Pressure testing method for mobile phone screen

Abstract

The invention discloses a pressure testing method for a mobile phone screen, and relates to the field of mobile phone testing, and the method comprises the steps: collecting pressure data and coordinate data of different regions of the mobile phone screen; constructing a screen pressure signal sensitivity distribution diagram according to the pressure data and the coordinate data; dividing the mobile phone screen into a first area and a second area according to the sensitivity distribution diagram; constructing a convex optimization model according to the signal continuity problem when the pressure is moved from the first region to the second region; the convex optimization model is solved through an optimization algorithm, the optimal parameter configuration of second area pressure signal enhancement is obtained, and the optimal parameter configuration comprises the optimal sampling frequency # imgabs0 #, the optimal signal smoothing factor # imgabs1 # and the optimal historical memory weight # imgabs2 #. In order to solve the problem of poor continuity of cross-regional pressure signals of a mobile phone screen, a convex optimization model with clear constraint conditions is constructed according to the problem of continuity of the cross-regional signals, an optimization algorithm is used for solving, and smooth transition of the pressure signals from a high-SNR region to a low-SNR region is achieved.

Description

Technical Field

[0001] The present application relates to the field of mobile phone testing, and in particular to a pressure testing method for a mobile phone screen. Background Art

[0002] With the continuous advancement of smartphone interaction technology, under-screen pressure sensors, as a new form of human-computer interaction, have gradually been adopted in high-end smartphones. By sensing the amount of pressure applied by the user on the screen, under-screen pressure sensors enable a richer set of operational commands, such as light pressure to preview and heavy pressure to invoke a menu, providing users with a more intuitive and natural interactive experience. Currently, major mobile phone manufacturers have adopted pressure sensing technology as a key feature to differentiate their products, and it is widely used in various application scenarios such as game control, image editing, and text selection.

[0003] As users' demands for a more responsive pressure-sensitive experience continue to rise, higher requirements are being placed on the stability and consistency of screen pressure sensor performance. The advent of the full-screen era, in particular, requires pressure sensors to provide uniform and consistent pressure sensing across a wider range. However, current pressure sensing technology still faces numerous challenges in practical application due to the complexity of screen structures and manufacturing process limitations.

[0004] The existing under-screen pressure sensor technology has the following main problems:

[0005] First, under-display pressure sensors are affected by the screen's structure and material properties, resulting in significant differences in signal-to-noise ratio (SNR) values ​​across different areas. A typical smartphone screen typically consists of multiple layers, including a protective glass, touch layer, display layer, and sensor layer. Differences in mechanical structure in specific areas, such as the bezel, camera area, and button locations, lead to varying pressure signal transmission efficiencies, ultimately manifesting as significant differences in pressure sensing capabilities across different areas of the screen. Existing testing methods can typically only assess individual points or average performance, making it difficult to comprehensively evaluate and characterize the pressure sensing distribution characteristics across the entire screen.

[0006] Secondly, when the user moves the touch pressure from a high SNR area to a low SNR area during use, the pressure signal may mutate, be lost, or be misjudged due to the sudden change in signal quality. For example, when performing a pressure sliding operation, the user may experience inconsistent pressure feedback, resulting in interruption or erroneous operation, seriously affecting the user experience. Existing technologies mainly use simple signal amplification or fixed parameter filtering, which cannot effectively solve the problem of cross-region signal continuity.

[0007] Third, ensuring accurate detection and stability of pressure signals in areas with low SNR is a technical challenge. Especially when the SNR falls below a certain threshold, traditional signal processing methods often struggle to effectively extract valid signals, resulting in unreliable pressure manipulation in these areas. Existing technologies primarily address this by increasing hardware sensitivity or lowering the detection threshold, but this also increases false trigger rates and system complexity.

[0008] Existing solutions primarily include: 1) increasing the number and density of sensors, which significantly increases cost and power consumption; 2) using fixed-parameter signal filtering algorithms, which struggle to adapt to the differentiated needs of different regions; and 3) simply increasing signal gain in weak areas, which can easily introduce excessive noise. None of these approaches address the cross-region continuity issue of screen pressure signals at a systemic level.

[0009] Therefore, there is an urgent need for a testing method that can comprehensively evaluate the sensitivity distribution of screen pressure signals and effectively solve the problem of cross-region signal continuity, so as to improve the signal reliability in low SNR areas and ensure that users can obtain a consistent and smooth pressure interaction experience across the entire screen. Summary of the Invention

[0010] In response to the poor continuity of pressure signals across regions of mobile phone screens, this application provides a pressure testing method for mobile phone screens. By constructing the cross-region signal continuity problem into a convex optimization model with clear constraints and solving it using an optimization algorithm, a smooth transition of the pressure signal from the high SNR region to the low SNR region is achieved.

[0011] The purpose of this application is achieved through the following technical solutions.

[0012] The present application provides a pressure testing method for a mobile phone screen, comprising: collecting pressure data and coordinate data of different areas of the mobile phone screen; constructing a screen pressure signal sensitivity distribution diagram based on the pressure data and coordinate data, wherein the sensitivity distribution diagram includes the signal-to-noise ratio (SNR) value distribution of each area of ​​the screen, and comprising: applying a preset pressure to multiple test points on the mobile phone screen within a preset pressure range, recording the pressure response signal and the ambient noise signal of each test point; and calculating the mean value of the pressure response signal for each test point. and standard deviation , and the mean of the ambient noise signal and standard deviation According to the formula The signal-to-noise ratio (SNR) value for each test point is calculated and associated with the corresponding screen coordinate data (x, y) to generate a three-dimensional dataset (x, y, SNR). A bilinear interpolation algorithm is used to interpolate the SNR values ​​across the entire screen area, forming a complete screen pressure signal sensitivity distribution map. The bilinear interpolation algorithm, based on a weighted average of four adjacent points, ensures mathematical continuity and smoothness of the interpolation result. This continuous distribution model overcomes the limitations of traditional discrete point testing.

[0013] Pressure data refers to the raw measurement data collected by the under-display pressure sensor array when a preset force is applied at different locations on the screen. It includes the mapping between pressure values ​​P(x, y, t) and two-dimensional coordinates (x, y). The pressure response signal is the electrical signal sequence containing valid pressure information output by the sensor system when a preset pressure is applied to the test point. It reflects the effective response of the sensor to the external pressure stimulus. The time-varying signal S(t) containing valid pressure information is also referred to as the environmental noise signal. The environmental noise signal refers to the background interference signal that still exists in the sensor system when no external pressure is applied. This includes unwanted signals generated by various interference factors such as electronic noise, mechanical vibration, and temperature drift.

[0014] In particular, traditional pressure testing methods often only focus on the performance of a few representative points and cannot reflect the full-screen pressure response characteristics. This application establishes a complete SNR distribution map through systematic multi-point testing, expanding the "point" test to a "surface" characterization sensitivity distribution map to reveal the differences in pressure sensing capabilities in different areas of the screen, and intuitively demonstrates the spatial non-uniformity of sensor performance.

[0015] According to the sensitivity distribution diagram, the mobile phone screen is divided into a first area and a second area. The first area represents a strong sensitive area with an SNR value greater than a threshold T0, and the second area represents an SNR value less than or equal to the threshold. The weak-sensing area, including: According to the sensitivity distribution map, calculate the distribution statistics of the full-screen SNR value, including the average value and standard deviation According to the formula Determine the signal-to-noise ratio threshold, where k is the adjustment coefficient, which is set according to product quality requirements and is usually in the range of [0.5, 1.5]. Different batches and different models of equipment may have different basic SNR levels. The traditional fixed threshold may lead to division deviation. This application divides based on the statistical characteristics of the SNR distribution of the equipment itself, so that the division result reflects the actual performance status of the equipment. The area is marked as the first area, that is, the strong sensitive area; the SNR value is less than or equal to the threshold The area is marked as the second area, i.e. the weakly sensitive area; the area boundary is smoothed to avoid scattered area divisions and ensure the boundary continuity between the first area and the second area.

[0016] The signal continuity problem when the pressure moves from the first area to the second area is constructed as a convex optimization model, which includes:

[0017] Establish the objective function F: ,in, represents the pressure signal at time t, represents the pressure signal at coordinate (x, y), for The adjacent coordinate points of , α and β are weight coefficients; that is, the pressure signal continuity problem is transformed into a variational problem, and the signal changes in the time domain and space domain are considered at the same time. Constrains the temporal continuity of the signal; Constrains the spatial uniformity of the signal.

[0018] Create constraints:

[0019] Sampling frequency ; Among them, the sampling frequency refers to the number of times the pressure signal is sampled per unit time, which determines the time resolution of signal acquisition and is usually measured in Hertz (Hz).

[0020] Signal smoothing factor The signal smoothing factor controls the intensity of pressure signal filtering, determining the filter algorithm's cutoff frequency or smoothing window size, and thus affecting the degree of signal smoothing. An exponential decay function was used to achieve a negative correlation between the smoothing factor and SNR. From the perspective of signal processing theory, low SNR regions require stronger smoothing to suppress noise. Enhancing the smoothing effect in low SNR regions reduces the impact of random fluctuations on pressure perception, while weakening the smoothing effect in high SNR regions preserves the pressure signal's sensitivity and response speed.

[0021] Historical memory weight The historical memory weight controls the ratio of the current signal to the historical signal. It determines the degree of time domain signal smoothing and affects the temporal continuity of the pressure signal. Increasing the historical signal weight in low SNR regions reduces the impact of sudden changes and abnormal fluctuations. In high SNR regions, decreasing the historical signal weight improves the signal's real-time response.

[0022] in, is the initial sampling frequency; is the baseline signal-to-noise ratio; is the initial signal smoothing factor; is the initial historical memory weight; is the sampling frequency adjustment coefficient; Represents the signal smoothing factor attenuation coefficient; Indicates the historical memory weight adjustment coefficient; determines the parameter value range: , , ;in, and are the minimum and maximum values ​​of the sampling frequency adjustment coefficient, and are the minimum and maximum values ​​of the signal smoothing factor, respectively. and are the minimum and maximum values ​​of the historical memory weight respectively; among them, the parameter value range constraint defines a convex set, which, combined with the convex objective function, constitutes a standard convex optimization problem.

[0023] The convex optimization model is solved by using an optimization algorithm to obtain the optimal parameter configuration for the pressure signal enhancement in the second region, which includes the optimal sampling frequency. , optimal signal smoothing factor and optimal historical memory weight ,include:

[0024] (1) Convert the convex optimization model into an augmented Lagrangian function ,in, is the obstacle parameter, is the logarithmic barrier function of the constraint condition and initializes the barrier parameter , convergence threshold and the maximum number of iterations Max_iter; the barrier parameter μ represents the coefficient of the control constraint transformation strength, which is a positive number and the initial value It must be large enough to ensure that the initial solution of the interior point method satisfies all constraints and follows the formula Gradually decrease.

[0025] In particular, the augmented Lagrangian function is a new objective function that combines the original objective function and the constraints through the obstacle term, which is expressed as , which is used to transform the constrained optimization problem into a sequence of unconstrained optimization problems. L is composed of the original objective function F (time-space continuity objective) and the obstacle term Composition, obstacles follow The optimal solution gradually approaches the true optimal solution of the original problem as the value decreases. This achieves the equivalent transformation of complex constrained problems to sequential unconstrained problems, making high-dimensional constrained optimization computable.

[0026] In particular, the barrier parameter μ is a coefficient that controls the strength of the constraint transformation, determines the weight of the barrier term in the augmented Lagrangian function, and represents the degree of penalty for constraint violation. The value makes the solution far away from the constraint boundary, and small The value allows the solution to approach the constraint boundary, the initial value It needs to be large enough to ensure that the initial solution satisfies all constraints, and the gradual reduction of the barrier parameter forms a smooth path from the interior of the feasible region to the optimal solution.

[0027] Logarithmic barrier function of the constraints , the expression is: ,in, Represents the bound constraint transfer function of the sampling frequency f, so that f is within the specified range. within the scope; Represents the boundary constraint transfer function of the signal smoothing factor λ, so that λ is within the specified within the scope; Represents the boundary constraint conversion function of the historical memory weight ω, so that ω is within the specified within the range.

[0028] In particular, the logarithmic barrier function is a mathematical mapping that transforms hard constraints into soft penalties, which is expressed as , used to create constraint boundaries. It takes the form of negative logarithms, such as ; When the parameter approaches the constraint boundary, the function value increases sharply and tends to infinity; within the feasible domain, the function value is finite and changes slowly; the constraint of each parameter corresponds to a component function, which together constitute a complete barrier function; the negative logarithmic form ensures the continuity and differentiability of the function within the feasible domain.

[0029] (2) According to the SNR value interval distribution in the sensitivity distribution diagram, set the initial coefficient estimate , , ; Based on the initial coefficient estimates , , and constraints, calculate the initial sampling frequency , initial signal smoothing factor and initial history memory weight ; Using the test data from the first area to the second area, calculate the initial objective function value ;

[0030] (3) For fixed obstacle parameters , perform Newton iteration until local convergence, including:

[0031] Compute the augmented Lagrangian function The gradient ∇L and the Hessian matrix H, where the gradient ∇L represents the function L on the adjustment coefficient The first-order derivative vector of L, the Hessian matrix H represents the second-order derivative matrix of the function L with respect to the adjustment coefficient; in particular, the signal transition from the strong-sensing area to the weak-sensing area presents a highly nonlinear characteristic, and traditional first-order methods (such as gradient descent) cannot effectively capture this complex curvature change. The Newton method captures the second-order curvature information of the target function through the Hessian matrix H, and can accurately model the signal characteristic changes in the transition area. In addition, the three key parameters (sampling frequency, smoothing factor, and historical memory weight) have a complex interdependence in the SNR change area. The Newton method uses the cross-derivative terms in the Hessian matrix to , accurately capturing the coupling effects between parameters, ensuring coordinated optimization of the three parameters rather than simple independent adjustments. Finally, in weakly sensitive areas, optimal performance often requires parameters to be close to constraint boundaries (such as the maximum smoothing factor). The Newton method combined with a logarithmic barrier function provides precise parameter adjustment near constraint boundaries, avoiding boundary oscillation issues that are common with conventional methods.

[0032] According to the gradient ∇L and the Hessian matrix H, the search direction is calculated using Newton's method ,in, represents the direction of descent in parameter space; in particular, From the perspective of differential geometry, the optimal descent direction of the function at the current point is expressed. For convex functions, the positive definiteness of H ensures that d is a strict descent direction. The Newton direction implicitly incorporates information about both step size and direction, taking small steps in directions of sharp signal change (where the corresponding eigenvalues ​​of the Hessian matrix are large) and large steps in directions of gentle change. This adaptability is crucial for handling abrupt SNR changes from areas of strong to weak signal sensitivity. Furthermore, the Newton direction is essentially the conjugate direction of the local quadratic approximation of the objective function, enabling the algorithm to rapidly approach the optimal solution along the most efficient path, avoiding the zigzag search path that can occur in transition regions with traditional methods, thereby reducing the adverse effects of parameter fluctuations on signal continuity. Finally, the positive definiteness of the Hessian matrix in convex optimization problems ensures the validity of the descent direction, which is crucial for ensuring monotonic improvement in parameter iteration during pressure signal processing and preventing signal instability caused by parameter fluctuations.

[0033] Set the search step size , including: from the initial step size Initially, if the current step size does not satisfy: , where c is a small constant, usually 0.01, and the step size is reduced to ,in, is the reduction factor, usually 0.5, until a step size that meets the conditions is found; in particular, the backtracking strategy The decreasing mechanism ensures the gradualness of parameter adjustment. When the finger slides from the strong-sensing area to the weak-sensing area, this gradual adjustment prevents the pressure signal from jumping due to sudden parameter changes.

[0034] According to the search direction d and the search step length η, the adjustment coefficient is updated: , , , generate new iteration points;

[0035] According to the updated adjustment coefficient, calculate the augmented Lagrangian function L value and repeat the above iterative steps until the local convergence condition is met: gradient norm Less than the preset convergence threshold of the current iteration stage .

[0036] (4) After local convergence, according to the formula Update barrier parameters ,in, is the barrier parameter attenuation coefficient, and returns to step (3) to continue to perform Newton iteration until the barrier parameter μ is less than the preset threshold and the gradient norm , or reaches the maximum number of iterations Max_iter; in particular, fixed Internal iteration of values ​​and The decreasing outer iterations together achieve the central path tracking, each The value corresponds to a point on the center path, From large to small, the corresponding solution gradually approaches the boundary constraint from the inside of the feasible region. When μ→0, the solution converges to the optimal solution of the original problem. In addition, the double termination condition μ<threshold and This ensures the global convergence of the algorithm. Small enough to ensure that the solution is close to the optimal solution of the original problem, and small enough for the gradient norm to ensure that the solution is the local optimal point of the current augmented problem. Finally, this application only has 3 parameters The algorithm has remarkable computational efficiency. The dimension of the Hessian matrix is ​​only 3×3, and the inversion cost is low. The low-dimensional parameter space enables the interior point method to converge quickly. Usually, a high-precision solution can be achieved within 10 to 20 outer iterations.

[0037] (5) Extract the optimal coefficients from the final converged solution ; According to the optimal coefficient and constraints, calculate the optimal sampling frequency , optimal signal smoothing factor and optimal historical memory weight .

[0038] Processing the pressure signal of the second area using the optimal parameter configuration, including: , resample the pressure signal of the second area to obtain the resampled signal , where t represents the time point; according to the optimal signal smoothing factor Resampled signal Perform filtering to obtain the filtered signal ; According to the optimal historical memory weight , after filtering Perform time domain weighted fusion to obtain the processed pressure signal , ,in, represents the processed pressure signal at the current time t, represents the filtered signal at the current time t, Represents the processed pressure signal at the previous moment t-1; the pressure signal , substitute the objective function F and calculate the optimized objective function value ; Compare the optimized objective function values , and the objective function value calculated with the original signal ,when Less than , it indicates that the optimal parameter configuration is valid.

[0039] Compared with the existing technology, the advantages of this application are:

[0040] Mobile phone screen pressure sensors, influenced by the screen's structure and material properties, exhibit significant variations in SNR (signal-to-noise ratio) values ​​across different areas. This results in weak pressure sensing in some areas. When a user's touch pressure shifts from a high-SNR area to a low-SNR area, the pressure signal may mutate, be lost, or be misjudged, resulting in an inconsistent user experience. In areas with low SNR, how can we ensure accurate pressure signal detection and stability? In particular, how can we evaluate and improve signal detection accuracy when the SNR falls below a specific threshold?

[0041] Therefore, this application models the cross-region signal continuity problem as a convex optimization problem with dual constraints in both time and space. It transforms the constrained optimization into an unconstrained optimization problem using a barrier function, and efficiently solves it using gradients and Hessian matrices. This effectively addresses the problem of signal abrupt changes and loss when pressure moves from a high-sensitivity area to a low-sensitivity area, significantly improving signal reliability in low-SNR areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present application will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein:

[0043] Figure 1 This is an exemplary flow chart of a method for pressure testing a mobile phone screen according to some embodiments of the present application;

[0044] Figure 2is a schematic diagram of a test track according to some embodiments of the present application;

[0045] Figure 3 is a comparison diagram of the original signal and the optimized signal in the track 1 shown in some embodiments of the present application;

[0046] Figure 4 is a schematic diagram of a second test track according to some embodiments of the present application;

[0047] Figure 5 is a comparison diagram of the original signal and the optimized signal in the second track shown in some embodiments of the present application;

[0048] Figure 6 is a third schematic diagram of a test track according to some embodiments of the present application;

[0049] Figure 7 This is a comparison diagram of the original signal and the optimized signal in trajectory three shown in some embodiments of the present application. DETAILED DESCRIPTION

[0050] The method and system provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0051] like Figure 1 As shown, pressure data and coordinate data of different areas of the mobile phone screen are collected; based on the pressure data and coordinate data, a screen pressure signal sensitivity distribution map is constructed, and the sensitivity distribution map includes the signal-to-noise ratio (SNR) value distribution of each area of ​​the screen; the mobile phone screen is divided into a first area and a second area according to the sensitivity distribution map, the first area represents a strong sensitivity area with an SNR value greater than a threshold value T0, and the second area represents a weak sensitivity area with an SNR value less than or equal to the threshold value T0; the signal continuity problem when the pressure moves from the first area to the second area is constructed as a convex optimization model, and the convex optimization model includes: establishing an objective function F:

[0052] , where represents the pressure signal at time t, represents the pressure signal at coordinate (x, y), for The adjacent coordinate points , α and β are weight coefficients; the convex optimization model is solved by the optimization algorithm to obtain the optimal parameter configuration for the pressure signal enhancement in the second region, which includes the optimal sampling frequency , optimal signal smoothing factor and optimal historical memory weight .

[0053] This example tests a flagship smartphone equipped with an under-display pressure sensor. The phone features a 6.7-inch AMOLED screen with a resolution of 3200×1440 pixels and is equipped with four pressure sensors located at the four corners of the screen. The test environment temperature was controlled at 25±2°C, and the relative humidity was kept within 45%±5% to ensure consistent testing conditions.

[0054] The testing equipment includes a high-precision pressure-applying device (accuracy 0.01N), a data acquisition system (sampling rate up to 5kHz), a screen coordinate positioning system (accuracy 0.1mm), and a signal processing and analysis platform. The pressure-applying device uses a silicone indenter with a diameter of 8mm to simulate the real-life scenario of a human fingertip touching the screen.

[0055] To collect pressure and coordinate data from different areas of the phone screen, we designed a uniform 15×25 grid with 375 test points to comprehensively evaluate the screen's pressure sensing characteristics. Each test point was spaced approximately 4.5mm apart to ensure coverage of all screen areas, including the edges, camera area, and center. To facilitate coordinate data processing, the upper-left corner of the screen was defined as the coordinate origin (0, 0), and the lower-right corner as (1, 1). The coordinates of all test points were normalized to this range.

[0056] For each test point, perform the following pressure data acquisition process: Aim the pressure-applying device at the test point to ensure that the center of the pressure head accurately coincides with the test point. Collect environmental noise signals: Without applying any pressure, record the sensor output signal for 5 seconds, set the sampling frequency to 200Hz, and obtain a total of 1000 sample points as environmental noise data. Collect pressure response signals: Apply five different levels of standard pressure of 0.5N, 1.0N, 1.5N, 2.0N, and 2.5N in turn. Each pressure level is maintained for 3 seconds. The sampling frequency is also 200Hz, and 600 sample points are obtained for each pressure level. To ensure the stability of the test, each test point is repeatedly collected 3 times, and the average value is taken as the final result.

[0057] The collected raw data undergoes preliminary preprocessing: The voltage signal is converted into pressure values ​​in Newtons (N) using a pre-calibrated pressure-voltage conversion curve. Outliers are removed: A 3σ criterion is used to identify and remove anomalous data points to ensure data quality. The signal is accurately segmented into an ambient noise segment and five pressure response segments at different pressure levels based on timestamps. In this example, data from four representative test points were randomly selected for the test phone with ID SM-N9870, as shown in Table 1:

[0058] Table 1: Pressure data of representative test points (unit: N)

[0059] Test point coordinates Average ambient noise Standard deviation of ambient noise 1.0N pressure response average 1.0N pressure response standard deviation (0.25,0.25) 0.023 0.008 0.982 0.035 (0.75,0.25) 0.021 0.009 0.975 0.042 (0.25,0.75) 0.025 0.015 0.968 0.048 (0.75,0.75) 0.029 0.021 0.952 0.063

[0060] From the data in Table 1, it can be observed that even when the same standard pressure is applied, there are differences in the pressure response at different positions of the screen. In particular, the signal fluctuation in the lower right corner (0.75, 0.75) is large, and the standard deviation is significantly higher than that in the upper left corner, which preliminarily reflects the unevenness of the screen pressure sensing.

[0061] Construct the screen pressure signal sensitivity distribution map. For the data of each test point, we use the formula Calculate the signal-to-noise ratio: is the mean value of the pressure response signal: for each pressure level, the average value of 600 sample points is calculated. In this embodiment, we select 1.0N pressure response as the standard evaluation pressure. The standard deviation of the pressure response signal is calculated from the 600 sample points at the same pressure level. The mean of the ambient noise signal is calculated by averaging 1000 ambient noise sample points. The standard deviation of the ambient noise signal is calculated by taking 1000 ambient noise sample points.

[0062] Taking the position (0.25, 0.25) in Table 1 as an example, the calculation process is as follows: Similarly, calculate the SNR values ​​at other positions: (0.75, 0.25) position: ; (0.25, 0.75) position: ; (0.75, 0.75) position: ; From the calculation results of these four representative points, it can be seen that the SNR value of the upper left area of ​​the screen is significantly higher than that of the lower right area, reflecting the difference in pressure sensing capabilities in different areas.

[0063] Following the above method, the SNR values ​​were calculated for all 375 test points. The results showed that the SNR values ​​for the entire screen ranged from 8.5 to 29.7, with significant differences. Combining the test point coordinate data, a three-dimensional dataset (x, y, SNR) was generated. Some examples of these data points are shown in Table 2:

[0064] Table 2: 3D dataset of some test points (x, y, SNR)

[0065] Serial number Normalized x-coordinate Normalized y coordinate SNR value 1 0.067 0.04 23.56 2 0.133 0.04 25.23 ... ... ... ... 374 0.933 0.96 12.35 375 1 0.96 9.42

[0066] Due to the limited number of test points, in order to obtain the SNR value estimate for each pixel position on the screen, the discrete SNR data needs to be interpolated. This embodiment uses a bilinear interpolation algorithm to scale the screen resolution to 320×640 grid points and calculate the SNR value for each grid point.

[0067] For any point in the grid , find the four nearest test points surrounding the point 、 、 、 . ,in, , Through bilinear interpolation, we obtained a high-resolution SNR distribution map covering the entire screen. Analysis of this distribution map reveals that the SNR values ​​in the upper left corner of the screen are generally higher, averaging around 25; while the SNR values ​​in the lower right corner are lower, averaging around 12; and the SNR values ​​in the four corners are the lowest, all below 10. This distribution characteristic is related to the layout of the sensor below the screen and the rigidity of the screen structure.

[0068] Screen area division and boundary smoothing, according to the sensitivity distribution map, calculate the statistical characteristics of the full screen SNR value: average value , standard deviation . Select the adjustment coefficient k=1.0, according to the formula Calculate the threshold: ,Therefore, the SNR value of 13.33 is used as the threshold for dividing the strong-sensitive area and the weak-sensitive area.

[0069] According to the threshold The interpolated sensitivity distribution map is then divided into regions: grid points with SNR > 13.33 are marked as 1, indicating a highly sensitive area (the first region); grid points with SNR ≤ 13.33 are marked as 0, indicating a weakly sensitive area (the second region). This initial division results in a binary image B(x, y), where approximately 68% of the screen area is highly sensitive and 32% is weakly sensitive. The weakly sensitive area is primarily distributed around the edges of the screen, particularly in the lower right corner.

[0070] Preferably, if the initially delineated boundary contains jagged or scattered irregular areas, smoothing is required. A Gaussian filter with a radius of 12 pixels is applied to blur the binary image B(x, y) to produce a blurred boundary image G(x, y). The standard deviation of the Gaussian filter is set to 5, and the kernel size is 25×25. The blurred image G(x, y) is then subjected to secondary segmentation using a threshold of 0.5, resulting in a smoothed region segmentation image S(x, y). Morphological operations are then applied to further optimize the region segmentation: an opening operation (erosion followed by dilation) is performed to eliminate isolated, strongly sensitive regions smaller than 200 pixels; a closing operation (dilation followed by erosion) is performed to fill isolated, less sensitive regions smaller than 200 pixels. A contour tracing algorithm is used to extract the region boundaries, resulting in three primary boundary curves. Each boundary curve is fitted with a B-spline curve, with the number of control points set to 1 / 10 of the number of boundary points, to generate smooth, continuous region demarcation lines. After smoothing, the boundary between the strong-sensitive area and the weak-sensitive area becomes continuous and smooth, eliminating the scattered points and jagged edges in the initial division, laying the foundation for subsequent convex optimization processing.

[0071] Construct a convex optimization model. In the experiment, through preliminary tests and empirical analysis, the following parameters are set: time continuity weight , spatial continuity weight . Initial sampling frequency , baseline signal-to-noise ratio , initial signal smoothing factor , initial history memory weight . Sampling frequency range: , , signal smoothing factor range: , , historical memory weight range: , .

[0072] The objective function is defined as the weighted sum of temporal continuity and spatial continuity: . Time domain continuity term Indicates the difference between pressure signals at adjacent moments. The smaller this term is, the better the temporal continuity of the pressure signal. In the actual calculation, we select a typical trajectory moving from the first area to the second area, which contains pressure data at 50 consecutive time points. Spatial domain continuity term Represents the difference in pressure signals between adjacent spatial locations. A smaller value indicates better spatial continuity of the pressure signal. For each location (x, y), consider the four adjacent points above, below, left, and right, and calculate the sum of the absolute values ​​of the pressure difference.

[0073] Based on the need to enhance the signal in weakly sensitive areas, constraint equations for three key parameters are designed to enable adaptive adjustment of the parameters as the SNR value changes:

[0074] Sampling frequency constraints: ,As the SNR value decreases, the sampling frequency increases, and the signal acquisition density is improved, is the sampling frequency adjustment coefficient, which controls the sensitivity of the sampling frequency to SNR changes. Signal smoothing factor constraint: ,As the SNR value decreases, the smoothing factor increases, enhancing the filtering effect, is the signal smoothing factor attenuation coefficient, which controls the rate at which the smoothing factor changes with SNR. History memory weight constraint: ,As the SNR value decreases, the historical memory weight increases, enhancing the temporal continuity, is the historical memory weight adjustment coefficient, which controls the degree to which the historical memory weight changes with SNR. These three constraints work together to enable the signal processing parameters to be adaptively adjusted according to the regional SNR value, improving the signal quality in weakly sensitive areas.

[0075] Solve the convex optimization model and convert it into an augmented Lagrangian function: . Initialization parameters: initial value of obstacle parameter , barrier parameter attenuation coefficient , convergence threshold , the maximum number of iterations Max_iter = 100, line search small constant , step size reduction factor According to the SNR value distribution in the sensitivity distribution graph, set the initial coefficient estimate: (initial value of sampling frequency adjustment coefficient), (Initial value of signal smoothing factor attenuation coefficient), (Initial value of historical memory weight adjustment coefficient).

[0076] Calculate the initial parameter configuration: For SNR=10 (typical weak sensitivity area): Initial sampling frequency: , initial signal smoothing factor: , initial history memory weight: , use the test data to calculate the initial objective function value .

[0077] For a fixed obstacle parameter μ, perform the Newton iterative solution process: calculate the gradient and Hessian matrix of the augmented Lagrangian function: Gradient , the Hessian matrix H is the corresponding second-order derivative matrix.

[0078] Detailed process of the first iteration: Calculate the gradient: , calculate the Hessian matrix: .

[0079] Calculate the search direction: ; Line search determines the step size: .

[0080] Update coefficient: ; ; ; Calculate the updated function value: L=11.23; Continue iterating until convergence, the convergence criterion is the gradient norm : After the fifth iteration, the gradient norm drops to 0.09, which is lower than the preset convergence threshold of 0.1 for the current iteration stage. The coefficient value is: , , .

[0081] After local convergence, update the barrier parameters: . Return to perform Newton iteration to obtain a new local optimal solution: After 4 iterations, the coefficient value is obtained: , , . Continue updating the barrier parameters and iterating: , , .

[0082] When the barrier parameter (preset threshold) and gradient norm When -6, the algorithm converges as a whole and obtains the final optimal coefficient: (optimal value of sampling frequency adjustment coefficient), (Optimal value of signal smoothing factor attenuation coefficient), (Optimal value of historical memory weight adjustment coefficient).

[0083] Based on the optimal coefficients and constraints, calculate the optimal parameter configuration for different SNR values:

[0084] When SNR=8 (extremely weak sense area): optimal sampling frequency: ;Optimal signal smoothing factor: ;Optimal historical memory weight: .

[0085] When SNR=10 (typical weak-sensing area): optimal sampling frequency: ;Optimal signal smoothing factor: ;Optimal historical memory weight: .

[0086] When SNR=13 (near the threshold): optimal sampling frequency: ;Optimal signal smoothing factor: ;Optimal historical memory weight: .

[0087] The pressure signal of the second area (weakly sensitive area) is processed: the corresponding optimal parameter configuration is calculated according to the SNR value of each point; the original pressure signal is resampled, filtered and time-domain weighted fusion processed; the resampling method is based on cubic spline interpolation; the filtering process uses a Butterworth low-pass filter, and the cutoff frequency is based on Dynamic adjustment; time domain weighted fusion uses a recursive formula: . Objective function value before processing ; The objective function value after processing ; Optimization effect: (12.56-5.83) / 12.56=53.6% improvement rate.

[0088] To verify the effectiveness of this method, three typical sliding trajectories from a strong-sensing area to a weak-sensing area were selected for testing:

[0089] Track 1: Figure 2 As shown, from the center to the lower right corner, the tester starts from the center of the screen (0.5, 0.5), uniformly applies 1.5N of pressure, and slides to the lower right corner (0.9, 0.9). Record the original pressure signal and the pressure signal after optimization. Figure 3 As shown, the chart contains three key curves: the red curve represents the original pressure signal, which gradually decreases from 1.48N at the starting point to 0.86N at the end point; the blue curve represents the optimized pressure signal, which remains stable between 1.43N and 1.52N throughout the entire trajectory; the gray dotted line represents the changing trend of the SNR value along the trajectory, which gradually decreases from the high value in the center to the low value in the lower right corner. The comparative analysis results show that: in the process of the SNR value of the original signal decreasing from 20 to 9, the pressure value gradually decreased from 1.48N to 0.86N, resulting in a significant decrease in pressure perception. On the same trajectory, the pressure value of the optimized signal remains stable between 1.43N and 1.52N, with a maximum deviation of only 0.09N. The signal volatility (standard deviation) is reduced from the original 0.21N to 0.08N, which improves the stability of the signal.

[0090] Track 2: Figure 4 As shown, from the upper left corner to the upper right corner, the tester slides along the upper edge of the screen from the upper left corner (0.1, 0.1) to the upper right corner (0.9, 0.1), applying a constant pressure of 2.0N. Special structural areas such as the camera, earpiece, and buttons are often "blind spots" of the under-screen pressure sensor, and the original signal is easily lost in these areas. Figure 5The figure shows a mobile phone screen with a camera notch, with the low SNR area marked with a red dashed circle. The test trajectory extends horizontally from the upper left corner (0.1, 0.1) to the upper right corner (0.9, 0.1), passing through the camera area. The figure clearly marks the SNR distribution of each area. The SNR in the camera area is only 7, far lower than the starting point (SNR=18) and end point (SNR=12) of the trajectory. The results show that the original signal experiences significant signal loss when passing through the camera area in the upper right corner of the screen, with the pressure value dropping to a minimum of 0.7N. The optimized signal maintains pressure values ​​above 1.85N in this area, effectively avoiding signal loss. The average signal continuity index (the average of the pressure difference between adjacent points) improves from 0.18N to 0.06N.

[0091] Track 3: Figure 6 As shown in the figure, the tester slides along the complex Z-shaped trajectory from the upper left to the lower right, passing through multiple areas with obvious SNR changes. Figure 7 As shown, the Z-shaped trajectory starts from the upper left corner, passes through the upper right corner, the lower left corner, and finally reaches the lower right corner. Five key SNR areas are clearly marked on the screen: the central high SNR area (22), two medium SNR areas (14-15), and two low SNR areas (7-8). The red points A, B, C, and D represent key signal transition points, which are located at the junction of different SNR areas. The optimized blue curve shows an ideal smooth state without any obvious breakpoints or mutations, and remains stable even in complex environments with an SNR value span of 15 units (from 7 to 22). This global smoothness ensures that users can obtain consistent pressure feedback when performing complex gesture operations. The results show that the original signal has multiple obvious step changes in the SNR change area, with the maximum step amplitude reaching 0.8N. The optimized signal maintains a smooth transition throughout the entire process, and the maximum step amplitude is reduced to 0.15N. The temporal domain continuity is improved by 62.7%, and the spatial domain continuity is improved by 58.3%.

Claims

1. A pressure testing method for a mobile phone screen, characterized in that: include: Collect pressure data and coordinate data of different areas of the mobile phone screen; Constructing a screen pressure signal sensitivity distribution map based on the pressure data and the coordinate data, wherein the sensitivity distribution map includes a signal-to-noise ratio (SNR) value distribution of each area of ​​the screen; Divide the mobile phone screen into a first area and a second area according to the sensitivity distribution map, wherein the first area represents a strong sensitivity area with an SNR value greater than a threshold T0, and the second area represents a weak sensitivity area with an SNR value less than or equal to the threshold T0; The signal continuity problem when the pressure moves from the first area to the second area is constructed as a convex optimization model, which includes: Establish the objective function F: , where represents the pressure signal at time t, represents the pressure signal at coordinate (x, y), for The adjacent coordinate points , α and β are weight coefficients; The convex optimization model is solved by using an optimization algorithm to obtain the optimal parameter configuration for the pressure signal enhancement in the second region, which includes the optimal sampling frequency. , optimal signal smoothing factor and optimal historical memory weight .

2. The pressure testing method for a mobile phone screen according to claim 1, characterized in that: Convex optimization models also include: Create constraints: Sampling frequency ; Signal smoothing factor ; Historical memory weight ; in, is the initial sampling frequency; is the baseline signal-to-noise ratio; is the initial signal smoothing factor; is the initial historical memory weight; is the sampling frequency adjustment coefficient; Represents the signal smoothing factor attenuation coefficient; Indicates the historical memory weight adjustment coefficient; Determine the parameter value range: , , ;in, and are the minimum and maximum values ​​of the sampling frequency adjustment coefficient, and are the minimum and maximum values ​​of the signal smoothing factor, respectively. and are the minimum and maximum values ​​of historical memory weight respectively.

3. The pressure testing method for a mobile phone screen according to claim 2, characterized in that: The convex optimization model is solved using an optimization algorithm to obtain the optimal parameter configuration for pressure signal enhancement in the second region, including: (1) Convert the convex optimization model into an augmented Lagrangian function ,in, is the obstacle parameter, is the logarithmic barrier function of the constraint condition and initializes the barrier parameter , convergence threshold and the maximum number of iterations Max_iter; (2) According to the SNR value interval distribution in the sensitivity distribution diagram, set the initial coefficient estimate , , ; Based on the initial coefficient estimates , , and constraints, calculate the initial sampling frequency , initial signal smoothing factor and initial history memory weight ; Using the test data from the first region to the second region, calculate the initial objective function value ; (3) For fixed obstacle parameters , perform Newton iteration until local convergence; (4) After local convergence, according to the formula Update barrier parameters ,in, is the barrier parameter attenuation coefficient, and returns to step (3) to continue the Newton iteration until the barrier parameter Less than the preset threshold and the gradient norm , or the maximum number of iterations Max_iter is reached; (5) Extract the optimal coefficients from the final converged solution ; According to the optimal coefficient and constraints, calculate the optimal sampling frequency , optimal signal smoothing factor and optimal historical memory weight ; The pressure signal of the second region is processed using the optimal parameter configuration.

4. The pressure testing method for a mobile phone screen according to claim 3, characterized in that: For fixed obstacle parameters , perform Newton iteration until local convergence, including: Compute the augmented Lagrangian function The gradient ∇L and the Hessian matrix H, where the gradient ∇L represents the function L on the adjustment coefficient The first-order derivative vector of , the Hessian matrix H represents the second-order derivative matrix of the function L with respect to the adjustment coefficient; According to the gradient ∇L and the Hessian matrix H, the search direction is calculated using Newton's method ,in, represents the descent direction in parameter space; Set the search step size ; According to the search direction d and search step size , update the adjustment coefficient: , , , generate new iteration points; According to the updated adjustment coefficient, calculate the augmented Lagrangian function L value and repeat the above iterative steps until the local convergence condition is met: gradient norm Less than the preset convergence threshold of the current iteration stage .

5. The pressure testing method for a mobile phone screen according to claim 4, characterized in that: Set the search step size ,include: From the initial step Initially, if the current step size does not satisfy: , where c is a small constant, usually 0.01, and the step size is reduced to ,in, is the reduction factor, usually 0.5, until a step value that meets the conditions is found; T represents transpose.

6. The pressure testing method for a mobile phone screen according to claim 3, characterized in that: Logarithmic barrier function of the constraints , the expression is: ； in, Represents the bound constraint transfer function of the sampling frequency f, so that f is within the specified range. within the scope; Represents the boundary constraint transfer function of the signal smoothing factor λ, so that λ is within the specified within the scope; Represents the boundary constraint conversion function of the historical memory weight ω, so that ω is within the specified within the range.

7. The pressure testing method for a mobile phone screen according to claim 3, characterized in that: The pressure signal of the second area is processed using the optimal parameter configuration, including: According to the optimal sampling frequency , resample the pressure signal of the second area to obtain the resampled signal , where t represents the time point; According to the optimal signal smoothing factor Resampled signal Perform filtering to obtain the filtered signal ; According to the optimal historical memory weight , after filtering Perform time domain weighted fusion to obtain the processed pressure signal ; The pressure signal , substitute the objective function F and calculate the optimized objective function value ; Compare the optimized objective function values , and the objective function value calculated with the original signal ,when Less than , it indicates that the optimal parameter configuration is valid.

8. The pressure testing method for a mobile phone screen according to claim 7, characterized in that: Get the processed pressure signal , through the following formula: ； in, represents the processed pressure signal at the current time t, represents the filtered signal at the current time t, Represents the processed pressure signal at the previous time t-1.

9. The pressure testing method for a mobile phone screen according to any one of claims 2 to 8, characterized in that: Construct a screen pressure signal sensitivity distribution map, including: Collect pressure response signals and environmental noise signals at multiple test points within a preset pressure range; Calculate the SNR value of each test point based on the pressure response signal and the environmental noise signal; According to the SNR value and the coordinates of the test points, a sensitivity distribution map is constructed.

10. The pressure testing method for a mobile phone screen according to claim 9, characterized in that: Threshold , using the following formula: ,in, Indicates the average value of the full-screen signal-to-noise ratio SNR value. represents the standard deviation of the full-screen signal-to-noise ratio (SNR), and k represents the threshold adjustment coefficient.

Images