Touch screen pressure sensing method, device, equipment and storage medium

Abstract

The present invention relates to a touch screen pressure sensing method, device, equipment and storage medium. The method: based on the first touch screen response parameter, the capacitance signal data of the touch screen is collected, and the initial feature vector is calculated; according to the initial feature vector, a pressure mapping model is established, and pressure distribution data is generated; feature extraction and fuzzy quantization are performed to obtain a fused feature vector; the fused feature vector is input into a Gaussian mixture model to calculate a dynamic pressure threshold parameter; the dynamic pressure threshold parameter is calculated for the rate of change, acceleration and degree of jitter to obtain a time domain feature sequence; the time domain feature sequence is comprehensively calculated with the pressure value and the contact area to obtain a pressure influence factor, and the first touch screen response parameter is adjusted according to the pressure influence factor, and the second touch screen response parameter is output. The implementation of the present invention improves the continuity and stability of pressure detection, and finally achieves a more accurate and smooth touch experience.

Description

Technical Field

[0001] The present invention relates to the field of touch screen technology, and in particular to a touch screen pressure sensing method, device, equipment and storage medium. Background Art

[0002] With the rapid development and widespread application of touch screen technology, users' requirements for touch screen interactive experience are constantly increasing, especially the demand for pressure perception is becoming increasingly prominent. Traditional touch screens mainly realize touch positioning based on the principle of capacitive sensing, but there are still obvious deficiencies in pressure detection accuracy, dynamic response and user experience.

[0003] At present, the touch screen pressure sensing solutions on the market mainly rely on a single pressure mapping model and fixed pressure threshold parameters, which cannot effectively adapt to the touch habits and dynamic touch scenarios of different users. At the same time, the existing pressure detection algorithms often ignore the timing characteristics and pressure change trends during the touch process, resulting in insufficient continuity and stability of the pressure response, affecting the user's operating experience. In addition, when dealing with multi-point touch pressure detection, the existing technology often encounters signal interference and crosstalk, the pressure feature extraction is not accurate enough, and the feature fusion method is relatively simple, making it difficult to accurately capture subtle changes in touch pressure. Summary of the invention

[0004] The main purpose of the present invention is to provide a touch screen pressure sensing method, device, equipment and storage medium. The present invention improves the continuity and stability of pressure detection, and ultimately achieves a more accurate and smooth touch experience.

[0005] To achieve the above object, the present invention provides a touch screen pressure sensing method, comprising the following steps:

[0006] Collecting capacitance signal data of the touch screen based on the first touch screen response parameter, and calculating an initial feature vector;

[0007] Establishing a pressure mapping model according to the initial eigenvector and generating pressure distribution data;

[0008] Performing feature extraction and fuzzy quantization on the pressure distribution data, and performing feature fusion to obtain a fused feature vector;

[0009] The fused feature vector is input into a Gaussian mixture model, and a dynamic pressure threshold parameter is calculated by an expectation maximization algorithm;

[0010] Calculating the rate of change, acceleration and degree of jitter of the dynamic pressure threshold parameter to obtain a time domain feature sequence;

[0011] The time domain feature sequence is comprehensively calculated with the pressure value and the contact area to obtain a pressure influence factor, and the first touch screen response parameter is adjusted according to the pressure influence factor to output a second touch screen response parameter.

[0012] The present invention also provides a touch screen pressure sensing device, comprising:

[0013] An acquisition module, used for acquiring capacitance signal data of the touch screen based on the first touch screen response parameter, and calculating an initial feature vector;

[0014] An establishing module, used to establish a pressure mapping model according to the initial characteristic vector and generate pressure distribution data;

[0015] An extraction module, used for performing feature extraction and fuzzy quantization on the pressure distribution data, and performing feature fusion to obtain a fused feature vector;

[0016] A processing module, used for inputting the fused feature vector into a Gaussian mixture model, and calculating a dynamic pressure threshold parameter by an expectation maximization algorithm;

[0017] A calculation module, used to calculate the change rate, acceleration and jitter degree of the dynamic pressure threshold parameter to obtain a time domain feature sequence;

[0018] The adjustment module is used to perform comprehensive calculation on the time domain feature sequence, the pressure value and the contact area to obtain a pressure influence factor, and adjust the first touch screen response parameter according to the pressure influence factor to output a second touch screen response parameter.

[0019] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0020] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0021] In summary, the technical solution provided by the present invention has the following beneficial effects by establishing a dual-mode pressure mapping model, a multi-feature fusion mechanism and a dynamic parameter adjustment strategy: the accuracy of pressure detection is significantly improved by adopting a segmented mapping strategy and a dynamic threshold adjustment mechanism; the robustness of pressure perception is enhanced by multi-dimensional feature extraction and hierarchical feature fusion; based on time domain feature analysis and pressure influencing factor calculation, adaptive adjustment of touch parameters is realized, and the dynamic response performance of the system is improved; multi-scale time window analysis and double sliding average filtering are adopted to effectively suppress signal noise and interference; through the time series correlation analysis and nonlinear combination of pressure features, the continuity and stability of pressure detection are improved, and ultimately a more accurate and smooth touch experience is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a schematic diagram of the steps of a touch screen pressure sensing method in one embodiment of the present invention;

[0023] Figure 2 is a structural block diagram of a touch screen pressure sensing device in one embodiment of the present invention;

[0024] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0025] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0027] Reference Figure 1 , this embodiment provides a touch screen pressure sensing method, comprising the following steps:

[0028] S1, collecting capacitance signal data of the touch screen based on the first touch screen response parameter, and calculating an initial feature vector;

[0029] Among them, the response parameters of the touch screen are analyzed to extract the touch sensitivity parameters, pressure response speed parameters and signal response mode parameters. The touch sensitivity parameter determines the response sensitivity of the touch screen to different pressures and directly affects the selection of the sampling frequency; the pressure response speed parameter is an indicator of the speed at which the touch screen responds to pressure changes and determines the setting of the sampling duration; and the signal response mode parameter is a mode setting for the touch screen response, such as single-point touch, multi-point touch or other complex modes, which affects the working mode setting of the entire touch screen. The settings of these three parameters jointly determine the sampling efficiency and signal accuracy. The sampling frequency is set based on the touch sensitivity parameter, the sampling duration is set based on the pressure response speed parameter, and the working mode of the touch screen is set based on the signal response mode parameter to ensure that the signal of the touch event can be efficiently and accurately collected. Based on the set sampling frequency and sampling duration, the touch screen is signaled and the capacitance signal data including the touch point position coordinates, contact area and capacitance value are obtained. The signal data includes the capacitance response of the touch screen at each moment, reflecting various characteristics of the touch, such as the movement trajectory of the touch point, the pressure distribution and the change of the contact area. The capacitance signal data is filtered by a Butterworth low-pass filter to obtain smoother and more realistic capacitance signal data, reducing the interference of noise on subsequent analysis. The filtered capacitance signal data is grouped according to the position coordinates of the touch point, so that the capacitance signal subsequence corresponding to each touch point can be processed separately. The capacitance signal subsequence is normalized according to the signal response mode parameters to ensure that the data of different touch events are within a unified scale range, avoiding the influence of the amplitude difference between different touch events on the feature extraction. The normalized capacitance signal subsequence is subjected to time domain feature extraction to extract parameters that can accurately characterize the characteristics of the touch point. The first characteristic parameter characterizing the average capacitance strength of the touch point is calculated, which reflects the overall pressure level of the touch point; the second characteristic parameter characterizing the capacitance fluctuation amplitude of the touch point is calculated to reflect the stability and fluctuation of the pressure of the touch point. The pressure change trend of the touch point and the duration of the touch event are reflected by extracting the time domain feature. The normalized capacitance signal subsequence is subjected to peak detection and energy accumulation calculation. Peak detection helps identify the maximum capacitance strength of the touch point during the touch process, which helps to evaluate the pressure strength of the touch point; while energy accumulation calculation can characterize the persistence of the touch point, reflecting the degree of pressure maintained by the touch point on the screen over a certain period of time. The first, second, third and fourth characteristic parameters are weighted and combined to obtain a comprehensive initial characteristic vector.

[0030] S2, establishing a pressure mapping model based on the initial eigenvector and generating pressure distribution data;

[0031] Specifically, the initial feature vector is classified according to the size of the pressure value to obtain a light pressure feature subvector and a heavy pressure feature subvector. The touch point features within different pressure ranges are processed separately so that different modeling methods can be used for analysis in subsequent steps. When processing the light pressure feature subvector, a least square linear mapping model is constructed. By constructing a feature matrix and a pressure value matrix, and using matrix operations to calculate the linear mapping coefficient matrix, a linear mapping equation is obtained. The linear mapping model expresses the relationship between the pressure value and the initial feature vector in mathematical form, and can effectively model the touch points in the light pressure interval. For the heavy pressure feature subvector, a three-layer radial basis function (RBF) neural network is used to model the nonlinear mapping. A three-layer RBF neural network structure is constructed, in which the dimension of the input layer is n, which is the dimension of the initial feature vector, and the feature vector of each touch point will be used as the input of the network. The number of neurons in the hidden layer is determined by the K-means clustering algorithm. An optimal number of neurons m is determined by clustering, so that the network has sufficient expression ability when processing heavy pressure data. The dimension of the output layer is 1, and the output result is the corresponding pressure value. For each heavy pressure feature subvector sample, the Euclidean distance from the center of the hidden layer neuron is calculated, and then the distance is input into the Gaussian radial basis function (RBF) for kernel function transformation. The orthogonal least squares (OLS) method is used to optimize the connection weights from the hidden layer to the output layer to minimize the root mean square error (RMSE). In this way, an optimal nonlinear mapping network parameter is obtained to effectively describe the touch point characteristics in the heavy pressure interval. The linear mapping equation and the nonlinear mapping network parameters are combined to construct a dual-mode pressure mapping model. In this model, a pressure value selection function is set. When the pressure value corresponding to the input feature vector is less than 1N, linear mapping is selected; when the pressure value corresponding to the input feature vector is greater than or equal to 1N, nonlinear mapping is selected. Through the dual-mode selection method, the touch points in different pressure ranges are modeled more accurately, so that the model can take into account both accuracy and efficiency. After the initial feature vector is input into the dual-mode pressure mapping model, the pressure value selection function automatically selects the appropriate mapping method to calculate the pressure mapping value of each touch point. The pressure mapping value of each touch point is combined with its two-dimensional coordinates and contact area into a four-dimensional data point to form a comprehensive data structure containing pressure, position and contact area. The four-dimensional data points are spatially interpolated by constructing a cubic spline interpolation function. The cubic spline interpolation function can ensure the continuity of the value and first-order derivative at the data point during the interpolation process, making the interpolation result smoother and more natural. Through interpolation calculation, complete pressure distribution data is obtained, including the spatial position, contact area and corresponding pressure value distribution of the touch point, reflecting the pressure state of each touch point on the touch screen.

[0032] S3, extracting features and fuzzy quantizing the pressure distribution data, and performing feature fusion to obtain a fused feature vector;

[0033] It should be noted that the pressure value, position coordinates and contact area data of the touch point are extracted from the pressure distribution data to construct the pressure feature matrix, position feature matrix and area feature matrix. By normalizing the maximum and minimum values ​​of the feature matrix, all feature data are standardized to the same scale range, usually the range between [0,1], to eliminate the influence of different units and dimensions between different features. Fuzzy quantization is performed on the standardized feature data. A trapezoidal fuzzy membership function is constructed to divide different pressure features into different fuzzy intervals, including a low pressure interval [0,0.3], a medium pressure interval [0.3,0.7] and a strong pressure interval [0.7,1]. Each interval has a corresponding trapezoidal membership function to reflect the fuzzy membership degree of the feature data in different intervals. At the same time, the position features and area features are also divided in a similar way to ensure that the data of different features can be compared and processed with the pressure features at the same fuzzy semantic level. Through fuzzy quantization, the standardized feature data is converted into a fuzzy feature vector to improve the interpretability of the features and the robustness of the model. By constructing a reference sequence and a comparison sequence, the fuzzy feature vector is subjected to a grey system analysis to calculate the correlation coefficient between each feature. The grey correlation coefficient matrix is ​​obtained by calculating the grey correlation coefficient between each feature vector and the reference sequence, which reflects the degree of correlation between each feature. Grey system theory is very effective in dealing with uncertainty and fuzzy information. Introducing this theory in touch screen pressure perception can better reflect the relationship between different features in complex touch events. In order to extract the correlation between features, the grey correlation coefficient matrix is ​​converted into a feature correlation matrix, and it is symmetric to obtain a symmetric correlation matrix. The correlation matrix is ​​converted into a symmetric form for subsequent eigendecomposition operations. The symmetric correlation matrix is ​​subjected to eigendecomposition operations to obtain the eigenvalue sequence of each feature. Through the decomposition of eigenvalues, it is analyzed which features play a major role in the overall structure of the data, and dimensionality reduction processing is performed. By calculating the cumulative contribution rate η(k) of the first k eigenvalues, the number of principal components k is determined. When the cumulative contribution rate reaches or exceeds 85%, it is considered that these principal components can explain most of the information in the data, so the corresponding k eigenvectors are selected to construct the projection matrix. The fuzzy feature vector is multiplied by the projection matrix to perform dimensionality reduction transformation to obtain the principal component features. The feature weight is calculated according to the variance contribution rate of each principal component. The variance contribution rate indicates the contribution of each principal component to the overall variance. A higher variance contribution rate means that the principal component has a stronger ability to explain the data and should be given a higher weight. The principal component features and the corresponding feature weights are weighted and summed to obtain the final fused feature vector.

[0034] S4, inputting the fused feature vector into the Gaussian mixture model, and calculating the dynamic pressure threshold parameter through the expectation maximization algorithm;

[0035] Specifically, the Gaussian components of the fused feature vector are initialized. Based on the different levels of touch screen pressure, three Gaussian components are set, corresponding to the three pressure levels of light pressure, medium pressure and heavy pressure respectively. The initial mean vector and covariance matrix of each Gaussian component are calculated using the K-means clustering method to determine the initial parameters of the Gaussian mixture model. The initial parameters of the Gaussian mixture model are substituted into the multivariate Gaussian distribution function to construct a mixed probability density function. The mixed probability density function is composed of a weighted combination of three Gaussian components, and each component represents the distribution of a pressure level, which can describe the initial distribution model of the entire pressure level. The multivariate Gaussian distribution function represents the distribution of light pressure, medium pressure and heavy pressure in the feature space in a parameterized manner. The expected step calculation is performed on the initial distribution model of the pressure level. In this step, the posterior probability of each sample data point belonging to each pressure level is calculated by using the Bayesian formula to obtain the probability distribution matrix of the pressure level. The matrix contains the probability of belonging to each fused feature vector in different pressure levels. Based on the probability distribution matrix of the pressure level obtained in the expectation step, the maximization step calculation is performed to update the mixing coefficient, mean vector and covariance matrix of the Gaussian mixture model. By maximizing the posterior probability, the parameters of the model are continuously adjusted so that the model can better fit the actual data distribution. After multiple iterative updates, the updated Gaussian mixture model parameters are obtained, which reflect the optimal distribution state of the pressure data. The updated model parameters are re-substituted into the multivariate Gaussian distribution function to calculate the value of the log-likelihood function, and the convergence of the model is judged based on the value. When the increment of the log-likelihood function value is less than the set threshold, or the number of iterations reaches the preset upper limit, the expectation maximization algorithm stops iterating and obtains the final Gaussian mixture model parameters. According to the final Gaussian mixture model parameters, the intra-class dispersion and inter-class distance of each pressure level are calculated, and the discrimination criteria of the pressure level are constructed. The intra-class dispersion indicates the degree of dispersion of samples within the same pressure level, while the inter-class distance indicates the degree of separation between different pressure levels. Through these two indicators, the rationality of the pressure level division is evaluated, and whether the model can effectively distinguish between light pressure, medium pressure and heavy pressure is judged. The probability density functions of adjacent pressure levels are optimized by boundary optimization. The optimal segmentation threshold between pressure levels is determined by calculating the points where two probability density functions are equal, and the dynamic threshold interval of pressure levels is obtained. The purpose of this optimization step is to find the most appropriate boundary position between each pressure level to minimize the fuzzy area between different pressure levels, so that the model is more accurate in pressure recognition. The result of boundary optimization is a set of dynamic thresholds that are adaptively adjusted according to changes in different touch conditions to ensure the accuracy of pressure perception. The dynamic threshold interval of pressure level is combined and mapped with the Gaussian probability density function of each pressure level to obtain the dynamic pressure threshold parameter.The dynamic pressure threshold parameter can be adaptively adjusted according to the actual touch situation, thus providing a flexible threshold setting for the pressure perception of the touch screen. Through the dynamic adjustment mechanism, the touch screen optimizes the response sensitivity and accuracy of its pressure perception according to the actual operation of the user, providing a more intelligent and accurate touch feedback experience.

[0036] S5, calculating the change rate, acceleration and jitter degree of the dynamic pressure threshold parameter to obtain a time domain feature sequence;

[0037] Among them, the first length value and the second length value of the feature extraction window are determined based on the duration of the touch event, and the two length values ​​are kept in a ratio of 2:1. The first length value is relatively large, while the second length value is relatively small. The changes in the pressure perception process are analyzed in a multi-scale manner through the ratio setting. On this basis, the first and second length values ​​are divided by sliding steps, wherein the first sliding step is set to 100ms and the second sliding step is set to 50ms, to obtain the first time window sequence and the second time window sequence. The dynamic pressure threshold parameter is subjected to a first differential operation in the first time window sequence, and the pressure value difference between adjacent sampling points is calculated and divided by the first sampling time interval to obtain the first pressure change rate sequence, which describes the speed of change of pressure in a longer time interval. Similarly, a second differential operation is performed in the second time window sequence to obtain a second pressure change rate sequence. Compared with the first change rate sequence, the second change rate sequence has a higher time resolution and can better capture the detailed changes of pressure in a shorter time interval. The acceleration feature is calculated for the pressure change rate sequence. In the first time window, by performing a second-order difference operation between adjacent time windows on the first pressure change rate sequence, a first pressure acceleration sequence is obtained to describe the acceleration of pressure change and reflect the dynamic characteristics of pressure change. Similarly, by performing a second-order difference operation on the second pressure change rate sequence in the second time window, a second pressure acceleration sequence is obtained to capture the acceleration of pressure at a higher time resolution. The degree of pressure jitter is analyzed. By calculating the standard deviation of the first pressure change rate sequence and the first pressure acceleration sequence in the first time window, a first pressure jitter sequence is obtained, and the jitter sequence reflects the stability of the pressure signal in the time window. Similarly, by calculating the standard deviation of the second pressure change rate sequence and the second pressure acceleration sequence in the second time window, a second pressure jitter sequence is obtained. A feature matrix is ​​constructed based on the pressure change rate, acceleration and jitter sequence. A first three-dimensional feature matrix is ​​constructed using the first pressure change rate sequence, the first pressure acceleration sequence and the first pressure jitter sequence; similarly, a second three-dimensional feature matrix is ​​constructed using the second pressure change rate sequence, the second pressure acceleration sequence and the second pressure jitter sequence. These three-dimensional feature matrices are used to describe the multi-scale variation characteristics of pressure during touch. The first and second three-dimensional feature matrices are rearranged and merged according to the sliding order of the time window to form a combined time series feature array. In order to smooth the noise in the combined time series feature array, a double sliding average filter is performed on it. Five adjacent time windows in the first time window sequence are selected for mean calculation to obtain the first mean result; ten adjacent time windows in the second time window sequence are selected for mean calculation to obtain the second mean result.The first mean result and the second mean result are weighted and combined to obtain a multi-scale smoothed feature sequence, which effectively reduces the noise in the pressure signal while retaining the main trend and change information of the pressure feature. The multi-scale smoothed feature sequence is hierarchically accumulated to extract its long-term trend information. The first exponential weight is applied to the feature values ​​corresponding to the first time window sequence for combination; similarly, the second exponential weight is applied to the feature values ​​corresponding to the second time window sequence for combination, and the ratio of the first exponential weight to the second exponential weight is kept at 3:2 to obtain a hierarchical cumulative feature vector. By setting the weights, the features at different time scales have different importance, and the long-term and short-term feature information are effectively integrated. The hierarchical cumulative feature vector is reconstructed in time order, and the reconstructed feature vector is normalized in sections. In the overlapping interval of the first time window sequence and the second time window sequence, the weighted average strategy is used for feature fusion to ensure that the features at different scales can transition smoothly. In this way, a time domain feature sequence is finally obtained, which has continuity and stability in time and scale, and can fully reflect the dynamic change characteristics of the touch screen pressure under different time windows.

[0038] S6, performing comprehensive calculation on the time domain feature sequence, the pressure value and the contact area to obtain a pressure influence factor, and adjusting the first touch screen response parameter according to the pressure influence factor to output a second touch screen response parameter.

[0039] Specifically, the time domain feature sequence is divided into pressure time-varying feature subsequences, acceleration time-varying feature subsequences and jitter time-varying feature subsequences according to the data structure, which respectively reflect the different dynamic characteristics of pressure during the touch process. Each subsequence is standardized to eliminate the dimensional differences between different features and obtain a standardized time-varying feature set. The pressure time-varying feature subsequence in the standardized time-varying feature set is cross-correlated with the pressure value of the touch point to obtain the correlation between them, that is, the first feature correlation coefficient. Similarly, the acceleration time-varying feature subsequence and the contact area of ​​the touch point are cross-correlated to obtain the second feature correlation coefficient. By quantifying the correlation between the features and the physical parameters, the influence of different touch features on the pressure response of the touch screen is clarified. Based on the first feature correlation coefficient and the second feature correlation coefficient, a feature correlation model is constructed. In order to capture the change of feature correlation over time, a time decay factor is introduced to perform time-series weighting on the feature correlation to obtain a dynamic feature correlation matrix, which reflects the correlation of features such as pressure and acceleration in time evolution. At the same time, by introducing the time decay factor, the recent feature correlation has a greater weight than the historical feature correlation, which meets the demand for fast response during the touch process. The dynamic feature association matrix is ​​subjected to singular value decomposition to extract the main feature directions. Through singular value decomposition, the complex feature space is simplified into several main feature directions, which can explain the most important change trends in the feature data. The main feature directions are projected with the standardized time-varying feature set to obtain a pressure feature contribution vector, which reflects the projection of the standardized features on the main feature directions, that is, the contribution of different features in the pressure response process. The components in the pressure feature contribution vector are nonlinearly combined to obtain a pressure influence factor. The complex contribution relationship of each feature to the pressure response is fully considered, so that the final pressure influence factor not only includes linear contributions, but also includes interactions and nonlinear effects between features, and more accurately reflects the influence of pressure on the touch screen. According to the pressure influence factor, the response parameters of the first touch screen are adjusted, including the adjustment of touch sensitivity parameters, pressure response speed parameters and signal response mode parameters respectively. The touch sensitivity parameters are linearly adjusted using the pressure influence factor, and the touch sensitivity is directly adjusted according to the size of the pressure influence factor to improve the response accuracy of the touch screen under different pressures. The pressure response speed parameter is adjusted exponentially, so that the touch screen responds slowly at low pressure, but responds significantly faster at high pressure, thus meeting the different needs of users for light touch and heavy press. For the signal response mode parameter, a segmented adjustment method is adopted to dynamically adjust the working mode of the touch screen according to the interval of the pressure influencing factor, so that it has the optimal response mode under different pressure conditions. The adjusted response parameter matrix is ​​input into the parameter optimization model, and the parameter combination is optimized using the least squares method.The least squares method obtains the optimal parameter combination by minimizing the error between the response and the target. These parameter combinations can meet the actual needs of users in touch operations to the greatest extent. The optimized response parameter matrix contains the optimal touch sensitivity, pressure response speed and signal response mode, ensuring that the touch screen performs best under various pressures. The optimal response parameter combination is repackaged as the second touch screen response parameter and finally output. The second touch screen response parameter contains the optimized touch sensitivity parameter, pressure response speed parameter and signal response mode parameter, and these parameters are integrated into a whole, so that the touch screen can adaptively adjust its own response characteristics according to the changes in the pressure influencing factors.

[0040] In one example, collecting capacitance signal data of the touch screen based on the first touch screen response parameter and calculating an initial feature vector includes:

[0041] Parse the first touch screen response parameter to obtain a touch sensitivity parameter, a pressure response speed parameter, and a signal response mode parameter, set a sampling frequency based on the touch sensitivity parameter, set a sampling duration based on the pressure response speed parameter, and set a working mode of the touch screen based on the signal response mode parameter;

[0042] According to the sampling frequency and sampling time, the touch screen is sampled to obtain capacitance signal data including the touch point position coordinates, contact area and capacitance value;

[0043] The capacitance signal data is filtered by a Butterworth low-pass filter to obtain filtered capacitance signal data, and the filtered capacitance signal data is grouped according to the touch point position coordinates to obtain a capacitance signal subsequence corresponding to each touch point, and the capacitance signal subsequence is normalized based on the signal response mode parameter to obtain a normalized capacitance signal subsequence;

[0044] Performing time domain feature extraction on the normalized capacitance signal subsequence, and calculating a first characteristic parameter representing an average capacitance strength of the touch point and a second characteristic parameter representing a capacitance fluctuation amplitude of the touch point;

[0045] Performing peak detection and energy accumulation calculation on the normalized capacitance signal subsequence to obtain a third characteristic parameter characterizing the maximum capacitance strength of the touch point and a fourth characteristic parameter characterizing the persistence of the touch point;

[0046] The first characteristic parameter, the second characteristic parameter, the third characteristic parameter and the fourth characteristic parameter are weightedly combined to obtain an initial characteristic vector.

[0047] In this example, the touch screen response is parsed. The touch sensitivity parameter (denoted as ) is an indicator to measure the sensitivity of the touch screen to different touch forces. The higher the value, the more sensitive the screen is to slight touch. Pressure response speed parameter (denoted as ) indicates the response time of the touch screen to pressure changes, and is used to set the appropriate sampling time length to cover the entire process of pressure changes. Signal response mode parameters (denoted as ) is used to set the parameters of the touch screen working mode, including single-point touch, multi-point touch, or enhanced mode. The specific choice depends on the touch application scenario that the screen needs to support. After parsing these parameters, according to the touch sensitivity parameters Set the sampling frequency , using the formula To determine, It is the conversion factor from sensitivity to sampling frequency, which is used to map the sensitivity parameter to a reasonable sampling frequency. The higher the sampling frequency, the more times the signal is sampled per second, and the more touch details can be captured. Set the sampling duration , then through To confirm, It is an empirical factor used to ensure that the sampling time is long enough to cover the entire process from the beginning of the touch to the leaving of the screen. Based on the signal response mode parameters , set the working mode of the touch screen. , the touch screen enters single-touch mode. , then enter the multi-touch mode, and if , it enters enhanced mode, which allows the screen to respond to complex multi-point pressure to adapt to more diverse application scenarios. According to the sampling frequency and sampling time, the touch screen is sampled to obtain the touch point position coordinates on the touch screen. , contact area And the capacitance value .in, Indicates the two-dimensional coordinate position of the touch point on the screen. represents the contact area of ​​the touch point, and It's about time The capacitance value function reflects the capacitance change during the touch process. The original capacitance signal data is filtered. The Butterworth low-pass filter is used to smooth the signal. The design goal of the Butterworth filter is to maximize the original characteristics of the signal while eliminating high-frequency noise. The transfer function of the low-pass filter is:

[0048] ;

[0049] in, is a complex frequency domain variable, is the cutoff frequency of the filter, is the order of the filter. By properly choosing the cutoff frequency and order , effectively filtering out high-frequency noise. According to the position coordinates of the touch point Group the filtered signals to obtain the capacitance signal subsequence of each touch point , which is convenient for independent analysis of the pressure characteristics of each touch point. The capacitance signal subsequence is normalized according to the signal response mode parameters. The minimum-maximum normalization method is used to map the signal to the [0,1] interval. The normalization formula is:

[0050] ;

[0051] in, and are the minimum and maximum values ​​of the signal, respectively, and the normalized signal Ensure that the capacitance signals between different touch points are on the same scale, so that the subsequent feature extraction is more accurate. Perform time domain feature extraction and calculate the first characteristic parameter representing the average capacitance strength of the touch point , and its calculation formula is: ;

[0052] in, Indicates that during the entire sampling period The average value of the internal capacitance signal is an important indicator for describing the average pressure of the touch point. Calculate the second characteristic parameter that characterizes the fluctuation amplitude of the touch point capacitance , by finding the standard deviation of the capacitance signal:

[0053] ;

[0054] parameter It reflects the fluctuation of the capacitance value of the touch point and is an important feature to describe the stability of touch pressure. Peak detection and energy accumulation calculation are performed on the normalized capacitance signal subsequence. Peak detection is used to calculate the maximum capacitance strength of the touch point. , the formula is:

[0055] ;

[0056] in, Indicates the maximum pressure during the entire touch process, which is the key feature for identifying hard press and light touch. The fourth characteristic parameter that characterizes the persistence of the touch point is obtained by calculating the energy accumulation of the signal. ;

[0057] parameter It reflects the energy distribution during the entire touch process and describes the continuity and stability of the touch. Perform weighted combination to obtain the initial eigenvector The formula for the weighted combination is:

[0058] ;

[0059] in, is the weight coefficient, which is used to adjust the contribution of different features according to application requirements. For example, if the application scenario pays more attention to the stability of touch pressure, increase and weights to highlight the impact of volatility and persistence characteristics.

[0060] In one example, a pressure mapping model is established according to the initial feature vector, and pressure distribution data is generated, including:

[0061] The initial feature vector is classified according to the pressure value to obtain a light pressure feature sub-vector and a heavy pressure feature sub-vector;

[0062] The least square linear mapping model is constructed for the light pressure characteristic sub-vector. By constructing the characteristic matrix and the pressure value matrix, the linear mapping coefficient matrix is ​​calculated by matrix operation to obtain the linear mapping equation.

[0063] A three-layer radial basis function neural network structure is constructed, including: an input layer with a dimension of n, where n is the dimension of the initial feature vector; a hidden layer with m neurons, where m is the optimal number of neurons determined by K-means clustering; an output layer with a dimension of 1, used to output the pressure value;

[0064] The Euclidean distance from each sample of the gravity pressure feature subvector to the center of the hidden layer neuron is calculated, and the Euclidean distance is input into the Gaussian radial basis function for kernel function transformation. The orthogonal least squares algorithm is used to optimize the connection weights from the hidden layer to the output layer. The objective function is constructed by minimizing the root mean square of the output error. The gradient descent method is used to iteratively optimize the weight matrix to obtain the nonlinear mapping network parameters.

[0065] Integrate the linear mapping equation and nonlinear mapping network parameters, build a dual-mode pressure mapping model, and establish a pressure value selection function. When the pressure value corresponding to the input feature vector is less than 1N, select linear mapping; when the pressure value corresponding to the input feature vector is greater than or equal to 1N, select nonlinear mapping;

[0066] The initial feature vector is input into the dual-mode pressure mapping model, the mapping mode is automatically selected through the pressure value selection function, the pressure mapping value of each touch point is calculated, and the pressure mapping value is combined with the two-dimensional coordinates and contact area of ​​the touch point into a four-dimensional data point;

[0067] A cubic spline interpolation function is constructed to perform spatial interpolation on four-dimensional data points. The interpolation function satisfies the continuity conditions of the value and the first-order derivative at the data point, and generates pressure distribution data. The pressure distribution data contains the distribution information of the touch point position, contact area and pressure value.

[0068] In this example, the initial feature vector is classified according to the pressure value to obtain light pressure feature sub-vectors and heavy pressure feature sub-vectors. Set a pressure threshold ,choose , the part of the initial feature vector corresponding to the pressure value less than 1N is classified as the light pressure feature subvector, and the part greater than or equal to 1N is classified as the heavy pressure feature subvector. The purpose of classification is to facilitate the application of different mapping models to different pressure ranges to achieve more accurate pressure perception. Linear mapping modeling is performed on the light pressure feature subvector, using the least squares linear mapping model. Construct the feature matrix And the pressure value matrix , where the matrix express The feature vectors of samples, each feature vector has a dimension of ,matrix Represents the corresponding pressure value. The least squares method determines the linear mapping coefficient matrix by minimizing the sum of squares of the error , its objective function is:

[0069] ;

[0070] In order to find the optimal , using matrix operations to minimize the objective function. The closed-form expression of the least squares solution is:

[0071] ;

[0072] Through this linear mapping coefficient matrix , we get the linear mapping equation:

[0073] ;

[0074] This equation is used to predict the pressure value of the characteristic subvector in the light pressure area, and is mainly used for response modeling under low pressure. For the heavy pressure characteristic subvector, a more complex nonlinear model is constructed to capture the complex relationship between pressure and characteristics. A three-layer radial basis function neural network structure is used. The input layer dimension of the network is , which is the dimension of the feature vector. The hidden layer contains neurons, The value of is determined by the K-means clustering algorithm. Clustering can determine the optimal number of neurons according to the distribution of data and build a suitable hidden layer structure. The dimension of the output layer is 1, which is used to output the final pressure value. In the radial basis function network, for each sample of the heavy pressure feature subvector, the Euclidean distance to the center of the hidden layer neuron is calculated. Suppose the input feature vector is , the center of the hidden layer neuron is , then the Euclidean distance It is expressed as:

[0075] ;

[0076] The calculated Euclidean distance Input into Gaussian radial basis function for kernel function transformation to get activation value :

[0077] ;

[0078] in, is the bandwidth parameter of the radial basis function, which is used to control the smoothness of the radial basis function. The orthogonal least squares (OLS) method is used to optimize the connection weights from the hidden layer to the output layer in order to minimize the root mean square of the output error (RMSE). The objective function is:

[0079] ;

[0080] in, Represents the weight vector from the hidden layer to the output layer, by minimizing Finding the optimal weights The weight matrix is ​​iteratively optimized using the gradient descent method, and the update formula is:

[0081] ;

[0082] in, is the learning rate, and the parameters of the nonlinear mapping network are finally obtained through multiple iterations. The linear mapping equation is integrated with the nonlinear mapping network parameters to build a dual-mode pressure mapping model. Define a pressure value selection function , when the pressure value corresponding to the input eigenvector is less than 1 When the pressure value is greater than or equal to When the pressure is high, the nonlinear mapping method is selected. The dual-mode pressure mapping model can automatically select the most appropriate mapping method according to the input features to ensure the accuracy of pressure perception. The initial feature vector is input into the dual-mode pressure mapping model, and the pressure value is selected by the function Automatically select the mapping method and calculate the pressure mapping value of each touch point . Then map the pressure value , the two-dimensional coordinates of the touch point And the contact area Combined into a four-dimensional data point These four-dimensional data points fully describe the spatial position, contact area and pressure information of the touch point. In order to obtain the spatial distribution of pressure, a cubic spline interpolation function is constructed to perform spatial interpolation on the four-dimensional data points. The form of the cubic spline interpolation function is:

[0083] ;

[0084] in, are the coefficients of the spline, is the position of the interpolation node. Cubic spline interpolation satisfies the continuity conditions of the value and first-order derivative at the data point to ensure the smoothness and accuracy of the interpolation result. Through cubic spline interpolation, complete pressure distribution data including touch point position, contact area and pressure value are generated.

[0085] In one example, feature extraction and fuzzy quantization are performed on the pressure distribution data, and feature fusion is performed to obtain a fused feature vector, including:

[0086] Extract the touch point pressure value, position coordinates and contact area data from the pressure distribution data, construct the pressure feature matrix, position feature matrix and area feature matrix, perform maximum and minimum value normalization on the pressure feature matrix, position feature matrix and area feature matrix, and obtain standardized feature data;

[0087] The standardized feature data is fuzzy quantized, a trapezoidal fuzzy membership function is constructed, the pressure features are divided into low pressure interval [0, 0.3], medium pressure interval [0.3, 0.7] and strong pressure interval [0.7, 1], the position features and area features are divided into corresponding levels, and the fuzzy feature vector is obtained;

[0088] Construct reference sequence and comparison sequence for fuzzy feature vectors, calculate correlation coefficient, and obtain grey correlation coefficient matrix;

[0089] The grey correlation coefficient matrix is ​​converted into a characteristic correlation matrix, the characteristic correlation matrix is ​​symmetricized to obtain a symmetric correlation matrix, and the symmetric correlation matrix is ​​subjected to characteristic decomposition to obtain an eigenvalue sequence;

[0090] Calculate the cumulative contribution rate η(k) of the first k eigenvalues. When η(k) ≥ 85%, determine the number of principal components k, and select the corresponding k eigenvectors to construct the projection matrix.

[0091] The fuzzy feature vector is multiplied by the projection matrix on the left to perform dimensionality reduction transformation to obtain the principal component features; the feature weights are calculated according to the variance contribution rate of each principal component, and the principal component features and the feature weights are weighted summed to obtain the fused feature vector.

[0092] In this example, the pressure value of each touch point on the touch screen is , location coordinates and contact area Extract. This information is constructed into pressure feature matrices , position feature matrix and area characteristic matrix .set up is the number of touch points, the matrix express The pressure value of each touch point, matrix Represents the position coordinates of the touch point. Each touch point has two-dimensional coordinates. ,matrix Represents the contact area of ​​each touch point. For the pressure feature matrix , position feature matrix and area characteristic matrix Perform maximum and minimum value normalization. The normalization formula is:

[0093] ;

[0094] ;

[0095] ;

[0096] in, and Respectively represent the minimum and maximum pressure values, Respectively represent the minimum and maximum values ​​of the position coordinates, and is the minimum and maximum value of the contact area. Through normalization, all pressure, position and area data are mapped to the [0,1] interval to eliminate the influence of different eigenvalue ranges and obtain standardized feature data. The standardized feature data is fuzzy quantized and a trapezoidal fuzzy membership function is constructed to divide different features into corresponding fuzzy levels. For pressure characteristics, it is divided into a low pressure interval [0,0.3], a medium pressure interval [0.3,0.7], and a strong pressure interval [0.7,1]. Each interval uses a trapezoidal membership function to describe the degree to which the pressure value falls into each interval. The expression of the fuzzy membership function is:

[0097] ;

[0098] ;

[0099] ;

[0100] Similarly, the position features and area features are fuzzy divided to obtain the fuzzy feature vector of each touch point, which reflects the uncertainty and fuzziness of the touch point features and improves the robustness of the model to noise. In order to analyze the correlation between features, a reference sequence and a comparison sequence are constructed for the fuzzy feature vectors, and the correlation coefficient is calculated using the correlation analysis method in the grey system theory. The reference sequence is defined as , the comparison sequence is The calculation formula of grey correlation coefficient is:

[0101] ;

[0102] in, is the resolution coefficient, which is usually taken as 0.5. The grey correlation coefficient matrix is ​​obtained by calculation , where each element Represents the degree of association between the reference sequence and the comparison sequence. The grey correlation coefficient matrix is ​​converted into a characteristic correlation matrix and symmetricized to obtain a symmetric correlation matrix. The symmetric processing is completed by averaging the correlation matrix and its transposed matrix to ensure the symmetry of the matrix. Then the eigendecomposition operation is performed on the symmetric correlation matrix to obtain the eigenvalue sequence These eigenvalues ​​reflect the contribution of each feature to the overall structure of the data. In order to determine the number of principal components , before calculation The cumulative contribution rate of the eigenvalues

[0103] ;

[0104] when When eigenvectors to construct the projection matrix The projection matrix is ​​used to reduce the dimension of high-dimensional features to low-dimensional space in order to reduce redundant information and improve computational efficiency. The fuzzy feature vector is multiplied by the projection matrix to perform dimensionality reduction transformation to obtain the principal component features:

[0105] ;

[0106] in, is the matrix of the original fuzzy eigenvector, is the projection matrix, is the principal component feature matrix after dimensionality reduction. By reducing the dimension of the principal component features, most of the information of the original features is retained, while reducing the feature dimension and reducing the computational complexity. In order to fuse the principal component features, the feature weights are calculated according to the variance contribution rate of each principal component. Suppose the variance contribution rate is , then The weight of the principal component is . Perform a weighted sum operation on the principal component features and the corresponding weights to obtain the fused feature vector

[0107] ;

[0108] in, For the principal component features, is the corresponding weight. Through weighted summation, the fusion feature vector integrates the contributions of multiple principal component features to obtain a more stable and robust feature representation.

[0109] In one example, the fused feature vector is input into the Gaussian mixture model, and the dynamic pressure threshold parameter is calculated by the expectation maximization algorithm, including:

[0110] Initialize the Gaussian components of the fused feature vector, set three Gaussian components according to the number of pressure levels, corresponding to light pressure, medium pressure and heavy pressure, calculate the initial mean vector and covariance matrix of each component through K-means clustering, and obtain the initial parameters of the Gaussian mixture model;

[0111] Substitute the initial parameters of the Gaussian mixture model into the multivariate Gaussian distribution function to construct a mixture probability density function, which is composed of a weighted combination of three Gaussian components to obtain an initial distribution model of the pressure level;

[0112] The expected step calculation is performed on the initial distribution model of the pressure level, and the posterior probability of each sample data point belonging to each pressure level is calculated using the Bayesian formula to obtain the probability distribution matrix of the pressure level;

[0113] Based on the probability distribution matrix of the pressure level, the maximization step calculation is performed to update the mixing coefficient, mean vector and covariance matrix of the Gaussian mixture model to obtain the updated model parameters;

[0114] Substitute the updated model parameters into the multivariate Gaussian distribution function again, calculate the log-likelihood function value, and when the increment of the log-likelihood function value is less than the set threshold or the number of iterations reaches the upper limit, obtain the final Gaussian mixture model parameters;

[0115] The intra-class dispersion and inter-class distance of each pressure level are calculated according to the final Gaussian mixture model parameters, and the pressure level discrimination criterion is constructed. The pressure level discrimination criterion is used to evaluate the rationality of the pressure level division.

[0116] The probability density functions of adjacent pressure levels are optimized, and the optimal segmentation threshold of the pressure level is determined by calculating the points of equal probability density to obtain the dynamic threshold interval of the pressure level. The dynamic threshold interval of the pressure level is combined and mapped with the Gaussian probability density function of each pressure level to obtain the dynamic pressure threshold parameter.

[0117] In this example, the Gaussian components of the fused feature vector are initialized, and three Gaussian components are set according to the number of pressure levels, corresponding to the three pressure levels of light pressure, medium pressure and heavy pressure. The fused feature vector is clustered to determine the initial parameters of each Gaussian component. The K-means clustering method is used to calculate the initial mean vector and covariance matrix of each component in the Gaussian mixture model. is the total number of fused feature vectors, is the number of Gaussian components set. Through K-means clustering, the data is divided into 3 categories, each category represents a pressure level. For each category , calculate the initial mean vector and the covariance matrix The mean vector represents the center position of the data point of this category, and the covariance matrix reflects the distribution of the data points of this category in each dimension. and the covariance matrix Then, set the mixing coefficient , which represents the weight of each Gaussian component in the whole. Initially, the mixing coefficient of each component is set to , that is, the weight of each component is equal. After obtaining the initial parameters, construct the initial parameter set of the Gaussian mixture model Substitute the initial parameters into the multivariate Gaussian distribution function to construct a mixed probability density function. For each Gaussian component , its probability density function is expressed as:

[0118] ;

[0119] in, is the fused feature vector, is the dimension of the feature vector, represents the determinant of the covariance matrix, represents the inverse matrix of the covariance matrix. The entire mixture probability density function It is composed of a weighted combination of three Gaussian components and has the form:

[0120] ;

[0121] The mixed probability density function describes the probability distribution of the fused feature vector under different pressure levels, thus forming the initial distribution model of the pressure level. In order to optimize the Gaussian mixture model, the expectation maximization (EM) algorithm is used to iteratively update the model parameters. The expectation step (E step) is performed, and the posterior probability of each sample data point belonging to each pressure level is calculated using the Bayesian formula. Assume For the The data point belongs to The posterior probability of a Gaussian component is calculated as:

[0122] ;

[0123] By calculating the posterior probability of all sample data points, the probability distribution matrix of pressure level is obtained , where each element Representation sample Belongs to the pressure level After obtaining the posterior probability matrix, a maximization step (M step) is performed to update the parameters of the Gaussian mixture model. The update formula of the mixture coefficient is:

[0124] ;

[0125] The update formula of the mean vector is:

[0126] ;

[0127] The update formula of the covariance matrix is:

[0128] ;

[0129] Through the above update formula, the updated mixing coefficient, mean vector and covariance matrix are obtained. Substitute the updated model parameters into the multivariate Gaussian distribution function again to calculate the log-likelihood function value

[0130] ;

[0131] Iteratively update the model parameters until the increment of the log-likelihood function value is less than the set threshold Or the number of iterations reaches the upper limit When , the iteration is stopped and the final Gaussian mixture model parameters are obtained. According to the final Gaussian mixture model parameters, the intra-class dispersion and inter-class distance of each pressure level are calculated to construct the discrimination criteria of the pressure level. It is represented by the trace of the covariance matrix:

[0132] ;

[0133] Distance between classes It is calculated by the Euclidean distance between the two mean vectors:

[0134] ;

[0135] The pressure level discrimination criterion evaluates the rationality of the pressure level division by comparing the intra-class discreteness and the inter-class distance. If the inter-class distance is much larger than the intra-class discreteness, it means that the pressure level division effect is good. In order to optimize the boundary between different pressure levels, the probability density function of adjacent pressure levels is optimized. By calculating the equal point of the probability density of adjacent pressure levels, the optimal segmentation threshold of the pressure level is determined. Assume that the probability density function of two adjacent Gaussian components is and , by solving the equation:

[0136] ;

[0137] Get the optimal segmentation threshold for the pressure level Based on these segmentation thresholds, the dynamic threshold intervals of the pressure levels are obtained, and the dynamic threshold intervals are combined and mapped with the Gaussian probability density functions of each pressure level, and finally the dynamic pressure threshold parameters are obtained.

[0138] In one example, the change rate, acceleration and jitter degree of the dynamic pressure threshold parameter are calculated to obtain a time domain feature sequence, including:

[0139] Determine a first length value and a second length value of the feature extraction window based on the duration of the touch event, wherein the ratio of the first length value to the second length value is 2:1, and divide the first length value and the second length value of the feature extraction window into a first sliding step and a second sliding step, respectively, wherein the first sliding step is 100 ms and the second sliding step is 50 ms, to obtain a first time window sequence and a second time window sequence;

[0140] A first differential operation is performed on the dynamic pressure threshold parameter within the first time window sequence, and the pressure value difference between adjacent sampling points is calculated and divided by the first sampling time interval to obtain a first pressure change rate sequence; a second differential operation is performed on the dynamic pressure threshold parameter within the second time window sequence, and the pressure value difference between adjacent sampling points is calculated and divided by the second sampling time interval to obtain a second pressure change rate sequence;

[0141] Performing a first second-order difference operation on the first pressure change rate sequence between adjacent time windows to obtain a first pressure acceleration sequence; performing a second second-order difference operation on the second pressure change rate sequence between adjacent time windows to obtain a second pressure acceleration sequence;

[0142] Performing a first standard deviation calculation on the first pressure change rate sequence and the first pressure acceleration sequence within the first time window sequence to obtain a first pressure jitter degree sequence; performing a second standard deviation calculation on the second pressure change rate sequence and the second pressure acceleration sequence within the second time window sequence to obtain a second pressure jitter degree sequence;

[0143] A first three-dimensional feature matrix is ​​constructed based on the first pressure change rate sequence, the first pressure acceleration sequence and the first pressure jitter sequence, a second three-dimensional feature matrix is ​​constructed based on the second pressure change rate sequence, the second pressure acceleration sequence and the second pressure jitter sequence, and the first three-dimensional feature matrix and the second three-dimensional feature matrix are rearranged and merged according to the sliding order of the time window to obtain a combined time series feature array;

[0144] Perform double sliding average filtering on the combined time series feature array, select 5 adjacent time windows in the first time window sequence to perform a first mean calculation to obtain a first mean calculation result, select 10 adjacent time windows in the second time window sequence to perform a second mean calculation to obtain a second mean calculation result, perform weighted combination of the first mean calculation result and the second mean calculation result to obtain a multi-scale smoothed feature sequence;

[0145] Perform hierarchical data accumulation on the multi-scale smoothed feature sequence, combine the eigenvalues ​​corresponding to the first time window sequence by the first exponential weight, and combine the eigenvalues ​​corresponding to the second time window sequence by the second exponential weight, where the ratio of the first exponential weight to the second exponential weight is 3:2, and obtain a hierarchical cumulative feature vector;

[0146] The hierarchical cumulative feature vectors are reconstructed in time sequence, and the reconstructed features are normalized piecewise, wherein a weighted average strategy is adopted in the overlapping interval of the first time window sequence and the second time window sequence to obtain a time domain feature sequence.

[0147] In this example, depending on the duration of the touch event Determine the first length value and the second length value of the feature extraction window. Let the first length value be and the second length value is , and make ,Right now For example, if the duration of a touch event is 600ms, set ms and ms. The two feature extraction windows are divided by sliding steps, where the first sliding step is 100ms and the second sliding step is 50ms, to obtain the first time window sequence and the second time window sequence. The length of each window in the first time window sequence is , and slides in steps of 100ms; the window length of the second time window sequence is , sliding with a step length of 50ms, thus ensuring that the feature extraction of touch events has multi-scale temporal resolution. The dynamic pressure threshold parameters are differentiated to calculate the pressure change rate. Indicates the dynamic pressure threshold parameter at time For each time window in the first time window sequence, perform the first difference operation to calculate the pressure difference between adjacent sampling points and divide it by the first sampling time interval (i.e. 100ms) to obtain the first pressure change rate sequence The formula is: ;

[0148] in, Represents the pressure change rate between every two adjacent sampling points in the first time window. Similarly, the second time window sequence is subjected to a second difference operation, and the pressure difference between adjacent sampling points is calculated and divided by the second sampling time interval (i.e., 50ms) to obtain the second pressure change rate sequence

[0149] ;

[0150] Perform a second-order difference operation on the first pressure change rate sequence between adjacent time windows to obtain the first pressure acceleration sequence The calculation formula is:

[0151] ;

[0152] in, It represents the acceleration in the first time window, reflecting the change of the pressure change rate. Similarly, the second pressure change rate sequence is operated by the second-order difference to obtain the second pressure acceleration sequence :

[0153] ;

[0154] These sequences are statistically analyzed to evaluate the fluctuation degree of the pressure signal, i.e., the jitter. and the first pressure-acceleration sequence Calculate the standard deviation in the first time window sequence to obtain the first pressure jitter sequence

[0155] ;

[0156] in, represents the number of sampling points in the first time window, and Represent the mean of pressure change rate and acceleration respectively. Similarly, for the second pressure change rate sequence and the second pressure acceleration sequence Calculate the standard deviation to obtain the second pressure jitter sequence After obtaining the pressure change rate, acceleration and jitter, a three-dimensional feature matrix is ​​constructed based on these sequences. The first three-dimensional feature matrix From the first pressure change rate sequence , the first pressure acceleration sequence and the first pressure jitter sequence Composition, namely:

[0157] ;

[0158] Similarly, construct the second three-dimensional feature matrix The second pressure change rate sequence , the second pressure acceleration sequence and the second pressure jitter sequence composition:

[0159] ;

[0160] For the first three-dimensional feature matrix and the second three-dimensional feature matrix Rearrange and merge according to the sliding order of the time window to obtain a combined time series feature array The combined time series feature array contains feature information at different time scales, which helps to fully capture the dynamic change characteristics of touch events. In order to smooth the noise in the combined time series feature array, Double sliding average filtering is performed. Five adjacent time windows are selected in the first time window sequence for mean calculation to obtain the first mean sequence.

[0161] ;

[0162] Select 10 adjacent time windows in the second time window sequence for mean calculation to obtain the second mean sequence

[0163] ;

[0164] The first mean calculation result and the second mean calculation result are weighted combined to obtain a multi-scale smoothed feature sequence , the formula is:

[0165] ;

[0166] in and is a weight parameter used to balance the influence of different time scales. Perform layered data accumulation on the multi-scale smoothed feature sequence. The feature value corresponding to the first time window sequence is calculated by the first exponential weight. Combine to get the cumulative feature vector

[0167] ;

[0168] Similarly, the eigenvalues ​​corresponding to the second time window sequence are weighted by the second exponential weight. Combine to get the cumulative feature vector

[0169] ;

[0170] in, and The ratio is 3:2, which is used to adjust the accumulation weights of different time scales. The feature vector obtained by accumulation and Further merge to obtain the hierarchical cumulative feature vector . The hierarchical cumulative feature vector Reconstruct in chronological order, and perform piecewise normalization on the reconstructed features. In the overlapping interval of the first time window sequence and the second time window sequence, a weighted average strategy is used to ensure the smoothness of the features in the transition interval. The final time domain feature sequence is continuous and stable, and can well reflect the dynamic change characteristics of touch events at different time scales.

[0171] In one example, the time domain feature sequence is comprehensively calculated with the pressure value and the contact area to obtain the pressure influence factor, and the first touch screen response parameter is adjusted according to the pressure influence factor to output the second touch screen response parameter, including:

[0172] The time domain feature sequence is divided into a pressure time-varying feature subsequence, an acceleration time-varying feature subsequence and a jitter time-varying feature subsequence according to the data structure, and the pressure time-varying feature subsequence, the acceleration time-varying feature subsequence and the jitter time-varying feature subsequence are standardized respectively to obtain a standardized time-varying feature set;

[0173] Performing a cross-correlation operation on the pressure time-varying feature subsequence in the standardized time-varying feature set and the pressure value of the touch point, and performing a cross-correlation operation on the acceleration time-varying feature subsequence and the contact area of ​​the touch point to obtain a first feature correlation coefficient and a second feature correlation coefficient;

[0174] A feature correlation model is constructed based on the first feature correlation coefficient and the second feature correlation coefficient, and a dynamic feature correlation matrix is ​​obtained by performing time-series weighting on the feature correlation by introducing a time decay factor;

[0175] Perform singular value decomposition on the dynamic feature correlation matrix, extract the main feature direction, perform projection operation on the main feature direction and the standardized time-varying feature set, and obtain the pressure feature contribution vector;

[0176] Perform nonlinear combination on each component in the pressure characteristic contribution vector to obtain the pressure influence factor;

[0177] According to the pressure influence factor, a touch sensitivity parameter in the first touch screen response parameter is linearly adjusted, a pressure response speed parameter is exponentially adjusted, and a signal response mode parameter is segmentally adjusted to obtain an adjusted response parameter matrix;

[0178] The adjusted response parameter matrix is ​​input into the parameter optimization model, and the parameter combination is optimized by the least square method to obtain the optimal response parameter combination, and the optimal response parameter combination is repackaged as the second touch screen response parameter, wherein the second touch screen response parameter includes the optimized touch sensitivity parameter, pressure response speed parameter and signal response mode parameter.

[0179] In this example, the time domain feature sequence is divided into three main subsequences according to the data structure, namely, the pressure time-varying feature subsequence, the acceleration time-varying feature subsequence, and the jitter time-varying feature subsequence. is a complete time domain feature sequence, which is divided into (pressure time-varying characteristic subsequence), (acceleration time-varying characteristic subsequence) and (Jitter time-varying feature subsequence). Each subsequence contains different features, pressure time-varying features Reflects the dynamic changes of touch point pressure and acceleration characteristics Reflects the speed of pressure change, while the jitter characteristics Reflects the fluctuation of the signal. These subsequences are standardized to eliminate the differences in the range of eigenvalues, so that all features can be analyzed on the same scale in the subsequent analysis, and a standardized time-varying feature set is obtained. The pressure time-varying feature subsequence in the standardized time-varying feature set is cross-correlated with the pressure value of the touch point. Suppose the pressure value sequence of the touch point is , the pressure time-varying characteristic subsequence is , then their cross-correlation function It is expressed as:

[0180] ;

[0181] in, is the time delay, calculated by The correlation between the pressure feature and the pressure value of the touch point is obtained. Similarly, for the acceleration time-varying feature subsequence Contact area sequence with touch points Perform cross-correlation operation to obtain the second characteristic correlation coefficient

[0182] ;

[0183] These two correlation coefficients and They represent the correlation between the pressure feature and the pressure value, and the correlation between the acceleration feature and the contact area. Based on the first characteristic correlation coefficient and the second characteristic correlation coefficient , construct a feature correlation model. In this model, the time decay factor is introduced The feature correlation is time-weighted to better reflect the change of features over time. The time decay factor is used to give higher weights to the most recent samples, and its form is:

[0184] ;

[0185] in, is the dynamic feature association matrix, is the time decay factor. The introduction of time decay can ensure that the latest features have a greater impact on the current system, while the impact of past features gradually decreases. Perform singular value decomposition to extract the main feature directions. Singular value decomposition converts the matrix Decomposed into the product of three matrices:

[0186] ;

[0187] in, and is an orthogonal matrix, is a diagonal matrix containing By Sort the singular values ​​in and determine the main feature direction. Let the main feature direction be , reflecting the main trend of feature changes. Perform projection operation with the standardized time-varying feature set to obtain the pressure feature contribution vector :

[0188] ;

[0189] in, It is the matrix representation of the standardized time-varying feature set. Through the projection operation, the standardized features are reduced to a new feature space, retaining the most important feature information. The components in the pressure feature contribution vector are nonlinearly combined to obtain the pressure influence factor. Assume that the components of the pressure feature contribution vector are , then the pressure influence factor It is represented by the following nonlinear function:

[0190] ;

[0191] in, is the hyperbolic tangent function, which is used to enhance nonlinear characteristics. is the weight coefficient, which indicates the contribution of each feature to the final pressure influencing factor. , adjust the touch screen response parameters. After linear adjustment, the sensitivity parameter is:

[0192] ;

[0193] in, is the adjustment coefficient, which is used to control the influence of the pressure influence factor on the sensitivity parameter. Perform exponential adjustment, and the pressure response speed after adjustment is:

[0194] ;

[0195] in, For the adjustment coefficient, the response speed is rapidly increased when the pressure is high through exponential adjustment. , adopt the segmented adjustment method and adjust according to the interval where the pressure influencing factor is located:

[0196] ;

[0197] in, and is the preset threshold used to determine the switching of the signal response mode. Combined into a response parameter matrix In order to find the optimal response parameter combination, the matrix is ​​input into the parameter optimization model, and the least squares method is used to optimize the parameters. The goal of the least squares method is to minimize the error between the predicted value and the target value, and its optimization goal is:

[0198] ;

[0199] in, is the target output, is the weight vector to be optimized. By solving the optimization problem, the optimal response parameter combination is obtained. The optimal response parameter combination is repackaged as the second touch screen response parameter, which includes the optimized touch sensitivity parameter, pressure response speed parameter and signal response mode parameter. This process can ensure that the touch screen can adaptively adjust its response characteristics according to different touch characteristics and pressure influencing factors, thereby providing a more flexible and precise touch experience.

[0200] Reference Figure 2 , this embodiment provides a touch screen pressure sensing device, including:

[0201] An acquisition module 1 is used to acquire capacitance signal data of the touch screen based on a first touch screen response parameter, and calculate an initial feature vector;

[0202] Establishing module 2, for establishing a pressure mapping model according to the initial eigenvector and generating pressure distribution data;

[0203] Extraction module 3, used for performing feature extraction and fuzzy quantization on pressure distribution data, and performing feature fusion to obtain a fused feature vector;

[0204] Processing module 4, used for inputting the fused feature vector into the Gaussian mixture model, and calculating the dynamic pressure threshold parameter by using the expectation maximization algorithm;

[0205] Calculation module 5, used to calculate the change rate, acceleration and jitter degree of the dynamic pressure threshold parameter to obtain a time domain feature sequence;

[0206] The adjustment module 6 is used to perform comprehensive calculation on the time domain feature sequence, the pressure value and the contact area to obtain a pressure influence factor, and adjust the first touch screen response parameter according to the pressure influence factor to output a second touch screen response parameter.

[0207] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0208] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0209] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0210] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0211] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0212] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0213] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A touch screen pressure sensing method, characterized in that: The following steps are involved: Collecting capacitance signal data of the touch screen based on the first touch screen response parameter, and calculating an initial feature vector; Establishing a pressure mapping model according to the initial eigenvector and generating pressure distribution data; The pressure distribution data is subjected to feature extraction and fuzzy quantization, and feature fusion is performed to obtain a fused feature vector; specifically comprising: extracting the touch point pressure value, position coordinates and contact area data from the pressure distribution data, constructing a pressure feature matrix, a position feature matrix and an area feature matrix, and performing maximum and minimum value normalization on the pressure feature matrix, the position feature matrix and the area feature matrix to obtain standardized feature data; performing fuzzy quantization processing on the standardized feature data, constructing a trapezoidal fuzzy membership function, dividing the pressure features into a low pressure interval [0, 0.3], a medium pressure interval [0.3, 0.7] and a strong pressure interval [0.7, 1], and dividing the position features and the area features into corresponding levels to obtain a fuzzy feature vector; The fuzzy feature vector is used to construct a reference sequence and a comparison sequence, and the correlation coefficient is calculated to obtain a grey correlation coefficient matrix; the grey correlation coefficient matrix is ​​converted into a feature correlation matrix, the feature correlation matrix is ​​symmetricized to obtain a symmetric correlation matrix, and the symmetric correlation matrix is ​​subjected to a feature decomposition operation to obtain an eigenvalue sequence; the cumulative contribution rate η(k) of the first k eigenvalues ​​is calculated, and when η(k) ≥ 85%, the number of principal components k is determined, and the corresponding k eigenvectors are selected to construct a projection matrix; the fuzzy feature vector is multiplied by the projection matrix on the left to perform a dimensionality reduction transformation to obtain the principal component feature; the feature weight is calculated according to the variance contribution rate of each principal component, and the principal component feature and the feature weight are weighted summed to obtain a fused feature vector; The fused feature vector is input into a Gaussian mixture model, and a dynamic pressure threshold parameter is calculated by an expectation maximization algorithm; Calculating the rate of change, acceleration and degree of jitter of the dynamic pressure threshold parameter to obtain a time domain feature sequence; The time domain feature sequence is comprehensively calculated with the pressure value and the contact area to obtain a pressure influence factor, and the first touch screen response parameter is adjusted according to the pressure influence factor to output a second touch screen response parameter.

2. The touch screen pressure sensing method according to claim 1, characterized in that: The collecting of the capacitance signal data of the touch screen based on the first touch screen response parameter and calculating the initial feature vector includes: Parsing the first touch screen response parameter to obtain a touch sensitivity parameter, a pressure response speed parameter, and a signal response mode parameter, setting a sampling frequency based on the touch sensitivity parameter, setting a sampling duration based on the pressure response speed parameter, and setting a working mode of the touch screen based on the signal response mode parameter; According to the sampling frequency and the sampling duration, the touch screen is subjected to signal acquisition to obtain capacitance signal data including the touch point position coordinates, the contact area and the capacitance value; filtering the capacitance signal data through a Butterworth low-pass filter to obtain filtered capacitance signal data, grouping the filtered capacitance signal data according to the touch point position coordinates to obtain a capacitance signal subsequence corresponding to each touch point, and normalizing the capacitance signal subsequence based on the signal response mode parameter to obtain a normalized capacitance signal subsequence; Performing time domain feature extraction on the normalized capacitance signal subsequence to calculate a first characteristic parameter representing an average capacitance strength of a touch point and a second characteristic parameter representing a capacitance fluctuation amplitude of a touch point; Performing peak detection and energy accumulation calculation on the normalized capacitance signal subsequence to obtain a third characteristic parameter characterizing the maximum capacitance strength of the touch point and a fourth characteristic parameter characterizing the persistence of the touch point; The first feature parameter, the second feature parameter, the third feature parameter and the fourth feature parameter are weightedly combined to obtain an initial feature vector.

3. The touch screen pressure sensing method according to claim 2, characterized in that: The step of establishing a pressure mapping model according to the initial eigenvector and generating pressure distribution data includes: Classifying the initial feature vector according to the pressure value to obtain a light pressure feature sub-vector and a heavy pressure feature sub-vector; A least square linear mapping model is constructed for the light pressure characteristic sub-vector, and a linear mapping equation is obtained by calculating a linear mapping coefficient matrix by constructing a characteristic matrix and a pressure value matrix; Constructing a three-layer radial basis function neural network structure, including: an input layer with a dimension of n, where n is the dimension of the initial feature vector; a hidden layer with m neurons, where m is the optimal number of neurons determined by K-means clustering; and an output layer with a dimension of 1, for outputting pressure values; For each sample of the heavy pressure characteristic subvector, the Euclidean distance to the center of the hidden layer neuron is calculated, the Euclidean distance is input into the Gaussian radial basis function for kernel function transformation, the connection weights from the hidden layer to the output layer are optimized by using the orthogonal least squares algorithm, the objective function is constructed by minimizing the root mean square of the output error, the weight matrix is ​​iteratively optimized by the gradient descent method, and the nonlinear mapping network parameters are obtained; Integrate the linear mapping equation and the nonlinear mapping network parameters, construct a dual-mode pressure mapping model, establish a pressure value selection function, and select linear mapping when the pressure value corresponding to the input feature vector is less than 1N; select nonlinear mapping when the pressure value corresponding to the input feature vector is greater than or equal to 1N; Inputting the initial feature vector into the dual-mode pressure mapping model, automatically selecting a mapping mode through the pressure value selection function, calculating the pressure mapping value of each touch point, and combining the pressure mapping value with the two-dimensional coordinates and contact area of ​​the touch point into a four-dimensional data point; A cubic spline interpolation function is constructed to perform spatial interpolation on the four-dimensional data points, the interpolation function satisfies the value at the data point and the first-order derivative continuity conditions, and pressure distribution data is generated, wherein the pressure distribution data includes distribution information of the touch point position, the contact area and the pressure value.

4. The touch screen pressure sensing method according to claim 1, characterized in that: The step of inputting the fused feature vector into a Gaussian mixture model and calculating a dynamic pressure threshold parameter by an expectation maximization algorithm includes: Initialize the Gaussian components of the fused feature vector, set three Gaussian components according to the number of pressure levels, corresponding to the three pressure levels of light pressure, medium pressure and heavy pressure, calculate the initial mean vector and covariance matrix of each component through K-means clustering, and obtain the initial parameters of the Gaussian mixture model; Substituting the initial parameters of the Gaussian mixture model into the multivariate Gaussian distribution function, constructing a mixture probability density function, wherein the mixture probability density function is composed of a weighted combination of three Gaussian components, and obtaining an initial distribution model of the pressure level; Performing expected step calculation on the initial distribution model of the pressure level, calculating the posterior probability of each sample data point belonging to each pressure level using the Bayesian formula, and obtaining a probability distribution matrix of the pressure level; Performing maximization step calculation based on the probability distribution matrix of the pressure level, updating the mixing coefficient, mean vector and covariance matrix of the Gaussian mixture model, and obtaining updated model parameters; Substitute the updated model parameters into the multivariate Gaussian distribution function again, calculate the log-likelihood function value, and when the increment of the log-likelihood function value is less than the set threshold or the number of iterations reaches the upper limit, obtain the final Gaussian mixture model parameters; Calculating the intra-class dispersion and inter-class distance of each pressure level according to the final Gaussian mixture model parameters, and constructing a pressure level discrimination criterion, wherein the pressure level discrimination criterion is used to evaluate the rationality of the pressure level division; The probability density functions of adjacent pressure levels are optimized, and the optimal segmentation threshold of the pressure level is determined by calculating the probability density equal point to obtain the dynamic threshold interval of the pressure level. The dynamic threshold interval of the pressure level is combined and mapped with the Gaussian probability density function of each pressure level to obtain the dynamic pressure threshold parameter.

5. The touch screen pressure sensing method according to claim 4, characterized in that: The step of calculating the change rate, acceleration and jitter degree of the dynamic pressure threshold parameter to obtain a time domain feature sequence includes: Determine a first length value and a second length value of the feature extraction window based on the duration of the touch event, wherein the ratio of the first length value to the second length value is 2:1, and divide the first length value and the second length value of the feature extraction window into a first sliding step and a second sliding step, respectively, wherein the first sliding step is 100 ms and the second sliding step is 50 ms, to obtain a first time window sequence and a second time window sequence; Performing a first differential operation on the dynamic pressure threshold parameter within the first time window sequence, calculating the pressure value difference between adjacent sampling points and dividing it by the first sampling time interval, to obtain a first pressure change rate sequence; performing a second differential operation on the dynamic pressure threshold parameter within the second time window sequence, calculating the pressure value difference between adjacent sampling points and dividing it by the second sampling time interval, to obtain a second pressure change rate sequence; Performing a first second-order difference operation on the first pressure change rate sequence between adjacent time windows to obtain a first pressure acceleration sequence; performing a second second-order difference operation on the second pressure change rate sequence between adjacent time windows to obtain a second pressure acceleration sequence; Performing a first standard deviation calculation on the first pressure change rate sequence and the first pressure acceleration sequence within the first time window sequence to obtain a first pressure jitter degree sequence; performing a second standard deviation calculation on the second pressure change rate sequence and the second pressure acceleration sequence within the second time window sequence to obtain a second pressure jitter degree sequence; constructing a first three-dimensional feature matrix based on the first pressure change rate sequence, the first pressure acceleration sequence, and the first pressure jitter sequence, constructing a second three-dimensional feature matrix based on the second pressure change rate sequence, the second pressure acceleration sequence, and the second pressure jitter sequence, and rearranging and merging the first three-dimensional feature matrix and the second three-dimensional feature matrix according to the sliding order of the time window to obtain a combined time series feature array; Performing double sliding average filtering on the combined time series feature array, selecting 5 adjacent time windows in the first time window sequence to perform a first mean calculation to obtain a first mean calculation result, selecting 10 adjacent time windows in the second time window sequence to perform a second mean calculation to obtain a second mean calculation result, and performing weighted combination of the first mean calculation result and the second mean calculation result to obtain a multi-scale smoothed feature sequence; Performing hierarchical data accumulation on the multi-scale smoothed feature sequence, combining the feature values ​​corresponding to the first time window sequence by a first exponential weight, and combining the feature values ​​corresponding to the second time window sequence by a second exponential weight, wherein the ratio of the first exponential weight to the second exponential weight is 3:2, to obtain a hierarchical cumulative feature vector; The hierarchical cumulative feature vectors are reconstructed in time sequence, and the reconstructed features are subjected to piecewise normalization processing, wherein a weighted average strategy is adopted in the overlapping interval of the first time window sequence and the second time window sequence to obtain a time domain feature sequence.

6. The touch screen pressure sensing method according to claim 5, characterized in that: The step of performing a comprehensive operation on the time domain feature sequence, the pressure value, and the contact area to obtain a pressure influence factor, and adjusting the first touch screen response parameter according to the pressure influence factor to output a second touch screen response parameter includes: Dividing the time domain feature sequence into a pressure time-varying feature subsequence, an acceleration time-varying feature subsequence and a jitter time-varying feature subsequence according to the data structure, and performing standardization processing on the pressure time-varying feature subsequence, the acceleration time-varying feature subsequence and the jitter time-varying feature subsequence respectively to obtain a standardized time-varying feature set; Performing a cross-correlation operation on the pressure time-varying feature subsequence in the standardized time-varying feature set and the pressure value of the touch point, and performing a cross-correlation operation on the acceleration time-varying feature subsequence and the contact area of ​​the touch point to obtain a first feature correlation coefficient and a second feature correlation coefficient; A feature correlation model is constructed based on the first feature correlation coefficient and the second feature correlation coefficient, and a dynamic feature correlation matrix is ​​obtained by performing time-series weighting on the feature correlation by introducing a time decay factor; Performing singular value decomposition on the dynamic feature association matrix, extracting main feature directions, performing projection operation on the main feature directions and the standardized time-varying feature set, and obtaining a pressure feature contribution vector; Performing nonlinear combination on each component in the pressure characteristic contribution vector to obtain a pressure influence factor; According to the pressure influencing factor, a touch sensitivity parameter in the first touch screen response parameter is linearly adjusted, a pressure response speed parameter is exponentially adjusted, and a signal response mode parameter is segmentedly adjusted to obtain an adjusted response parameter matrix; The adjusted response parameter matrix is ​​input into a parameter optimization model, and the parameter combination is optimized by the least square method to obtain an optimal response parameter combination, and the optimal response parameter combination is repackaged as a second touch screen response parameter, wherein the second touch screen response parameter includes an optimized touch sensitivity parameter, a pressure response speed parameter, and a signal response mode parameter.

7. A touch screen pressure sensing device, characterized in that: For implementing the steps of the method according to any one of claims 1 to 6, the device comprises: An acquisition module, used for acquiring capacitance signal data of the touch screen based on the first touch screen response parameter, and calculating an initial feature vector; An establishing module, used to establish a pressure mapping model according to the initial characteristic vector and generate pressure distribution data; An extraction module, used for performing feature extraction and fuzzy quantization on the pressure distribution data, and performing feature fusion to obtain a fused feature vector; A processing module, used for inputting the fused feature vector into a Gaussian mixture model, and calculating a dynamic pressure threshold parameter by an expectation maximization algorithm; A calculation module, used to calculate the change rate, acceleration and jitter degree of the dynamic pressure threshold parameter to obtain a time domain feature sequence; The adjustment module is used to perform comprehensive calculation on the time domain feature sequence, the pressure value and the contact area to obtain a pressure influence factor, and adjust the first touch screen response parameter according to the pressure influence factor to output a second touch screen response parameter.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images