Time series-based touch response delay data analysis method and system

Abstract

The application discloses a time series-based touch response delay data analysis method and system, and relates to the technical field of big data analysis.The application constructs four types of time series data by collecting touch response delay, touch sampling rate fluctuation, touch point position sequence and touch area change rate; adopts a preset time window to intercept the delay sequence to generate a delay waveform, counts the numerical combination frequency of any two types of sequences to construct a correlation probability matrix, and draws a touch point curve from the touch point position; compares the delay waveform and the touch point curve with a pre-stored normal mode library respectively, calculates the morphological abnormality degree and the state deviation probability; and performs weighted fusion on the morphological abnormality degree and the state deviation probability to obtain a joint abnormal scalar, and outputs a data report when the joint abnormal scalar exceeds a preset threshold value, thereby overcoming the detection blind area of the traditional single-threshold detection method in the multi-parameter collaborative abnormal scene, and improving the comprehensiveness and accuracy of touch response delay abnormal detection.

Description

Technical Field

[0001] This invention relates to the field of big data analytics, specifically to a method and system for analyzing touch response latency data based on time series. Background Technology

[0002] Touch response latency is one of the core metrics for evaluating the interactive experience of touch devices, as its level directly determines the perceived smoothness of operation. In touch interaction scenarios, controlling touch latency and detecting anomalies have always been key areas of focus in the industry.

[0003] Currently, the main approach to touch response latency detection is threshold-based. However, in real-world touch usage scenarios, the perceived degradation in interaction is often not due to a single parameter exceeding its threshold, but rather the synergistic effect of multiple parameters within a specific time window. For example, touch latency itself is not the only factor affecting smoothness; fluctuations in the touchscreen's sampling rate, abnormal continuity of touch point sequences, and unexpected changes in the touch area also dynamically impact responsiveness. When a brief peak in touch latency coincides with a momentary drop in the touch sampling rate due to noise interference, and simultaneously the touch point position sequence index declines, the user will perceive noticeable stuttering or frame skipping. However, a single threshold detection mechanism might classify the latency peak as normal fluctuation, the decrease in sampling rate as occasional sensor noise, and the decrease in touch point continuity as normal touch point variation. Since each of these factors is within an acceptable range, this situation might not be flagged, thus affecting the device's smoothness. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for analyzing touch response delay data based on time series, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a touch response delay data analysis method based on time series, the touch response delay data analysis method comprising the following steps: Step S1: Collect touch response delay, touch sampling rate fluctuation, touch position sequence, and touch area change rate to construct four types of time series data. The four types of time series data include touch response delay sequence, touch sampling rate fluctuation sequence, touch position sequence, and touch area change rate sequence. Step S2: Use a time window of preset length to capture the touch response delay sequence, generate a delay waveform, and at the same time, count the numerical combinations of sequence data between any two types at the same timestamp to construct an association probability matrix. Draw the touch point curve through the touch point position sequence. Step S3: Calculate the morphological anomaly degree and state deviation probability of the current time window delay waveform and touch point curve respectively through the pre-stored normal mode library; The pre-stored normal mode library is formed by collecting several sets of touch response delay sequence samples under the conditions that the touch panel is free from abnormal lag, interference noise, and sufficient system resources, generating corresponding normal delay waveform templates and direction angle change rates, thus constituting the normal mode library. Step S4: Weight and fuse the morphological anomaly degree and the state deviation probability to obtain a joint anomaly scalar. When the joint anomaly scalar exceeds the system's preset threshold, a data report is generated and output. The system's preset threshold is determined through historical normal records, and can also be manually set and adjusted according to actual application scenarios.

[0006] Furthermore, the specific steps of step S1 are as follows: The raw data output by the touch panel is collected, including the timestamp, planar coordinates, contact area, touch sampling rate, and total response delay for each touch point. Based on the timestamp, the total response delay, touch sampling rate, and contact area of ​​each received touch point are time-aligned and mapped onto a unified time axis scale, forming four types of initial time series signals with fixed time intervals as sampling grids. The touch response delay sequence directly takes the delay value at each alignment moment, the touch sampling rate fluctuation sequence is obtained by differential operation of the touch sampling rates at adjacent sampling moments, the touch point position sequence is composed of the coordinate positions of adjacent exit points, and the touch area change rate sequence is determined by the difference of the contact area values ​​at adjacent moments.

[0007] Furthermore, the specific steps of step S2 are as follows: A time window of preset length is used to extract the touch response delay sequence. The deviation value is obtained by subtracting the arithmetic mean of all total response delays within the time window from each total response delay within the time window. The deviation values ​​are then sorted according to the timestamp to obtain the delay deviation sequence. The difference between the global maximum and global minimum total response delays within the time window is used as the normalization denominator. Each deviation value is divided by the denominator, and the calculation results are sorted according to the timestamp to generate the delay waveform. The touch response delay sequence, touch sampling rate fluctuation sequence, touch position sequence, and touch area change rate sequence are labeled as the first type sequence, the second type sequence, the third type sequence, and the fourth type sequence, respectively. The values ​​of any m-th type sequence and the n-th type sequence at the same timestamp are combined to form a value pair. The occurrence frequency of the value pair is counted, and the occurrence frequency is divided by the total number of all value pairs to obtain the joint state probability value. All the obtained joint state probability values ​​are filled into a 4x4 matrix with row index m and column index n. When m and n are equal, the values ​​at the corresponding positions in the matrix are set to zero to obtain the association probability matrix. Extract the horizontal and vertical coordinates of the touch point on the plane corresponding to each timestamp. Calculate the horizontal and vertical displacements based on the difference between adjacent timestamps. The horizontal and vertical displacements form a displacement vector. Extract the total response delay for the same timestamp. Multiply the displacement vector by the total response delay to obtain a delay-weighted displacement vector. Accumulate the delay-weighted displacement vectors calculated for consecutive timestamps to obtain the cumulative displacement vector from the start time to the current time. Record the coordinates of the endpoint of the cumulative displacement vector for each timestamp on the plane and connect them in chronological order to obtain the touch point curve.

[0008] Furthermore, the specific steps of step S3 are as follows: Read all normal delay waveform templates with the same length as the current delay waveform from the pre-stored normal pattern library. Calculate the difference between the current delay waveform and the corresponding position in each normal delay waveform template. Sum the absolute values ​​of the differences at all corresponding positions to obtain the sum of the absolute differences between the normal delay waveform template and the current delay waveform. Iterate through all normal delay waveform templates in the normal pattern library and take the minimum value among all sums of absolute differences. Use this minimum value as the initial morphological difference value. Obtain the difference between the global maximum value and the global minimum value in the current delay waveform as the amplitude range. Divide the initial morphological difference value by this amplitude range to obtain the morphological anomaly degree. The morphological anomaly is a quantitative indicator that measures the degree of difference between the shape characteristics of the touch response delay waveform within the current time window and the delay waveform template under normal operating conditions. Extract the planar horizontal and vertical coordinates corresponding to each time point from the touch point curve. Starting from the second time point, calculate the difference between the horizontal coordinate of the current time point and the horizontal coordinate of the previous time point to obtain the horizontal displacement. Calculate the difference between the vertical coordinate of the current time point and the vertical coordinate of the previous time point to obtain the vertical displacement. The horizontal and vertical displacements constitute the displacement vector of the current time point. Calculate the angle between this displacement vector and the positive horizontal direction, and use this angle as the instantaneous motion direction angle of the current time point. After calculating the instantaneous motion direction angle of each time point, calculate the absolute value of the difference between the instantaneous motion direction angles of two adjacent time points to obtain the direction angle change rate sequence. Obtain the statistical mean and statistical standard deviation of the direction angle change rate sequence in the pre-stored normal mode library. Subtract the statistical mean from each value in the current direction angle change rate sequence, and then divide by the statistical standard deviation to obtain the standardized deviation corresponding to that value. Take the absolute value of all standardized deviations and calculate their arithmetic mean to obtain the state deviation probability. The state deviation probability is a quantitative indicator that measures the degree to which the movement direction change characteristics of the touch point deviate from the statistical distribution of the direction angle change rate under normal working conditions within the current time window.

[0009] Furthermore, the specific steps of step S4 are as follows: Obtain the correlation probability matrix. Extract the three non-zero elements (excluding the main diagonal elements) from the first row of the correlation probability matrix, which represent the joint probability values ​​of the first type sequence with the second, third, and fourth type sequences respectively. Calculate the arithmetic mean of these three values ​​to obtain the delayed correlation coefficient. Then, extract all non-main diagonal elements from the matrix (excluding the first row), which represent the joint probability values ​​between the second, third, and fourth type sequences and between them and the first type sequence. Calculate the arithmetic mean of all non-main diagonal elements in the entire matrix to obtain the overall correlation coefficient. Divide the delayed correlation coefficient by the sum of the overall correlation coefficient and the delayed correlation coefficient to obtain the first fusion weight. Divide the overall correlation coefficient by the sum of the overall correlation coefficient and the delayed correlation coefficient to obtain the second fusion weight. Multiply the morphological anomaly degree by the first fusion weight to obtain the weighted morphological anomaly component; multiply the state deviation probability by the second fusion weight to obtain the weighted state deviation component; add the weighted morphological anomaly component and the weighted state deviation component to obtain the joint anomaly scalar. The joint anomaly scalar is a comprehensive anomaly determination index obtained by weighted fusion of morphological anomaly degree and state deviation probability. The joint anomaly scalar is compared with the system's preset threshold. If the current joint anomaly scalar is not greater than the system's preset threshold, it is not marked as an anomaly. If the current joint anomaly scalar is greater than the system's preset threshold, it is marked as an anomaly. At the same time, the start and end timestamps of the time window, the morphological anomaly degree, and the state deviation probability are recorded, and a data report is generated and output.

[0010] Furthermore, the touch response delay data analysis system includes a data acquisition module, a feature extraction module, an anomaly assessment module, and a fusion determination module; The data acquisition module is used to collect the raw data output by the touch panel and perform time axis alignment operations on the total response delay, touch sampling rate and contact area of ​​each touch point; the feature extraction module is used to extract the touch response delay sequence to generate delay waveform and touch point curve, and to count the frequency of numerical combinations between any two types of sequences to construct an association probability matrix; the anomaly evaluation module is used to compare the delay waveform and touch point curve of the current time window with the pre-stored normal mode library to calculate the morphological anomaly degree and state deviation probability; the fusion judgment module is used to perform weighted fusion of morphological anomaly degree and state deviation probability to generate a joint anomaly scalar, and to output a data report after comparing the scalar with a preset threshold.

[0011] Furthermore, the data acquisition module includes a raw data acquisition unit and a time alignment unit; The raw data acquisition unit is used to acquire the timestamp, planar coordinates, contact area, touch sampling rate and total response delay corresponding to each touch point output by the touch panel; the time alignment unit is used to map the data of each touch point to a unified time axis scale with a fixed time interval according to the timestamp, forming a touch response delay sequence, a touch sampling rate fluctuation sequence, a touch point position sequence and a touch area change rate sequence.

[0012] Furthermore, the feature extraction module includes a waveform trajectory generation unit and an association probability calculation unit; The waveform trajectory generation unit is used to center and normalize the delay values ​​within the intercepted time window to generate a delayed waveform, and to draw the touch point curve based on the cumulative displacement vector; the correlation probability calculation unit is used to count the number of times any two types of sequences appear in the same timestamp value combination, and fill it into the matrix to construct the correlation probability matrix.

[0013] Furthermore, the anomaly assessment module includes a morphological anomaly measurement unit and a state deviation measurement unit; The morphological anomaly measurement unit is used to calculate the sum of the minimum absolute differences between the current delayed waveform and all templates in the normal mode library, and divide by the amplitude range to obtain the morphological anomaly degree; the state deviation measurement unit is used to calculate the state deviation probability based on the instantaneous motion direction angle change rate of the current touch point curve and its normal statistical parameters.

[0014] Furthermore, the fusion determination module includes a weighted fusion unit and an anomaly report generation unit. The weighted fusion unit is used to extract the delayed correlation coefficient and the overall correlation coefficient from the correlation probability matrix to calculate the fusion weight, and to perform a weighted summation on the morphological anomaly degree and the state deviation probability; the anomaly report generation unit is used to record the start and end timestamps, morphological anomaly degree and state deviation probability of the time window when the joint anomaly scalar exceeds the preset threshold, and to summarize and output the data report.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention overcomes the limitations of traditional methods that rely solely on a single delay indicator for judgment by integrating four types of time series data: touch response delay, touch sampling rate fluctuation, touch position sequence, and touch area change rate. It can comprehensively perceive the touch interaction status from multiple dimensions such as touch response timeliness, data acquisition continuity, motion trajectory smoothness, and contact state stability.

[0016] 2. This invention constructs a correlation probability matrix between any two types of sequences to quantitatively capture the collaborative change relationship between touch response delay and sampling rate fluctuation, touch position jump, and area mutation. It also uses this matrix to dynamically calculate the fusion weight of morphological anomaly degree and state deviation probability, so that multiple parameters can be accurately identified when they are collaboratively abnormal within the same time window. This effectively solves the detection blind spot problem of existing technologies in multi-parameter joint anomaly scenarios.

[0017] 3. This invention uses a weighted fusion of the degree of morphological distortion of the delayed waveform and the statistical deviation of the rate of change of the touch point movement direction angle to form a unified joint anomaly scalar for judgment. It integrates anomaly signals from multiple dimensions into a comprehensive quantitative index. At the same time, it improves the robustness of the method under different touch scenarios and device conditions through adaptive weight allocation, and reduces the false alarm rate and false negative rate. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the touch response delay data analysis method based on time series according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1: As Figure 1 As shown, this invention provides a technical solution: a touch response delay data analysis method based on time series. The touch response delay data analysis method includes the following steps: Step S1: Collect touch response delay, touch sampling rate fluctuation, touch position sequence, and touch area change rate to construct four types of time series data. The four types of time series data include touch response delay sequence, touch sampling rate fluctuation sequence, touch position sequence, and touch area change rate sequence. The specific steps of step S1 are as follows: The raw data output by the touch panel is collected, including the timestamp, planar coordinates, contact area, touch sampling rate, and total response delay for each touch point. Based on the timestamp, the total response delay, touch sampling rate, and contact area of ​​each received touch point are time-aligned and mapped onto a unified time axis scale, forming four types of initial time series signals with fixed time intervals as sampling grids. The touch response delay sequence directly takes the delay value at each alignment moment, the touch sampling rate fluctuation sequence is obtained by differential operation of the touch sampling rates at adjacent sampling moments, the touch point position sequence is composed of the coordinate positions of adjacent exit points, and the touch area change rate sequence is determined by the difference of the contact area values ​​at adjacent moments.

[0021] Step S2: Use a time window of preset length to capture the touch response delay sequence, generate a delay waveform, and at the same time, count the numerical combinations of sequence data between any two types at the same timestamp to construct an association probability matrix. Draw the touch point curve through the touch point position sequence. The specific steps of step S2 are as follows: A time window of preset length is used to extract the touch response delay sequence. The deviation value is obtained by subtracting the arithmetic mean of all total response delays within the time window from each total response delay within the time window. The deviation values ​​are then sorted according to the timestamp to obtain the delay deviation sequence. The difference between the global maximum and global minimum total response delays within the time window is used as the normalization denominator. Each deviation value is divided by the denominator, and the calculation results are sorted according to the timestamp to generate the delay waveform. The touch response delay sequence, touch sampling rate fluctuation sequence, touch position sequence, and touch area change rate sequence are labeled as the first type sequence, the second type sequence, the third type sequence, and the fourth type sequence, respectively. The values ​​of any m-th type sequence and the n-th type sequence at the same timestamp are combined to form a value pair. The occurrence frequency of the value pair is counted, and the occurrence frequency is divided by the total number of all value pairs to obtain the joint state probability value. All the obtained joint state probability values ​​are filled into a 4x4 matrix with row index m and column index n. When m and n are equal, the values ​​at the corresponding positions in the matrix are set to zero to obtain the association probability matrix. Extract the horizontal and vertical coordinates of the touch point on the plane corresponding to each timestamp. Calculate the horizontal and vertical displacements based on the difference between adjacent timestamps. The horizontal and vertical displacements form a displacement vector. Extract the total response delay for the same timestamp. Multiply the displacement vector by the total response delay to obtain a delay-weighted displacement vector. Accumulate the delay-weighted displacement vectors calculated for consecutive timestamps to obtain the cumulative displacement vector from the start time to the current time. Record the coordinates of the endpoint of the cumulative displacement vector for each timestamp on the plane and connect them in chronological order to obtain the touch point curve.

[0022] Step S3: Calculate the morphological anomaly degree and state deviation probability of the current time window delay waveform and touch point curve respectively through the pre-stored normal mode library; The specific steps of step S3 are as follows: Read all normal delay waveform templates with the same length as the current delay waveform from the pre-stored normal pattern library. Calculate the difference between the current delay waveform and the corresponding position in each normal delay waveform template. Sum the absolute values ​​of the differences at all corresponding positions to obtain the sum of the absolute differences between the normal delay waveform template and the current delay waveform. Iterate through all normal delay waveform templates in the normal pattern library and take the minimum value among all sums of absolute differences. Use this minimum value as the initial morphological difference value. Obtain the difference between the global maximum value and the global minimum value in the current delay waveform as the amplitude range. Divide the initial morphological difference value by this amplitude range to obtain the morphological anomaly degree. Extract the planar horizontal and vertical coordinates corresponding to each time point from the touch point curve. Starting from the second time point, calculate the difference between the horizontal coordinate of the current time point and the horizontal coordinate of the previous time point to obtain the horizontal displacement. Calculate the difference between the vertical coordinate of the current time point and the vertical coordinate of the previous time point to obtain the vertical displacement. The horizontal and vertical displacements constitute the displacement vector of the current time point. Calculate the angle between this displacement vector and the positive horizontal direction, and use this angle as the instantaneous motion direction angle of the current time point. After calculating the instantaneous motion direction angle of each time point, calculate the absolute value of the difference between the instantaneous motion direction angles of two adjacent time points to obtain the direction angle change rate sequence. Obtain the statistical mean and statistical standard deviation of the direction angle change rate sequence in the pre-stored normal mode library. Subtract the statistical mean from each value in the current direction angle change rate sequence, and then divide by the statistical standard deviation to obtain the standardized deviation corresponding to that value. Take the absolute value of all standardized deviations and calculate their arithmetic mean to obtain the state deviation probability.

[0023] Step S4: Weight and fuse the morphological anomaly degree and the state deviation probability to obtain a joint anomaly scalar. When the joint anomaly scalar exceeds the system's preset threshold, a data report is generated and output. The specific steps of step S4 are as follows: Obtain the correlation probability matrix. Extract the three non-zero elements (excluding the main diagonal elements) from the first row of the correlation probability matrix, which represent the joint probability values ​​of the first type sequence with the second, third, and fourth type sequences respectively. Calculate the arithmetic mean of these three values ​​to obtain the delayed correlation coefficient. Then, extract all non-main diagonal elements from the matrix (excluding the first row), which represent the joint probability values ​​between the second, third, and fourth type sequences and between them and the first type sequence. Calculate the arithmetic mean of all non-main diagonal elements in the entire matrix to obtain the overall correlation coefficient. Divide the delayed correlation coefficient by the sum of the overall correlation coefficient and the delayed correlation coefficient to obtain the first fusion weight. Divide the overall correlation coefficient by the sum of the overall correlation coefficient and the delayed correlation coefficient to obtain the second fusion weight. Multiply the morphological anomaly degree by the first fusion weight to obtain the weighted morphological anomaly component; multiply the state deviation probability by the second fusion weight to obtain the weighted state deviation component; add the weighted morphological anomaly component and the weighted state deviation component to obtain the joint anomaly scalar. The joint anomaly scalar is compared with the system's preset threshold. If the current joint anomaly scalar is not greater than the system's preset threshold, it is not marked as an anomaly. If the current joint anomaly scalar is greater than the system's preset threshold, it is marked as an anomaly. At the same time, the start and end timestamps of the time window, the morphological anomaly degree, and the state deviation probability are recorded, and a data report is generated and output.

[0024] For example, the touch performance monitoring service of the device under test is activated, continuously collecting raw data output from the touch panel through the system's underlying interface. The user performs a straight-line swipe operation from left to right in the central area of ​​the screen at a speed of approximately 20 mm / s for 2 seconds. The data acquisition module collects the touch response delay sequence, touch sampling rate fluctuation sequence, touch point position sequence, and touch area change rate sequence at a fixed time interval Δt = 8.33 ms.

[0025] In this embodiment, the collected touch response delay sequence data is collected over 10 consecutive sampling times, with timestamps of 0ms, 8.33ms, 16.67ms, 25.00ms, 33.33ms, 41.67ms, 50.00ms, 58.33ms, 66.67ms, and 75.00ms respectively; the touch response delays are 32, 35, 38, 42, 45, 48, 52, 55, 58, and 62; and the corresponding touch sampling rates are 118Hz, 120Hz, 119Hz, 115Hz, 112Hz, 108Hz, 105Hz, 102Hz, 98Hz, and 95Hz respectively. The touch sampling rate fluctuation sequence is obtained by differential calculation of the touch sampling rates at adjacent sampling times, and is +2Hz, -1Hz, -4Hz, -3Hz, -4Hz, -3Hz, -3Hz, -4Hz, and -3Hz respectively.

[0026] In the touch point position sequence, the planar coordinates of adjacent timestamps are (100.0, 500), (100.8, 500), (101.6, 500), (102.4, 500), (103.2, 500), (104.0, 500), (104.8, 500), (105.6, 500), (106.4, 500), and (107.2, 500). The touch area change rate sequence is determined by the difference in contact area values ​​between adjacent times. In this embodiment, the contact area remains stable at each time point, and the area change rate is 0.

[0027] The touch response delay sequence was extracted using a time window of preset length L=10, resulting in the aforementioned 10 delay values. The arithmetic mean of all delay values ​​within this time window was calculated to be 46.7ms.

[0028] Subtracting the average value from each delay value yields a sequence of deviation values: -14.7, -11.7, -8.7, -4.7, -1.7, 1.3, 5.3, 8.3, 11.3, and 15.3. The difference between the global maximum delay of 62ms and the global minimum delay of 32ms within this time window is 30ms. Using this difference as the normalization denominator, each deviation value is divided by 30, resulting in a delay waveform sequence of -0.49, -0.39, -0.29, -0.16, -0.06, 0.04, 0.18, 0.28, 0.38, and 0.51.

[0029] Simultaneously, the frequency of numerical combinations of the four types of sequences at the same timestamp was statistically analyzed. Taking the touch response delay sequence and the touch sampling rate fluctuation sequence as examples, the numerical pairs formed at 10 timestamps are (-14.7, +2), (-11.7, -1), (-8.7, -4), (-4.7, -3), (-1.7, -4), (1.3, -3), (5.3, -3), (8.3, -4), (11.3, -3), and (15.3, -3). After counting the occurrences of each numerical pair, the result is divided by the total number of numerical pairs (10) to obtain the joint state probability value. This process is repeated to construct a complete 4×4 correlation probability matrix. In this matrix, the first row contains 0, 0.15, 0.12, and 0.08; the second row contains 0.15, 0, 0.10, and 0.06; the third row contains 0.12, 0.10, 0, and 0.04; and the fourth row contains 0.08, 0.06, 0.04, and 0.

[0030] Extract the planar horizontal and vertical coordinates corresponding to each timestamp, and calculate the displacement vector of adjacent timestamps. From timestamp 0 to 8.33ms, the horizontal displacement is 100.8 - 100.0 = 0.8 pixels, and the vertical displacement is 500 - 500 = 0 pixels, forming the displacement vector (0.8, 0). Extract the total response delay of 32ms for the same timestamp, and multiply the displacement vector by the total response delay to obtain the delay-weighted displacement vector (25.6, 0). Accumulate the delay-weighted displacement vectors under consecutive timestamps to obtain the cumulative displacement vector sequence. Record the endpoint coordinates of each cumulative displacement vector and connect them in chronological order to obtain the touch point curve.

[0031] Extract normal delayed waveform templates of length 10 from the normal pattern library. The template sequence closest to the current delayed waveform is -0.45, -0.35, -0.25, -0.15, -0.05, 0.05, 0.15, 0.25, 0.35, 0.45. Calculate the sum of the absolute differences between the current delayed waveform and the corresponding positions of these templates, which equals 0.30.

[0032] After traversing all templates in the normal pattern library, the minimum value of the sum of all absolute differences, 0.30, is taken as the initial morphological difference value. The amplitude range of the current delayed waveform is 1.00, and the morphological anomaly is 0.30.

[0033] Extract the planar coordinates corresponding to each timestamp from the touch point curve and calculate the instantaneous motion direction angle. Since the user performs a horizontal straight-line swipe in this embodiment, the displacement vector at each moment is (0.8, 0), and the angle with the positive horizontal direction is 0°. Therefore, the entire direction angle change rate sequence is 0. Obtain the statistical average of the direction angle change rate sequence (2.5° / ms) and the statistical standard deviation (1.2° / ms) from the pre-stored normal mode library. Subtract the statistical average of 2.5 from each value in the current direction angle change rate sequence and divide by the statistical standard deviation of 1.2 to obtain the standardized deviation corresponding to each timestamp, which is -2.08. Calculate the arithmetic mean of the absolute values ​​of all standardized deviations to obtain the state deviation probability of 2.08.

[0034] The three non-zero elements (excluding the main diagonal) extracted from the first row of the correlation probability matrix are 0.15, 0.12, and 0.08. Their arithmetic mean yields a delayed correlation coefficient of 0.117. The arithmetic mean of all non-main diagonal elements extracted from the matrix yields an overall correlation coefficient of 0.092.

[0035] The first fusion weight is 0.56; the second fusion weight is 0.44.

[0036] Multiplying the morphological anomaly degree of 0.30 by the first fusion weight of 0.56 yields a weighted morphological anomaly component of 0.168; multiplying the state deviation probability of 2.08 by the second fusion weight of 0.44 yields a weighted state deviation component of 0.915; the joint anomaly scalar is 0.168 + 0.915 = 1.083.

[0037] The system has a preset threshold T=1.00. The current joint anomaly scalar of 1.083 is greater than the system's preset threshold of 1.00, therefore the current time window is determined to be an anomaly. The system records the start timestamp of this time window (0ms) and end timestamp (75ms), the morphological anomaly degree (0.30), and the state deviation probability (2.08), and outputs a data report.

[0038] In another embodiment, a complex sliding trajectory with sharp turns is executed to verify the anomaly detection capability of the present invention in scenarios with abrupt changes in contact direction.

[0039] The user completes three directional swipe operations with sharp turns within 2 seconds, resulting in drastic changes in the displacement vector direction between adjacent timestamps in the touch point position sequence. Taking one segment of the directional change window as an example, the extracted instantaneous motion direction angle sequence is 0°, 2°, 5°, 8°, 12°, 15°, 45°, 80°, 115°, and 150°. Calculating the absolute value of the direction angle difference between adjacent timestamps yields the direction angle change rate sequence as 2° / ms, 3° / ms, 3° / ms, 4° / ms, 3° / ms, 30° / ms, 35° / ms, 35° / ms, and 35° / ms.

[0040] Obtain the statistical mean (3.0° / ms) and standard deviation (1.5° / ms) of the orientation angle change rate sequence from the pre-stored normal pattern library. Standardize the deviations of the first five values ​​(2, 3, 3, 4, 3) in the current orientation angle change rate sequence to -0.67, 0, 0, 0.67, and 0, respectively, and the standardized deviations of the last four values ​​(30, 35, 35, 35) to 18.0, 21.3, 21.3, and 21.3, respectively. Calculate the arithmetic mean of the absolute values ​​of all standardized deviations to obtain the state deviation probability of 9.25.

[0041] Meanwhile, the touch response delay sequence remains relatively stable within this window, with minimal morphological difference between the delay waveform and the normal pattern library, resulting in a morphological anomaly degree of 0.12. The delay correlation coefficient and the overall correlation coefficient are similar to those in Example 2, with a first fusion weight of approximately 0.55 and a second fusion weight of approximately 0.45.

[0042] The weighted morphological anomaly component is 0.066; the weighted state deviation component is 4.163; and the joint anomaly scalar is 4.229. This value is significantly greater than the system's preset threshold of 1.00, and the system determines it as an anomaly and outputs a data report.

[0043] In another embodiment, the present invention provides a time-series-based touch response delay data analysis system, which includes a data acquisition module, a feature extraction module, an anomaly assessment module, and a fusion determination module. The data acquisition module is used to acquire the raw data output by the touch panel and perform time axis alignment operation on the total response delay, touch sampling rate and contact area of ​​each touch point. The data acquisition module includes a raw data acquisition unit and a time alignment unit; The raw data acquisition unit is used to acquire the timestamp, planar coordinates, contact area, touch sampling rate and total response delay corresponding to each touch point output by the touch panel; the time alignment unit is used to map the data of each touch point to a unified time axis scale with a fixed time interval according to the timestamp, forming a touch response delay sequence, a touch sampling rate fluctuation sequence, a touch point position sequence and a touch area change rate sequence.

[0044] The feature extraction module is used to extract the touch response delay sequence to generate delay waveforms and touch point curves, and to count the frequency of numerical combinations between any two types of sequences to construct an association probability matrix; The feature extraction module includes a waveform trajectory generation unit and an association probability calculation unit; The waveform trajectory generation unit is used to center and normalize the delay values ​​within the intercepted time window to generate a delayed waveform, and to draw the touch point curve based on the cumulative displacement vector; the correlation probability calculation unit is used to count the number of times any two types of sequences appear in the same timestamp value combination, and fill it into the matrix to construct the correlation probability matrix.

[0045] The anomaly assessment module is used to compare the delay waveform and touch point curve of the current time window with the pre-stored normal mode library to calculate the morphological anomaly degree and the probability of state deviation. The anomaly assessment module includes a morphological anomaly measurement unit and a state deviation measurement unit; The morphological anomaly measurement unit is used to calculate the sum of the minimum absolute differences between the current delayed waveform and all templates in the normal mode library, and divide by the amplitude range to obtain the morphological anomaly degree; the state deviation measurement unit is used to calculate the state deviation probability based on the instantaneous motion direction angle change rate of the current touch point curve and its normal statistical parameters.

[0046] The fusion determination module is used to perform weighted fusion of morphological anomaly degree and state deviation probability to generate a joint anomaly scalar, and output a data report after comparing the scalar with a preset threshold. The fusion determination module includes a weighted fusion unit and an anomaly report generation unit. The weighted fusion unit is used to extract the delayed correlation coefficient and the overall correlation coefficient from the correlation probability matrix to calculate the fusion weight, and to perform a weighted summation on the morphological anomaly degree and the state deviation probability; the anomaly report generation unit is used to record the start and end timestamps, morphological anomaly degree and state deviation probability of the time window when the joint anomaly scalar exceeds the preset threshold, and to summarize and output the data report.

[0047] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for analyzing touch response delay data based on time series, characterized in that: The touch response delay data analysis method includes the following steps: Step S1: Collect touch response delay, touch sampling rate fluctuation, touch position sequence, and touch area change rate to construct four types of time series data. The four types of time series data include touch response delay sequence, touch sampling rate fluctuation sequence, touch position sequence, and touch area change rate sequence. Step S2: Use a time window of preset length to capture the touch response delay sequence, generate a delay waveform, and at the same time, count the numerical combinations of sequence data between any two types at the same timestamp to construct an association probability matrix. Draw the touch point curve through the touch point position sequence. Step S3: Calculate the morphological anomaly degree and state deviation probability of the current time window delay waveform and touch point curve respectively through the pre-stored normal mode library; Step S4: Weight and fuse the morphological anomaly degree and the state deviation probability to obtain a joint anomaly scalar. When the joint anomaly scalar exceeds the system's preset threshold, a data report is generated and output.

2. The touch response delay data analysis method based on time series according to claim 1, characterized in that: The specific steps of step S1 are as follows: The raw data output by the touch panel is collected, including the timestamp, planar coordinates, contact area, touch sampling rate, and total response delay for each touch point. Based on the timestamp, the total response delay, touch sampling rate, and contact area of ​​each received touch point are time-aligned and mapped onto a unified time axis scale, forming four types of initial time series signals with fixed time intervals as sampling grids. The touch response delay sequence directly takes the delay value at each alignment moment, the touch sampling rate fluctuation sequence is obtained by differential operation of the touch sampling rates at adjacent sampling moments, the touch point position sequence is composed of the coordinate positions of adjacent exit points, and the touch area change rate sequence is determined by the difference of the contact area values ​​at adjacent moments.

3. The touch response delay data analysis method based on time series according to claim 1, characterized in that: The specific steps of step S2 are as follows: A time window of preset length is used to extract the touch response delay sequence. The deviation value is obtained by subtracting the arithmetic mean of all total response delays within the time window from each total response delay within the time window. The deviation values ​​are then sorted according to the timestamp to obtain the delay deviation sequence. The difference between the global maximum and global minimum total response delays within the time window is used as the normalization denominator. Each deviation value is divided by the denominator, and the calculation results are sorted according to the timestamp to generate the delay waveform. The touch response delay sequence, touch sampling rate fluctuation sequence, touch position sequence, and touch area change rate sequence are labeled as the first type sequence, the second type sequence, the third type sequence, and the fourth type sequence, respectively. The values ​​of any m-th type sequence and the n-th type sequence at the same timestamp are combined to form a value pair. The occurrence frequency of the value pair is counted, and the occurrence frequency is divided by the total number of all value pairs to obtain the joint state probability value. All the obtained joint state probability values ​​are filled into a 4x4 matrix with row index m and column index n. When m and n are equal, the values ​​at the corresponding positions in the matrix are set to zero to obtain the association probability matrix. Extract the horizontal and vertical coordinates of the touch point on the plane corresponding to each timestamp. Calculate the horizontal and vertical displacements based on the difference between adjacent timestamps. The horizontal and vertical displacements form a displacement vector. Extract the total response delay for the same timestamp. Multiply the displacement vector by the total response delay to obtain a delay-weighted displacement vector. Accumulate the delay-weighted displacement vectors calculated for consecutive timestamps to obtain the cumulative displacement vector from the start time to the current time. Record the coordinates of the endpoint of the cumulative displacement vector for each timestamp on the plane and connect them in chronological order to obtain the touch point curve.

4. The touch response delay data analysis method based on time series according to claim 1, characterized in that: The specific steps of step S3 are as follows: Read all normal delay waveform templates with the same length as the current delay waveform from the pre-stored normal mode library, calculate the difference between the current delay waveform and the corresponding position in each normal delay waveform template, and sum the absolute values ​​of the differences at all corresponding positions to obtain the sum of the absolute differences between the normal delay waveform template and the current delay waveform. Traverse all normal delayed waveform templates in the normal mode library, take the minimum value among all the sums of absolute differences, and use this minimum value as the initial morphological difference value; obtain the difference between the global maximum value and the global minimum value in the current delayed waveform as the amplitude range, divide the initial morphological difference value by this amplitude range, and obtain the morphological anomaly degree. Extract the planar horizontal and vertical coordinates corresponding to each time point from the touch point curve. Starting from the second time point, calculate the difference between the horizontal coordinate of the current time point and the horizontal coordinate of the previous time point to obtain the horizontal displacement. Calculate the difference between the vertical coordinate of the current time point and the vertical coordinate of the previous time point to obtain the vertical displacement. The horizontal and vertical displacements constitute the displacement vector of the current time point. Calculate the angle between this displacement vector and the positive horizontal direction, and use this angle as the instantaneous motion direction angle of the current time point. After calculating the instantaneous motion direction angle of each time point, calculate the absolute value of the difference between the instantaneous motion direction angles of two adjacent time points to obtain the direction angle change rate sequence. Obtain the statistical mean and statistical standard deviation of the direction angle change rate sequence in the pre-stored normal mode library. Subtract the statistical mean from each value in the current direction angle change rate sequence, and then divide by the statistical standard deviation to obtain the standardized deviation corresponding to that value. Take the absolute value of all standardized deviations and calculate their arithmetic mean to obtain the state deviation probability.

5. The touch response delay data analysis method based on time series according to claim 1, characterized in that: The specific steps of step S4 are as follows: Obtain the correlation probability matrix. Extract the three non-zero elements (excluding the main diagonal elements) from the first row of the correlation probability matrix, which represent the joint probability values ​​of the first type sequence with the second, third, and fourth type sequences respectively. Calculate the arithmetic mean of these three values ​​to obtain the delayed correlation coefficient. Then, extract all non-main diagonal elements from the matrix (excluding the first row), which represent the joint probability values ​​between the second, third, and fourth type sequences and between them and the first type sequence. Calculate the arithmetic mean of all non-main diagonal elements in the entire matrix to obtain the overall correlation coefficient. Divide the delayed correlation coefficient by the sum of the overall correlation coefficient and the delayed correlation coefficient to obtain the first fusion weight. Divide the overall correlation coefficient by the sum of the overall correlation coefficient and the delayed correlation coefficient to obtain the second fusion weight. Multiply the morphological anomaly degree by the first fusion weight to obtain the weighted morphological anomaly component; Multiply the state deviation probability by the second fusion weight to obtain the weighted state deviation component; add the weighted morphological anomaly component and the weighted state deviation component to obtain the joint anomaly scalar. The joint anomaly scalar is compared with the system's preset threshold. If the current joint anomaly scalar is not greater than the system's preset threshold, it is not marked as an anomaly. If the current joint anomaly scalar is greater than the system's preset threshold, it is marked as an anomaly. At the same time, the start and end timestamps of the time window, the morphological anomaly degree, and the state deviation probability are recorded, and a data report is generated and output.

6. A time-series-based touch response delay data analysis system, applied to the time-series-based touch response delay data analysis method according to any one of claims 1-5, characterized in that: The touch response delay data analysis system includes a data acquisition module, a feature extraction module, an anomaly assessment module, and a fusion determination module. The data acquisition module is used to collect the raw data output by the touch panel and perform time axis alignment operations on the total response delay, touch sampling rate and contact area of ​​each touch point; the feature extraction module is used to extract the touch response delay sequence to generate delay waveform and touch point curve, and to count the frequency of numerical combinations between any two types of sequences to construct an association probability matrix; the anomaly evaluation module is used to compare the delay waveform and touch point curve of the current time window with the pre-stored normal mode library to calculate the morphological anomaly degree and state deviation probability; the fusion judgment module is used to perform weighted fusion of morphological anomaly degree and state deviation probability to generate a joint anomaly scalar, and to output a data report after comparing the scalar with a preset threshold.

7. The touch response delay data analysis system based on time series according to claim 6, characterized in that: The data acquisition module includes a raw data acquisition unit and a time alignment unit; The raw data acquisition unit is used to acquire the timestamp, planar coordinates, contact area, touch sampling rate and total response delay corresponding to each touch point output by the touch panel; the time alignment unit is used to map the data of each touch point to a unified time axis scale with a fixed time interval according to the timestamp, forming a touch response delay sequence, a touch sampling rate fluctuation sequence, a touch point position sequence and a touch area change rate sequence.

8. The touch response delay data analysis system based on time series according to claim 6, characterized in that: The feature extraction module includes a waveform trajectory generation unit and an association probability calculation unit; The waveform trajectory generation unit is used to center and normalize the delay values ​​within the intercepted time window to generate a delayed waveform, and to draw the touch point curve based on the cumulative displacement vector. The correlation probability calculation unit is used to count the number of times any two types of sequences appear in the same combination of timestamp values, and fill the matrix to construct the correlation probability matrix.

9. The touch response delay data analysis system based on time series according to claim 6, characterized in that: The anomaly assessment module includes a morphological anomaly measurement unit and a state deviation measurement unit; The morphological anomaly measurement unit is used to calculate the sum of the minimum absolute differences between the current delayed waveform and all templates in the normal mode library, and divide by the amplitude range to obtain the morphological anomaly degree; the state deviation measurement unit is used to calculate the state deviation probability based on the instantaneous motion direction angle change rate of the current touch point curve and its normal statistical parameters.

10. The touch response delay data analysis system based on time series according to claim 6, characterized in that: The fusion determination module includes a weighted fusion unit and an anomaly report generation unit. The weighted fusion unit is used to extract the delayed correlation coefficient and the overall correlation coefficient from the correlation probability matrix to calculate the fusion weight, and to perform a weighted summation on the morphological anomaly degree and the state deviation probability; the anomaly report generation unit is used to record the start and end timestamps, morphological anomaly degree and state deviation probability of the time window when the joint anomaly scalar exceeds the preset threshold, and to summarize and output the data report.

Images