of claims 1 to 7.

9. A method for establishing a calibration current setting value, applicable to a backlight module, the backlight module comprising a plurality of light-emitting units, the plurality of light-emitting units being driven by current to emit light, the method comprising: Drive multiple specific light-emitting units among the multiple light-emitting units, and measure the current of the multiple specific light-emitting units at multiple dimming levels to obtain a set of measured current values ​​corresponding to each of the multiple specific light-emitting units at the multiple dimming levels; The plurality of specific light-emitting units are calibrated according to the measured current value set to generate a preliminary set of calibration current values ​​for each of the plurality of specific light-emitting units; An input data matrix is ​​established based on the various dimming levels of the multiple specific light-emitting units, the measured current value group, and the preliminary setting value group of the correction current; as well as The input data matrix is ​​input into the trained model to generate an output data matrix, and the output data matrix contains a set of optimal correction current settings for each of the plurality of light-emitting units at the plurality of dimming levels.

10. The method of claim 9, further comprising a training phase for training the generator to obtain the trained model, wherein the training phase includes: A plurality of training backlight modules are obtained, wherein each of the plurality of training backlight modules includes a plurality of training light-emitting units; Drive the plurality of training light-emitting units and measure the current of the plurality of training light-emitting units at the various dimming levels to obtain a group of training measurement current values ​​corresponding to each of the plurality of training light-emitting units at the various dimming levels. Each of the plurality of training light-emitting units is calibrated according to the training measurement current value set to generate a preliminary set of training calibration current values ​​for each of the plurality of training light-emitting units. A real data matrix is ​​established based on the various dimming levels, the training measurement current value group, and the training correction current preliminary setting value group. A portion of the real data matrix is ​​sampled as the training input data matrix, and the training input data matrix is ​​input into the generator to generate the training output data matrix. The number of data entries in the training input data matrix is ​​less than the number of data entries in the training output data matrix, and the number of data entries in the training output data matrix is ​​equal to the number of data entries in the real data matrix. as well as The real data matrix and the training output data matrix are input into a discriminator, which compares the training output data matrix with the real data matrix to generate a discrimination result. The generator completes training based on the discrimination result to obtain the trained model.

11. The method for establishing according to claim 10, wherein, During the training phase, it is determined whether the discrimination result generated by the discriminator falls within the acceptable range. If it does not fall within the acceptable range, the process of sampling a portion of the real data matrix as the training input data matrix and inputting it into the generator to generate the training output data matrix is ​​repeated until the discrimination result falls within the acceptable range. The generator then completes training based on the discrimination result to obtain the trained model.

12. The method for establishing according to claim 10, wherein, During the training phase, a portion of the real data matrix is ​​sampled as the training input data matrix and input to the generator to produce the training output data matrix. The real data matrix and the training output data matrix are then input to the discriminator, which compares the training output data matrix with the real data matrix to generate the discrimination result. The above steps are performed multiple times to generate multiple discrimination results. When any of the multiple discrimination results does not fall within the acceptable range, the generator completes training based on the smallest value among the discrimination results to obtain the trained model.

13. The method for establishing according to claim 11 or 12 further includes: The loss function is obtained based on the discrimination result of the discriminator. as well as The generator and the discriminator are trained based on the loss function.

14. The method for establishing according to claim 13, wherein, During the training phase, the loss function of the generator and the discriminator is optimized to be less than a loss threshold.

15. The method for establishing according to claim 9, wherein, The input data matrix also includes the number, position, model, driving circuit model, driving circuit serial number, target current value and current preset setting value of each of the plurality of light-emitting units, wherein the target current value corresponds to the dimming level and the current preset setting value corresponds to the target current value.

16. The method for establishing according to claim 9, wherein, The plurality of specific light-emitting units are selected from at least two of the plurality of light-emitting units, and the shape of the plurality of light-emitting units arranged is any polygon, arbitrary shape, random arc or combination thereof.

17. The method for establishing according to claim 16, further comprising: If the shape of the plurality of light-emitting units arranged includes at least the arbitrary arc shape, then the curvature of the arbitrary arc shape is calculated; as well as When the curvature at a position among the plurality of light-emitting units is greater than a curvature threshold, at least one of the plurality of light-emitting units within the range of the position is defined as one of the plurality of specific light-emitting units.

18. The method for establishing according to claim 16, further comprising: If the shape in which the plurality of light-emitting units are arranged includes at least the arbitrary shape, then at least one of the plurality of light-emitting units within the endpoint range of each of the arbitrary shape is defined as one of the plurality of specific light-emitting units; and If the shape of the plurality of light-emitting units arranged includes at least any of the polygons, then at least one of the plurality of light-emitting units within the endpoint range or the included angle range of any polygon is defined as one of the plurality of specific light-emitting units.

19. The method of establishing according to claim 10, wherein, The plurality of training light-emitting units are arranged in any shape, any shape, any arc, or a combination thereof. The plurality of training light-emitting units are divided into a plurality of specific training light-emitting units and a plurality of non-specific training light-emitting units. The plurality of specific training light-emitting units are arranged around the plurality of non-specific training light-emitting units. During the training phase, in the training input data matrix, the proportion of the plurality of specific training light-emitting units is greater than the proportion of the plurality of non-specific training light-emitting units.

20. The method for establishing according to claim 19, wherein, The plurality of light-emitting units also include a plurality of non-specific light-emitting units. The input data matrix is ​​established based on the plurality of non-specific light-emitting units and the plurality of specific light-emitting units, the measured current value group and the preliminary setting value group of the correction current. When the input data matrix is ​​input into the trained model, the proportion of the plurality of specific light-emitting units in the input data matrix is ​​less than the proportion of the plurality of non-specific light-emitting units.

21. An electronic terminal for obtaining the optimal set of correction current settings for driving the plurality of light-emitting units of the backlight module, according to the method for establishing correction current settings as described in any one of claims 9 to 20.

22. A training method for calibration current setpoints, used to train a generator to complete training and obtain a trained model for generating multiple calibration current setpoints for multiple light-emitting units of a backlight module, the training method comprising: A plurality of training backlight modules are obtained, wherein each of the plurality of training backlight modules includes a plurality of training light-emitting units; Drive the plurality of training light-emitting units and measure the current of the plurality of training light-emitting units at various dimming levels to obtain a group of training measurement current values ​​corresponding to each of the plurality of training light-emitting units at the various dimming levels. Each of the plurality of training light-emitting units is calibrated according to the training measurement current value set to generate a preliminary set of training calibration current values ​​for each of the plurality of training light-emitting units. A real data matrix is ​​established based on the various dimming levels, the training measurement current value group, and the training correction current preliminary setting value group. A portion of the real data matrix is ​​sampled as the training input data matrix, and the training input data matrix is ​​input into the generator to generate the training output data matrix. The number of data entries in the training input data matrix is ​​less than the number of data entries in the training output data matrix, and the number of data entries in the training output data matrix is ​​equal to the number of data entries in the real data matrix. as well as The real data matrix and the training output data matrix are input into a discriminator, which compares the training output data matrix with the real data matrix to generate a discrimination result. The generator completes training based on the discrimination result to obtain the trained model.

23. The training method according to claim 22, further comprising: The discriminator determines whether the discrimination result generated by the discriminator falls within the acceptable range. If it does not fall within the acceptable range, the process of sampling a portion of the real data matrix as the training input data matrix and inputting it into the generator to generate the training output data matrix is ​​repeated until the discrimination result falls within the acceptable range. The generator then completes training based on the discrimination result to obtain the trained model.

24. The training method according to claim 22, further comprising: A portion of the real data matrix is ​​sampled as the training input data matrix and input to the generator to generate the training output data matrix. The real data matrix and the training output data matrix are then input to the discriminator, which compares the training output data matrix with the real data matrix to generate the discrimination result. The above steps are performed multiple times to generate multiple discrimination results. When each of the multiple discrimination results does not fall within the acceptable range, the generator completes training based on the smallest value among the discrimination results to obtain the trained model.

25. The training method according to claim 23 or 24, further comprising: The loss function is obtained based on the discrimination result of the discriminator. as well as The generator and the discriminator are trained based on the loss function.

26. The training method according to claim 25, wherein, The loss function of the generator and the discriminator is optimized to be less than the loss threshold.

27. The training method according to claim 22, wherein, The plurality of training light-emitting units are arranged in any shape, any shape, any arc, or a combination thereof. The plurality of training light-emitting units are divided into a plurality of specific training light-emitting units and a plurality of non-specific training light-emitting units. The plurality of specific training light-emitting units are arranged around the plurality of non-specific training light-emitting units. In the training input data matrix, the proportion of the plurality of specific training light-emitting units is greater than the proportion of the plurality of non-specific training light-emitting units.

28. An electronic terminal for training the generator to complete training and obtain the trained model according to the training method for the correction current setting value according to any one of claims 22 to 27, so as to generate the plurality of correction current setting values ​​for the plurality of light-emitting units of the backlight module.

Images