A vibration detection method and device for a force-sensing touch panel of a notebook computer

Abstract

The invention relates to the field of vibration detection technology, and discloses a vibration detection method and device for a force-sensing touch pad of a notebook computer. The method comprises collecting an original signal of the force-sensing touch pad based on an acceleration sensor, filtering the original signal to obtain an acceleration signal of the force-sensing touch pad, performing multi-level decomposition on the amplitude and phase information of each frequency component, converting all feature data of a multi-scale feature matrix into standard data with zero mean and unit variance, performing covariance matrix calculation on the standardized acceleration feature data, performing classification prediction processing on the reduced-dimensional feature matrix, obtaining a classification result and a confidence score, and generating a detection report in combination with the classification result and the confidence score; the vibration characteristics of the force-sensing touch pad in different areas are accurately captured, and the amplitude and phase characteristics of the vibration signal of the force-sensing touch pad can be accurately analyzed, thereby comprehensively reflecting the vibration intensity distribution in different areas, and effectively enhancing the detection capability of the vibration intensity deviating from the design value.

Description

Technical Field

[0001] The present invention relates to the technical field of vibration detection, and in particular to a vibration detection method and device for a force-sensing touch panel of a notebook computer. Background Art

[0002] In today's era of rapid technological development, laptops, as an important productivity tool, are increasingly used in work, study, and entertainment. As users' requirements for operating experience continue to increase, touch technology has developed rapidly. Forcepad, as a new type of touchpad based on pressure sensing technology, can achieve multi-dimensional interactive operations by detecting the amount of pressure applied by the user, significantly improving ease of use and operating efficiency. In order to ensure that Forcepad has stable performance and consistent user experience when it leaves the factory, the detection of its vibration characteristics has become a core link in product quality control. The goal of the vibration characteristics test is to verify whether the Forcepad has the expected vibration intensity and uniformity in the feedback response, so as to provide users with a reliable operating experience.

[0003] At present, the vibration performance detection of force-sensing touch panels mainly adopts the collection and analysis of single-point vibration signals, which can detect whether the function of the force-sensing touch panel is normal and analyze the sensitivity. However, it cannot fully reflect the vibration characteristics of the force-sensing touch panel in different areas, especially there are deficiencies in the detection of vibration intensity. When the feedback intensity deviates from the design value, it cannot be discovered in time or accurately quantified, which can easily cause users to have illusions or feedback delays during operation, thereby reducing the accuracy of the operation and user experience.

[0004] Therefore, it is necessary to provide a vibration detection method and device for a force-sensing touch pad of a notebook computer, so as to solve the problem that the vibration intensity detection of the force-sensing touch pad is insufficient. Summary of the invention

[0005] The main purpose of the present invention is to provide a vibration detection method and device for a force-sensitive touch pad of a notebook computer, aiming to solve the technical problems mentioned in the above background technology.

[0006] The present invention adopts the following technical solutions:

[0007] A vibration detection method for a force-sensing touch pad of a notebook computer, comprising:

[0008] Collecting the original signal of the force-sensing touch panel based on the acceleration sensor, filtering the original signal to obtain the acceleration signal of the force-sensing touch panel, wherein the acceleration signal includes acceleration data of three axes: X, Y, and Z;

[0009] Performing time domain analysis on the acceleration signal, and converting the acceleration signal into a frequency domain signal to obtain amplitude and phase information of each frequency component;

[0010] Performing multi-level decomposition on the amplitude and phase information of each frequency component, and extracting multi-scale time-frequency features in the acceleration signal to obtain a multi-scale feature matrix of the acceleration signal;

[0011] Converting all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data;

[0012] Calculate the covariance matrix of the normalized acceleration feature data, sort the calculated multiple eigenvalues, select the eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting results, and establish a dimension reduction feature matrix according to the eigenvectors;

[0013] The reduced-dimensional feature matrix is ​​subjected to classification prediction processing to obtain a classification result and a confidence score, and a detection report is generated in combination with the classification result and the confidence score.

[0014] Furthermore, the step of collecting the original signal of the force-sensitive touch panel based on the acceleration sensor and filtering the original signal to obtain the acceleration signal of the force-sensitive touch panel specifically includes:

[0015] Placing the acceleration sensor in contact with the force-sensing touch panel and initializing the acceleration sensor;

[0016] Acquire a timestamp of the vibration of the force-sensing touch panel, collect an original signal of the force-sensing touch panel based on the acceleration sensor, perform denoising on the original signal, filter out low-frequency noise and high-frequency noise of the original signal, and obtain a denoised acceleration signal;

[0017] Performing dynamic range adjustment on the denoised acceleration signal to eliminate amplitude deviation in the denoised acceleration signal and obtain an acceleration signal of uniform magnitude;

[0018] The uniform magnitude acceleration signal is time-aligned with the timestamp to obtain the time-aligned acceleration signal.

[0019] Furthermore, the step of performing time domain analysis processing on the acceleration signal and converting the acceleration signal into a frequency domain signal to obtain amplitude and phase information of each frequency component specifically includes:

[0020] Performing frame processing on the acceleration signal according to a preset time window length and a preset overlap rate to obtain a plurality of time window signals;

[0021] Performing time domain analysis and processing on each of the time window signals to obtain vibration characteristic data within each time window;

[0022] Performing fast Fourier transform on the vibration characteristic data in each time window to obtain frequency components of each frequency band;

[0023] A modular operation is performed on the complex value of each frequency component to obtain the amplitude of each frequency component, and a complex angle of each frequency component is calculated to obtain the phase information of each frequency component.

[0024] Furthermore, the step of performing multi-level decomposition on the amplitude and phase information of each of the frequency components and extracting the multi-scale time-frequency features in the acceleration signal to obtain the multi-scale feature matrix of the acceleration signal specifically includes:

[0025] Performing wavelet transform processing on the amplitude and phase information of each frequency component to obtain multi-level wavelet coefficients of the frequency domain signal;

[0026] Performing inverse transformation on the wavelet coefficients to reconstruct the time-frequency diagram of the acceleration signal at each scale;

[0027] Refining the time-frequency graph by thresholding and peak detection to extract characteristic components at multiple scales;

[0028] The characteristic components at each scale are standardized to obtain time-frequency characteristic data;

[0029] The time-frequency feature data is divided into multiple time windows according to time and frequency, and the feature data in each time window is organized into a multi-scale feature matrix according to scale.

[0030] Furthermore, the step of converting all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data specifically includes:

[0031] Performing column mean calculation on all feature data of the multi-scale feature matrix to obtain the mean of each column of feature data;

[0032] Performing a mean removal process on each column of the multi-scale feature matrix according to the mean value to obtain a mean removal feature matrix;

[0033] Calculating the column variance of the de-meaned feature matrix to obtain the variance of each column of feature data of the de-meaned feature matrix;

[0034] Standardizing the de-meaned feature matrix according to the variance to obtain a standardized feature data matrix with unit variance;

[0035] The standardized characteristic data matrix is ​​normalized, and outlier detection is performed on the normalized standardized characteristic data matrix to obtain standardized acceleration characteristic data.

[0036] Furthermore, the step of calculating the covariance matrix of the normalized acceleration feature data, sorting the calculated multiple eigenvalues, selecting the eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting results, and establishing the dimension reduction feature matrix according to the eigenvectors specifically includes:

[0037] According to the normalized acceleration feature data, the covariance between the dimensions of the acceleration feature data is calculated to obtain a covariance matrix;

[0038] Performing eigendecomposition on the covariance matrix to obtain eigenvalues ​​of the covariance matrix and eigenvectors corresponding to the eigenvalues;

[0039] Sorting the eigenvalues ​​in descending order, and selecting eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting result;

[0040] The characteristic vectors of the pre-set values ​​are organized into a dimension reduction matrix by columns, and an acceleration feature data matrix is ​​established for the acceleration feature data. Matrix multiplication is performed between the dimension reduction matrix and the acceleration feature data matrix to obtain the dimension reduction feature matrix.

[0041] Furthermore, the step of performing classification prediction processing on the reduced dimension feature matrix to obtain a classification result and a confidence score, and generating a test report in combination with the classification result and the confidence score specifically includes:

[0042] Normalizing each row of the reduced-dimensional feature matrix, calculating the eigenvalue range of the reduced-dimensional feature matrix by column and performing a preset scaling to obtain a normalized reduced-dimensional feature matrix;

[0043] Passing the normalized reduced-dimensional feature matrix as input data to a support vector machine classifier, performing nonlinear mapping processing on the input data according to a kernel function of the support vector machine classifier, generating a classification decision boundary, and outputting a preliminary classification result;

[0044] According to the decision function distance formula of the support vector machine classifier, the distance value from the input data to the classification decision boundary is calculated, and a confidence distribution is generated for the input data to obtain a confidence score matrix corresponding to the input data;

[0045] For input data whose confidence in the preliminary classification results is lower than a preset threshold, a random forest classifier is used for re-testing, and a revised classification result is generated by a majority voting method;

[0046] Divide the sample data according to the modified classification result and the confidence score matrix corresponding to the modified classification result, calculate the confidence mean, variance and coverage of the sample data of each category, obtain the comprehensive confidence eigenvalue of each category, and perform correlation analysis on the confidence eigenvalue of each category with the dimension reduction feature matrix to generate a comprehensive feature analysis table including category distribution and eigenvector relationship;

[0047] According to the comprehensive feature analysis table, the quantity distribution, confidence range and feature vector direction of the classified samples are statistically analyzed by category to generate a multidimensional performance evaluation index of the vibration characteristics, and the multidimensional performance evaluation index is compared with the preset detection standards item by item to determine whether the vibration performance of the force-sensing touch panel meets the requirements, and a detection report is generated, wherein the detection report includes the classified sample distribution statistics, the confidence analysis chart, the abnormal point feature list and the final pass / fail result.

[0048] The present invention also provides a vibration detection device for a force-sensing touch pad of a notebook computer, comprising:

[0049] An acquisition module, used for acquiring the original signal of the force-sensing touch panel, filtering the original signal to obtain the acceleration signal of the force-sensing touch panel, wherein the acceleration signal includes acceleration data of three axes: X, Y, and Z;

[0050] An analysis module, used to perform time domain analysis on the acceleration signal and convert the acceleration signal into a frequency domain signal to obtain the amplitude and phase information of each frequency component;

[0051] A decomposition module performs multi-level decomposition on the amplitude and phase information of each frequency component, and extracts multi-scale time-frequency features in the acceleration signal to obtain a multi-scale feature matrix of the acceleration signal;

[0052] A conversion module, used for converting all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data;

[0053] A calculation module, used to calculate the covariance matrix of the normalized acceleration feature data, sort the calculated multiple eigenvalues, select the eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting results, and establish a dimension reduction feature matrix according to the eigenvectors;

[0054] The generation module is used to perform classification prediction processing on the reduced-dimensional feature matrix to obtain a classification result and a confidence score, and generate a detection report in combination with the classification result and the confidence score.

[0055] The present invention further provides a computer device, which includes a processor, a memory, and a computer program stored in the memory and running on the processor, and the processor implements the above method when executing the computer program.

[0056] 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 above method is implemented.

[0057] Beneficial effects:

[0058] In the present invention, the X, Y, and Z three-axis acceleration data of the force touch panel are collected based on an acceleration sensor, and the original signal is filtered, analyzed in the time domain and frequency domain, so as to accurately capture the vibration characteristics of the force touch panel in different areas. Through multi-scale time-frequency feature extraction and multi-level decomposition, the amplitude and phase characteristics of the vibration signal of the force touch panel can be accurately analyzed, so as to fully reflect the distribution of vibration intensity in different areas. In particular, through standardization and dimensionality reduction processing, the detection ability of vibration intensity deviation from the design value is effectively enhanced, and the deficiency of the existing technology that it is difficult to quantify the vibration intensity is solved, so as to ensure the uniformity of the vibration intensity and the consistency of the response of the force touch panel, and provide users with accurate and reliable operation feedback experience. In addition, through the refined processing of multi-dimensional vibration feature data, the detection accuracy and algorithm efficiency are significantly improved. At the same time, by combining the test report generated by classification prediction and confidence scoring, the vibration performance of the force touch panel can be intuitively and quickly evaluated. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of the steps of a vibration detection method of a force-sensing touch pad of a notebook computer according to the present invention;

[0060] Figure 2 It is a structural schematic block diagram of a vibration detection device for a force-sensing touch pad of a notebook computer according to the present invention;

[0061] Figure 3 is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention;

[0062] 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

[0063] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0064] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0065] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0066] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may include that the first and second features are in direct contact, or may include that the first and second features are not in direct contact but are in contact through another feature between them. Moreover, a first feature being "above", "above" and "above" a second feature includes that the first feature is directly above and obliquely above the second feature, or simply indicates that the first feature is higher in level than the second feature. A first feature being "below", "below" and "below" a second feature includes that the first feature is directly below and obliquely below the second feature, or simply indicates that the first feature is lower in level than the second feature.

[0067] Embodiment 1

[0068] Reference Figure 1 The present invention proposes a vibration detection method for a force-sensing touch pad of a notebook computer, comprising the following steps:

[0069] S1: collecting the original signal of the force-sensing touch panel based on the acceleration sensor, and filtering the original signal to obtain the acceleration signal of the force-sensing touch panel, wherein the acceleration signal includes acceleration data of three axes: X, Y, and Z;

[0070] In step S1, the force-sensing touch pad of the laptop is contacted with the acceleration sensor, and the acceleration sensor selects the X, Y, and Z three-axis accelerometer to collect the vibration signal of the force-sensing touch pad. The acceleration sensor can capture the original acceleration data of the force-sensing touch pad in the three directions of X, Y, and Z. The original signal contains various interferences from inside and outside the device, so the collected signal needs to be filtered to remove the noise. Low-pass and high-pass filters can be used to filter out unnecessary frequency components, remove high-frequency noise such as electromagnetic interference, and also eliminate low-frequency vibration interference generated by the device under test itself. After filtering, there may still be some smaller noise components in the signal, so the signal can be further denoised using techniques such as wavelet denoising. At this time, the smoothness and stability of the signal are improved, which can more accurately reflect the actual acceleration characteristics of the force-sensing touch pad. After filtering and denoising, the signal will eventually obtain an acceleration signal containing X, Y, and Z three-axis acceleration data, which will serve as the basis for subsequent analysis.

[0071] S2: performing time domain analysis processing on the acceleration signal, and converting the acceleration signal into a frequency domain signal to obtain amplitude and phase information of each frequency component;

[0072] In step S2, the acceleration signal after filtering and denoising is first analyzed in the time domain. The goal of this stage is to understand the changing trend of the signal from the time domain perspective. The basic mode of vibration of the force touch pad can be observed by calculating the statistical characteristics of the signal such as the average value and peak value. Time domain analysis helps to preliminarily determine whether there is an abnormality in the force touch pad. After completing the time domain analysis, the signal can be converted to the frequency domain through the fast Fourier transform (FFT). The time domain signal is decomposed into multiple different frequency components. Each frequency component has a specific amplitude and phase. The amplitude reflects the energy of the frequency in the signal, and the phase describes the time offset of the frequency component relative to the signal. The frequency domain signal can assist in identifying which frequency ranges of vibration components are signs of normal operation of the force touch pad to be tested, and which frequency bands may represent abnormal vibration characteristics. For example, high-frequency components may correspond to failures or abnormal conditions of the force touch pad of a laptop computer.

[0073] S3: performing multi-level decomposition on the amplitude and phase information of each frequency component, and extracting multi-scale time-frequency features in the acceleration signal to obtain a multi-scale feature matrix of the acceleration signal;

[0074] In step S3, the amplitude and phase information of each frequency component obtained in the frequency domain are further analyzed at multiple levels. In order to reveal the changing characteristics of the signal at different time scales, time-frequency analysis methods such as wavelet transform can be used to decompose the signal at multiple levels. By decomposing the signal at different scales, the instantaneous frequency changes in the acceleration signal can be captured. At this stage, the time-frequency diagram of the signal is generated, which shows the amplitude changes of each frequency component in different time windows. Through this multi-scale decomposition, the changing rules of the signal in different frequency bands can be extracted, revealing the vibration characteristics of the force touch pad of the laptop computer in each frequency band. The obtained time-frequency features can help identify the differences between the force touch pad in normal working state and abnormal state. Finally, all the extracted multi-scale features are integrated into a very multi-scale feature matrix, which contains detailed information of the signal at different time scales and frequency bands, laying the foundation for subsequent feature processing and machine learning analysis.

[0075] S4: converting all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data;

[0076] In step S4, the multi-scale feature matrix obtained in step S3 is first standardized. The standardization process includes zero-meaning the data of each feature, that is, subtracting the mean of each feature so that the central value of each feature is zero. The purpose of this step is to eliminate possible offsets in the data so that all feature data are compared under the same benchmark. Next, the data of each feature is divided by the standard deviation of the feature to complete the unit variance standardization. In this process, the data of all features are converted into the form of unit variance, so that their fluctuation amplitudes become consistent, avoiding some features with larger magnitudes from having too much impact on subsequent analysis. The standardized acceleration feature data will be more suitable for subsequent statistical analysis and machine learning processing, ensuring that the weights between different features are balanced, and avoiding some features from having deviations in the results due to different dimensions.

[0077] S5: Calculate the covariance matrix of the normalized acceleration feature data, sort the calculated multiple eigenvalues, select the eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting results, and establish a dimension reduction feature matrix according to the eigenvectors;

[0078] In step S5, the standardized acceleration feature data is used to calculate its covariance matrix. The covariance matrix reflects the correlation between the various features and can assist in understanding the mutual influence between different features. After the covariance matrix is ​​calculated, the eigenvalue decomposition technique can be used to extract eigenvalues ​​and eigenvectors from the covariance matrix. The eigenvalue represents the variance of the principal component represented by each eigenvector, while the eigenvector is an indication of the important direction in the data. Sorting is performed according to the size of the eigenvalue, and the eigenvalues ​​and their corresponding eigenvectors of a preset number before sorting are selected. These eigenvectors represent the most important components in the data. By selecting these eigenvectors, a dimensionality reduction feature matrix can be established, which contains the most representative and informative features in the data. Dimensionality reduction can not only reduce computational complexity, but also help remove redundant information, making subsequent analysis more efficient.

[0079] S6: Perform classification prediction processing on the reduced-dimensional feature matrix to obtain a classification result and a confidence score, and generate a detection report in combination with the classification result and the confidence score.

[0080] In step S6, a classification algorithm is used to analyze and predict the feature matrix after dimensionality reduction. Machine learning algorithms including support vector machine (SVM), decision tree, random forest, etc. can be used to train based on the feature data after dimensionality reduction to learn the relationship between the acceleration signal and different vibration states (such as normal or abnormal). The classification model obtained through training can classify and predict the new acceleration signal and give a prediction result of whether the signal is in a normal working state or has an abnormality. At the same time, in order to evaluate the reliability of the classification, the classification model will also output a confidence score to indicate the credibility of the classification result. A result with a high confidence score indicates that the model is more confident in the judgment of the sample, while a result with a low confidence score requires further verification. Finally, based on the classification results and confidence score, the system will generate a test report, which lists in detail the current vibration state of the laptop force touchpad, the abnormal detection results, and the possible causes of the failure.

[0081] In summary, by collecting the X, Y, and Z three-axis acceleration data of the force touch panel based on the acceleration sensor, filtering the original signal, and analyzing the time domain and frequency domain, the vibration characteristics of the force touch panel in different areas are accurately captured. Through multi-scale time-frequency feature extraction and multi-level decomposition, the amplitude and phase characteristics of the vibration signal of the force touch panel can be accurately analyzed, so as to fully reflect the distribution of vibration intensity in different areas. In particular, through standardization and dimensionality reduction processing, the detection ability of vibration intensity deviation from the design value is effectively enhanced, and the deficiency of the existing technology that it is difficult to quantify the vibration intensity is solved, ensuring the uniformity of the vibration intensity and the consistency of the response of the force touch panel, providing users with accurate and reliable operation feedback experience. In addition, through the refined processing of multi-dimensional vibration feature data, the detection accuracy and algorithm efficiency are significantly improved. At the same time, by combining the test report generated by classification prediction and confidence scoring, the vibration performance of the force touch panel can be evaluated intuitively and quickly.

[0082] In an application scenario embodiment, the force-sensing touchpad, as a new touchpad technology for laptop computers, brings users a more convenient and efficient operating experience. It needs to undergo strict testing before leaving the factory to ensure product quality.

[0083] The acceleration acquisition module, motion control system and acceleration detection module are integrated into the detection assembly. The acceleration acquisition module includes an acceleration sensor module for acceleration measurement and a data acquisition card module for acceleration acquisition; the motion control system and acceleration detection module include motion control software integrated with the triggering and acceleration value monitoring of the data acquisition card, and the software establishes communication with the force-sensing touch panel of the laptop computer to be tested through the serial port, and triggers the vibration of the force-sensing touch panel of the laptop computer to be tested through the command trigger mode.

[0084] By connecting the acceleration sensor to a data acquisition card, triggering the data acquisition card through software and monitoring the acceleration value, the real-time acceleration of the object under test can be visualized, and the different positions of the object under test can be moved and measured through the software motion control system.

[0085] During the test of the force-sensing touch panel, the accelerometer, voice coil motor, pressure sensor and sensor overpressure alarm sensor can be used as test components to press the force-sensing touch panel. In order to maintain a constant pressure range between the accelerometer and the object to be tested when measuring acceleration and to avoid damage to the product due to excessive pressure, the voice coil motor and pressure sensor are used to quickly and accurately position the accelerometer within the specified pressure range to measure the acceleration of the object to be tested.

[0086] Preferably, the acceleration sensor uses an IEPE type three-axis acceleration sensor with a measurement range of 10g and a resolution of 0.1mg, the Z axis uses a voice coil motor with a positioning accuracy of ±5um, and the pressure sensor uses an electronic weighing instrument with an input sensitivity of 0.4mv / V~6mv / V. The data acquisition card uses a high-speed, high-precision 4-channel synchronous acquisition card with up to 24-bit accuracy and 156KSPS sampling rate, and the X, Y, and Z direction signal lines of the three-axis acceleration sensor are respectively connected to each channel access port of the data acquisition card.

[0087] In practical applications, this detection method can also effectively identify the vibration characteristics of the laptop force touch pad under different operating conditions. For example, when the user performs operations such as light touch and sliding, the vibration signal collected by the acceleration sensor can be used to analyze the impact of the operation force and speed on the vibration of the force touch pad. In addition, this method can also be used to detect the wear or aging of the force touch pad after long-term use. By comparing the vibration characteristics of the new and old force touch pads, their durability and reliability can be evaluated.

[0088] During the detection process, the software system will record and analyze vibration data in real time. Once an abnormal vibration pattern is detected, the system will immediately issue an alarm and provide corresponding fault diagnosis information. This not only helps to find and fix potential problems in a timely manner, but also provides a scientific basis for the repair and maintenance of laptops, comprehensively evaluates the performance and stability of force-sensing touchpads, and provides strong technical support for the quality control and user experience of laptops.

[0089] In one embodiment, the step of collecting the original signal of the force-sensitive touch panel based on the acceleration sensor and filtering the original signal to obtain the acceleration signal of the force-sensitive touch panel specifically includes:

[0090] Placing the acceleration sensor in contact with the force-sensing touch panel and initializing the acceleration sensor;

[0091] Acquire a timestamp of the vibration of the force-sensing touch panel, collect an original signal of the force-sensing touch panel based on the acceleration sensor, perform denoising on the original signal, filter out low-frequency noise and high-frequency noise of the original signal, and obtain a denoised acceleration signal;

[0092] Performing dynamic range adjustment on the denoised acceleration signal to eliminate amplitude deviation in the denoised acceleration signal and obtain an acceleration signal of uniform magnitude;

[0093] The uniform magnitude acceleration signal is time-aligned with the timestamp to obtain the time-aligned acceleration signal.

[0094] In the above embodiment, in order to ensure that the acceleration sensor can accurately capture the vibration signal generated by the force-sensing touch panel. The acceleration sensor has multiple axial vibration detection functions. By contacting the surface of the force-sensing touch panel or its related parts, the sensor can sense the directional vibrations borne by the force-sensing touch panel, thereby generating an original acceleration signal. The purpose of the sensor initialization process is to calibrate the zero point and deviation of the sensor to ensure its accuracy and stability.

[0095] After completing the initialization process, obtain the timestamp of the force touch panel vibration, and collect the original signal based on the acceleration sensor, that is, collect the acceleration signal generated by the force touch panel at each moment in real time through the acceleration sensor, and record the specific time when these signals occur. Timestamp recording is crucial for subsequent signal processing, because in dynamic signal processing, time alignment and synchronization are important steps to ensure data correctness and accuracy. The acceleration sensor collects the original signal together with the timestamp for subsequent analysis and processing.

[0096] During the signal acquisition process, environmental noise, noise from the device under test, or interference in signal transmission may affect the quality of the original signal, so denoising is necessary. Low-frequency noise in the original signal is usually caused by vibration in the external environment or non-ideal operation of the force-sensing touchpad of the laptop under test, while high-frequency noise may be caused by interference from electronic equipment, power fluctuations, and other factors. In order to effectively remove these noises, a variety of filtering techniques can be used, such as a bandpass filter that combines a low-pass filter and a high-pass filter, or advanced denoising algorithms based on wavelet transforms. The removal of low-frequency noise is mainly achieved by setting a suitable filter cutoff frequency to filter out signal components below a certain frequency, while the removal of high-frequency noise is achieved by filtering out signal components above a certain frequency in the same way. Through this denoising process, a purer acceleration signal can be obtained.

[0097] Since the accelerometer may produce different signal amplitudes due to different measurement scenarios, environmental changes or physical property differences in actual applications, it is necessary to adjust the dynamic range of the signal to eliminate the amplitude deviation. The purpose of dynamic range adjustment is to make the signals from different sources within a unified standard range to ensure that there will be no analysis errors in subsequent processing due to excessive differences in signal amplitude.

[0098] In the process of multi-channel data acquisition, the role of time alignment is to ensure that all signal data can be accurately matched on the same time axis to avoid data offset caused by different sampling intervals or sampling timing differences. The implementation method of time alignment is to accurately match the collected acceleration signal and the timestamp to ensure that the time point corresponding to each signal is accurate.

[0099] In one example, the step of performing time domain analysis on the acceleration signal and converting the acceleration signal into a frequency domain signal to obtain amplitude and phase information of each frequency component specifically includes:

[0100] Performing frame processing on the acceleration signal according to a preset time window length and a preset overlap rate to obtain a plurality of time window signals;

[0101] Performing time domain analysis and processing on each of the time window signals to obtain vibration characteristic data within each time window;

[0102] Performing fast Fourier transform on the vibration characteristic data in each time window to obtain frequency components of each frequency band;

[0103] A modular operation is performed on the complex value of each frequency component to obtain the amplitude of each frequency component, and a complex angle of each frequency component is calculated to obtain the phase information of each frequency component.

[0104] In the above embodiment, the time domain analysis and frequency domain conversion process of the acceleration signal is actually a multi-stage process of signal processing, in which each step has its specific function, the purpose is to extract the frequency component of the acceleration signal, and provide valuable physical characteristics for subsequent analysis through amplitude and phase information. It is necessary to perform frame processing on the acceleration signal, which is a very critical step in signal analysis. The acceleration signal is a continuous time domain signal, and in order to effectively perform frequency domain analysis, it must be divided into several small segments, each of which is local in time and can better capture the changing characteristics of the acceleration signal in the time dimension. Specifically, the original acceleration signal is framed according to the preset time window length and overlap rate. The time window length refers to the time span of each frame signal, and the overlap rate determines the overlapping part between adjacent frame signals. Reasonable time window length can balance the resolution of time domain and frequency domain analysis. A short time window provides a higher time resolution, but may cause the frequency domain information to be inaccurate; while a longer time window is conducive to obtaining more accurate frequency domain information, but it will reduce the time domain resolution. The setting of overlap rate also plays a key role. A higher overlap rate helps improve the continuity of analysis and reduce distortion caused by frame breaks. In practical applications, assuming that the sampling time of an acceleration signal is 10 seconds, the time window length is set to 2 seconds, and the overlap rate is 50%, then each time window covers 2 seconds of the signal, and there will be a 50% overlap between two adjacent time windows, which means that the starting point of each frame differs by 1 second.

[0105] For each frame of signal, time domain analysis processing is required. The main purpose of time domain analysis is to extract vibration characteristic data from the acceleration signal in each time window. These data can be the mean, variance, peak value, etc. of the signal, which are directly related to the vibration characteristics of the signal.

[0106] After the time domain feature data of each frame is processed, the next step is to perform a fast Fourier transform (FFT) on the vibration feature data in each time window to convert the time domain signal into a frequency domain signal. Through FFT, each time window signal can be decomposed into a series of sine waves of different frequencies, and then the frequency components of each frequency band can be obtained.

[0107] For the complex value of each frequency component obtained, a modular operation and an angle calculation need to be performed to obtain the amplitude and phase information of the frequency component respectively.

[0108] In one example, the step of performing multi-level decomposition on the amplitude and phase information of each of the frequency components and extracting the multi-scale time-frequency features in the acceleration signal to obtain the multi-scale feature matrix of the acceleration signal specifically includes:

[0109] Performing wavelet transform processing on the amplitude and phase information of each frequency component to obtain multi-level wavelet coefficients of the frequency domain signal;

[0110] Performing inverse transformation on the wavelet coefficients to reconstruct the time-frequency diagram of the acceleration signal at each scale;

[0111] Refining the time-frequency graph by thresholding and peak detection to extract characteristic components at multiple scales;

[0112] The characteristic components at each scale are standardized to obtain time-frequency characteristic data;

[0113] The time-frequency feature data is divided into multiple time windows according to time and frequency, and the feature data in each time window is organized into a multi-scale feature matrix according to scale.

[0114] In the above embodiment, the amplitude and phase information of each frequency component are processed by wavelet transform, and the signal can be subjected to local frequency analysis by selecting a suitable wavelet basis function (e.g., Morlet wavelet or Daubechies wavelet). It is possible to perform refined analysis on both time and frequency scales to help obtain the time distribution of different frequency components. When performing wavelet transform, the amplitude and phase information of each frequency component will be extracted, and the signal will be decomposed by wavelet coefficients of different scales. At this point, not only the spectrum of the signal can be obtained, but also its time local information can be retained.

[0115] By inverse transforming the obtained wavelet coefficients, the time-frequency diagram of the signal can be reconstructed. The time-frequency diagram can be regarded as the joint performance of the signal on the time and frequency axes, which can clearly show the energy distribution of specific frequency components of the signal in different time periods. Through inverse wavelet transform, the decomposed wavelet coefficients are restored to the time-frequency diagram, thereby reconstructing the specific performance of the acceleration signal at each scale, providing detailed frequency information of the signal in different time periods.

[0116] Among them, by inverse transforming the wavelet coefficients, the calculation expression of the time-frequency diagram of the acceleration signal at each scale is reconstructed as follows: , in this expression, is the reconstructed time-frequency diagram, Indicates scale, is the index of the time / space position, It is a scale The coefficients obtained by wavelet transform represent the characteristics of the signal at this scale. is the wavelet reconstruction function, is a constant, which is related to the choice of mother wavelet and is used to ensure the correctness of the inverse transformation process.

[0117] Furthermore, it also includes the expression of the wavelet reconstruction function: , is the mother wavelet function, is the scale factor, is the translation factor.

[0118] The wavelet coefficients The reconstruction function corresponding to the scale and position Multiplying and summing the results of all scales and positions can restore the time-frequency characteristics of the signal. and time location The inverse transform process is to combine the wavelet components of different scales and positions to reconstruct the complete signal. The wavelet coefficients are inversely transformed to obtain the performance of the signal in the time-frequency domain. The time-frequency diagram can provide the distribution information of the signal at different times and frequencies.

[0119] By refining the time-frequency graph through thresholding and peak detection, the key features of the acceleration signal can be further extracted. In practical applications, the time-frequency graph may contain a lot of noise or irrelevant frequency components, so it is necessary to process the image by setting a suitable threshold to remove background noise and retain only important signal features. Thresholding means setting a threshold so that only frequency components greater than the value will be retained, while smaller components will be ignored. Peak detection determines the key features of the signal by identifying local peaks in the time-frequency graph. The significant part of the signal will appear as peaks or local high-frequency areas in the time-frequency graph, so the frequency components corresponding to these peaks often represent important features of the signal.

[0120] The characteristic components at each scale are standardized to eliminate the impact of amplitude differences between different scales. The mean of each characteristic component is subtracted and divided by its standard deviation, so that the characteristic data at all scales have the same dimension and a balanced numerical range.

[0121] The time-frequency feature data is divided into multiple time windows according to time and frequency, and the feature data in each time window is organized into a multi-scale feature matrix according to scale. Specifically, the division of time windows is based on the sampling frequency of the signal and the analysis requirements. The selection of time windows needs to take into account the time-varying characteristics of the signal and the computational complexity. By allocating time-frequency feature data to different time windows, the changing characteristics of the signal in different time periods can be effectively captured, while avoiding information loss caused by analysis of too long time periods. In each time window, the feature data will be organized according to different scales to form a multi-scale feature matrix. This matrix not only contains the time-frequency characteristics of the acceleration signal, but also retains the multi-dimensional information of the feature components at each scale, thereby providing rich data support for subsequent signal classification, pattern recognition or other analysis tasks.

[0122] In one example, the step of converting all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data specifically includes:

[0123] Performing column mean calculation on all feature data of the multi-scale feature matrix to obtain the mean of each column of feature data;

[0124] Performing a mean removal process on each column of the multi-scale feature matrix according to the mean value to obtain a mean removal feature matrix;

[0125] Calculating the column variance of the de-meaned feature matrix to obtain the variance of each column of feature data of the de-meaned feature matrix;

[0126] Standardizing the de-meaned feature matrix according to the variance to obtain a standardized feature data matrix with unit variance;

[0127] The standardized characteristic data matrix is ​​normalized, and outlier detection is performed on the normalized standardized characteristic data matrix to obtain standardized acceleration characteristic data.

[0128] In the above embodiment, the column mean is calculated for all feature data of the multi-scale feature matrix to eliminate the influence of different dimensions or scales in the data. For each column of feature data in the multi-scale feature matrix, its mean is calculated, and the mean calculation is obtained by summing the values ​​of all data points in the column and dividing it by the number of elements in the column. The average level of each column of data can be understood, so as to identify the magnitude difference between different feature data. According to the calculated mean of each column, each column of data in the multi-scale feature matrix is ​​de-meaned, that is, the mean of each column of data is subtracted from each data point in the column, so that the mean of each column of data becomes zero. The data after de-meaning is called "de-meaning feature matrix", and each column of data fluctuates around zero. This process helps to eliminate the offset in the data, so that subsequent processing can pay more attention to the fluctuation characteristics of the data rather than its absolute value.

[0129] The column variance of the removed mean feature matrix is ​​calculated. The variance reflects the degree of data dispersion. The larger the variance, the stronger the data volatility. The variance of each column of the removed mean data is calculated. The square of the difference between each point in the column and the mean of the column is summed, and then divided by the number of elements in the column to obtain the variance of each column of data. This can measure the degree of fluctuation of each feature data in its column, so as to make appropriate scaling adjustments in the standardization process to ensure that the fluctuation amplitude of each feature is at the same order of magnitude.

[0130] Based on the variance of each column, the de-meaned feature matrix is ​​standardized so that each column of data has unit variance. Each de-meaned data point is divided by the standard deviation of the column (the standard deviation is the square root of the variance) to obtain a standardized feature data matrix with unit variance. After this process, each column of data in the de-meaned feature matrix will have the same scale, and the degree of fluctuation of each column is 1. This step is to eliminate the dimensional differences between different features due to different variances, so that each feature has the same influence on subsequent analysis or model training. Standardized data will help improve the performance of subsequent algorithms, especially in machine learning models involving multiple features. Standardization can prevent certain features from affecting the stability of model training due to excessively large or small values.

[0131] Normalization is to scale each column of data to a predetermined range, which can be [0, 1] or [-1, 1]. The purpose of normalization is to prevent data from having an adverse effect on model training in extreme cases. There are many normalization methods, which can be used to scale the data according to the maximum and minimum values ​​of the column so that all data points are mapped to the specified range.

[0132] Outlier detection is performed on the normalized feature data matrix. The outlier detection methods used may include statistical methods, such as the Z-score detection method, the box plot method (IQR), or model-based methods, such as the isolation forest, etc., which can help identify points that deviate from the normal data distribution and process them according to the actual situation, and finally obtain the standardized acceleration feature data.

[0133] In one embodiment, the steps of calculating the covariance matrix of the standardized acceleration feature data, sorting the calculated multiple eigenvalues, selecting the eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting results, and establishing the dimension reduction feature matrix according to the eigenvectors specifically include:

[0134] According to the normalized acceleration feature data, the covariance between the dimensions of the acceleration feature data is calculated to obtain a covariance matrix;

[0135] Performing eigendecomposition on the covariance matrix to obtain eigenvalues ​​of the covariance matrix and eigenvectors corresponding to the eigenvalues;

[0136] Sorting the eigenvalues ​​in descending order, and selecting eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting result;

[0137] The characteristic vectors of the pre-set values ​​are organized into a dimension reduction matrix by columns, and an acceleration feature data matrix is ​​established for the acceleration feature data. Matrix multiplication is performed between the dimension reduction matrix and the acceleration feature data matrix to obtain the dimension reduction feature matrix.

[0138] Among them, the calculation expression of the covariance matrix is: , the normalized acceleration feature data matrix is ​​X, and its size is , is the number of samples, is the feature dimension, is the covariance matrix, is the transpose of the data matrix.

[0139] The characteristic decomposition expression of the covariance matrix is: , is the covariance matrix The characteristic vector of is the eigenvector The corresponding eigenvalues ​​are decomposed through eigenvalue decomposition to obtain the eigenvalues ​​and eigenvectors of the covariance matrix.

[0140] Select the first The eigenvalues ​​corresponding to the eigenvectors , these eigenvectors form the dimension reduction matrix , whose column vector is this eigenvectors; reduced-dimensionality matrix for: , The size is , is the dimension of the original feature, is the dimension after dimensionality reduction.

[0141] The reduced dimension matrix is ​​obtained , which can be multiplied with the standardized acceleration feature data matrix X to obtain the reduced dimension feature matrix , is the dimension-reduced feature matrix, whose size is , is the number of samples, is the number of features after dimensionality reduction.

[0142] In the above embodiment, the covariance between the dimensions of the acceleration feature data is calculated to obtain a covariance matrix, which can describe the relationship between the features of each dimension, and each element of the matrix reflects the linear influence of a specific feature on another feature. The constructed covariance matrix is ​​subjected to eigendecomposition, and this process can identify the most representative feature direction in the data. The result of the eigendecomposition is to obtain a set of eigenvalues ​​and eigenvectors, wherein the eigenvalue reflects the data variance in the direction of the corresponding eigenvector, and the larger the eigenvalue, the more information in the feature direction.

[0143] These eigenvalues ​​are sorted in descending order, and the eigenvectors corresponding to the first several eigenvalues ​​after sorting are selected. This number of choices can be preset according to actual needs to ensure that the selected eigenvectors can retain the intrinsic information of the data to the greatest extent without introducing too much redundancy. After selecting these eigenvectors, they are combined into a new dimensionality reduction matrix by column. This dimensionality reduction matrix will serve as the basis for subsequent operations to help convert the original high-dimensional acceleration feature data into a low-dimensional feature space.

[0144] In order to achieve the ultimate goal of dimensionality reduction, matrix multiplication is performed between the dimensionality reduction matrix and the standardized acceleration feature data matrix to obtain a dimensionality reduction feature matrix. The dimension of this dimensionality reduction feature matrix will be reduced, and this new space can capture the most important feature information in the original data. Throughout the process, through effective mathematical operations and data processing, dimensionality reduction processing of high-dimensional acceleration feature data can be achieved, reducing the complexity of the data while improving the efficiency and accuracy of subsequent analysis and modeling. This dimensionality reduction method not only helps to improve the efficiency of the data processing process, but also reduces the risk of overfitting in fields such as machine learning and signal analysis, and improves the generalization ability of the model.

[0145] In one embodiment, the step of performing classification prediction processing on the reduced dimension feature matrix to obtain a classification result and a confidence score, and generating a test report in combination with the classification result and the confidence score specifically includes:

[0146] Normalizing each row of the reduced-dimensional feature matrix, calculating the eigenvalue range of the reduced-dimensional feature matrix by column and performing a preset scaling to obtain a normalized reduced-dimensional feature matrix;

[0147] Passing the normalized reduced-dimensional feature matrix as input data to a support vector machine classifier, performing nonlinear mapping processing on the input data according to a kernel function of the support vector machine classifier, generating a classification decision boundary, and outputting a preliminary classification result;

[0148] According to the decision function distance formula of the support vector machine classifier, the distance value from the input data to the classification decision boundary is calculated, and a confidence distribution is generated for the input data to obtain a confidence score matrix corresponding to the input data;

[0149] For input data whose confidence in the preliminary classification results is lower than a preset threshold, a random forest classifier is used for re-testing, and a revised classification result is generated by a majority voting method;

[0150] Divide the sample data according to the modified classification result and the confidence score matrix corresponding to the modified classification result, calculate the confidence mean, variance and coverage of the sample data of each category, obtain the comprehensive confidence eigenvalue of each category, and perform correlation analysis on the confidence eigenvalue of each category with the dimension reduction feature matrix to generate a comprehensive feature analysis table including category distribution and eigenvector relationship;

[0151] According to the comprehensive feature analysis table, the quantity distribution, confidence range and feature vector direction of the classified samples are statistically analyzed by category to generate a multidimensional performance evaluation index of the vibration characteristics, and the multidimensional performance evaluation index is compared with the preset detection standards item by item to determine whether the vibration performance of the force-sensing touch panel meets the requirements, and a detection report is generated, wherein the detection report includes the classified sample distribution statistics, the confidence analysis chart, the abnormal point feature list and the final pass / fail result.

[0152] In the above embodiment, each row of the reduced dimension feature matrix is ​​normalized so that its value is within a uniform standard range, thereby ensuring that the contribution of each feature is relatively balanced. The eigenvalue range of the reduced dimension feature matrix is ​​calculated column by column and scaled by a preset ratio to obtain a normalized reduced dimension feature matrix.

[0153] The normalized reduced-dimensional feature matrix is ​​passed as input data to the support vector machine classifier. The support vector machine classifier uses its kernel function to perform nonlinear mapping on the input data. The purpose is to map complex data that cannot be divided by a straight line to a high-dimensional space, so that the data becomes linearly separable in the high-dimensional space. Through this mapping, the support vector machine can find a classification decision boundary to divide the data into different categories. After classification, the support vector machine outputs a preliminary classification result, but this is only a preliminary judgment of the classification and cannot fully reflect the confidence of the input data. In order to evaluate the confidence of the preliminary classification result, it is necessary to calculate the distance value from each data point to the decision boundary. According to the decision function distance formula of the support vector machine, points closer to the decision boundary often mean that the model is not sure enough about the classification of the point, so the confidence is low. On the contrary, points farther from the decision boundary indicate that the reliability of the classification result is higher. Therefore, using this distance value, the support vector machine can generate a confidence distribution for each data point, and finally obtain a corresponding confidence score matrix, which reflects the classification confidence of each input data.

[0154] However, since the support vector machine may have misclassification or low confidence in some cases of fuzzy boundaries, the system will further re-test samples with confidence levels below the preset threshold. The re-test method is to use a random forest classifier to improve the accuracy of classification through the integration of multiple decision trees. By re-testing low-confidence samples, the random forest can re-judge the category of these data through the majority voting method, thereby correcting the classification results of the support vector machine and obtaining a corrected classification result.

[0155] For the corrected classification results, it is necessary to calculate the confidence mean, variance, and coverage of each category of samples. These statistical values ​​can reflect the stability and distribution of each category and help understand the confidence characteristics of each category. In addition, by performing correlation analysis on these confidence eigenvalues ​​and the dimensionality reduction feature matrix, a comprehensive feature analysis table can be generated, which contains the distribution of each category, confidence characteristics, and corresponding eigenvector relationships. This analysis result provides the necessary basis for subsequent multi-dimensional performance evaluation.

[0156] Based on the comprehensive feature analysis table, the multi-dimensional performance evaluation index of vibration characteristics will be generated by statistically analyzing the number of samples, confidence range and feature vector direction by category, which is crucial for determining whether the vibration performance of the laptop force touch pad meets the requirements. By comparing each item with the preset test standards, the system can determine whether the vibration performance of the laptop force touch pad is qualified. Finally, based on these evaluation results, a test report is generated. The report will include the distribution statistics of the classified samples, the confidence analysis chart, the list of abnormal point features and the final pass / fail results, thereby providing users with detailed and reliable test results.

[0157] Embodiment 2

[0158] Reference Figure 2 , a vibration detection device for a force-sensing touch pad of a notebook computer, comprising:

[0159] The acquisition module 100 is used to acquire the original signal of the force-sensing touch panel, and filter the original signal to obtain the acceleration signal of the force-sensing touch panel, wherein the acceleration signal includes acceleration data of three axes: X, Y, and Z;

[0160] The analysis module 200 is used to perform time domain analysis on the acceleration signal and convert the acceleration signal into a frequency domain signal to obtain the amplitude and phase information of each frequency component;

[0161] A decomposition module 300 performs multi-level decomposition on the amplitude and phase information of each frequency component, and extracts multi-scale time-frequency features in the acceleration signal to obtain a multi-scale feature matrix of the acceleration signal;

[0162] A conversion module 400 is used to convert all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data;

[0163] The calculation module 500 is used to calculate the covariance matrix of the normalized acceleration feature data, sort the calculated multiple eigenvalues, select the eigenvectors corresponding to the eigenvalues ​​with the preset values ​​before the sorting results, and establish the dimension reduction feature matrix according to the eigenvectors;

[0164] The generation module 600 is used to perform classification prediction processing on the reduced-dimensional feature matrix to obtain a classification result and a confidence score, and generate a detection report in combination with the classification result and the confidence score.

[0165] In the above embodiment, the X, Y, and Z three-axis acceleration data of the force touch panel are collected based on the acceleration sensor, and the original signal is filtered, analyzed in the time domain and frequency domain, so as to accurately capture the vibration characteristics of the force touch panel in different areas. Through multi-scale time-frequency feature extraction and multi-level decomposition, the amplitude and phase characteristics of the vibration signal of the force touch panel can be accurately analyzed, so as to fully reflect the distribution of vibration intensity in different areas. In particular, through standardization and dimensionality reduction processing, the detection ability of vibration intensity deviation from the design value is effectively enhanced, and the deficiency of the existing technology that it is difficult to quantify the vibration intensity is solved, so as to ensure the uniformity of the vibration intensity and the consistency of response of the force touch panel, and provide users with accurate and reliable operation feedback experience. In addition, through the refined processing of multi-dimensional vibration feature data, the detection accuracy and algorithm efficiency are significantly improved. At the same time, by combining the test report generated by classification prediction and confidence scoring, the vibration performance of the force touch panel can be intuitively and quickly evaluated.

[0166] Based on the same idea as the method in the above-mentioned embodiment, the vibration detection device of the force-sensing touch pad of a laptop computer provided in the present application can implement the method in the above-mentioned embodiment. For the convenience of explanation, the structural schematic diagram of the device embodiment only shows the parts related to the embodiment of the present application. Those skilled in the art can understand that the illustrated structure does not constitute a limitation on the device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0167] Reference Figure 3 In an embodiment of the present application, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, 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 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 data such as a vibration detection method for a force-sensitive touch pad of a laptop computer. 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, a control method for a rotating structure is implemented.

[0168] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, a vibration detection method for a force-sensing touch pad of a laptop computer is implemented, comprising: collecting an original signal of the force-sensing touch pad based on an acceleration sensor, filtering the original signal to obtain an acceleration signal of the force-sensing touch pad, wherein the acceleration signal includes acceleration data of three axes: X, Y, and Z; performing time domain analysis on the acceleration signal, and converting the acceleration signal into a frequency domain signal to obtain the amplitude and phase information of each frequency component; performing multi-level decomposition on the amplitude and phase information of each frequency component, and filtering the original signal to obtain an acceleration signal of the force-sensing touch pad; ... Extract multi-scale time-frequency features from the acceleration signal to obtain a multi-scale feature matrix of the acceleration signal; convert all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data; perform covariance matrix calculation on the standardized acceleration feature data, sort the calculated multiple eigenvalues, select eigenvectors corresponding to eigenvalues ​​with preset values ​​before the sorting results, and establish a reduced-dimensional feature matrix based on the eigenvectors; perform classification and prediction processing on the reduced-dimensional feature matrix to obtain classification results and confidence scores, and generate a detection report in combination with the classification results and confidence scores.

[0169] Those of ordinary skill 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 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 in this application 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 many 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 (RDRAM).

[0170] 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 vibration detection method for a force-sensitive touch pad of a notebook computer, characterized in that: include: Collecting the original signal of the force-sensing touch panel based on the acceleration sensor, filtering the original signal to obtain the acceleration signal of the force-sensing touch panel, wherein the acceleration signal includes acceleration data of three axes: X, Y, and Z; Performing time domain analysis on the acceleration signal, and converting the acceleration signal into a frequency domain signal to obtain amplitude and phase information of each frequency component; Performing multi-level decomposition on the amplitude and phase information of each frequency component, and extracting multi-scale time-frequency features in the acceleration signal to obtain a multi-scale feature matrix of the acceleration signal; Converting all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data; The covariance matrix of the normalized acceleration feature data is calculated, the calculated multiple eigenvalues ​​are sorted, the eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting results are selected, and the dimension reduction feature matrix is ​​established according to the eigenvectors, specifically including: According to the normalized acceleration feature data, the covariance between the dimensions of the acceleration feature data is calculated to obtain a covariance matrix; Performing eigendecomposition on the covariance matrix to obtain eigenvalues ​​of the covariance matrix and eigenvectors corresponding to the eigenvalues; Sorting the eigenvalues ​​in descending order, and selecting eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting result; The feature vectors of the pre-set values ​​are organized into a dimension reduction matrix by columns, and an acceleration feature data matrix is ​​established for the acceleration feature data, and matrix multiplication is performed on the dimension reduction matrix and the acceleration feature data matrix to obtain the dimension reduction feature matrix; The reduced-dimensional feature matrix is ​​subjected to classification prediction processing to obtain a classification result and a confidence score, and a detection report is generated in combination with the classification result and the confidence score.

2. The vibration detection method of a notebook computer force touch pad according to claim 1, characterized in that: The step of collecting the original signal of the force-sensing touch panel based on the acceleration sensor and filtering the original signal to obtain the acceleration signal of the force-sensing touch panel specifically includes: Placing the acceleration sensor in contact with the force-sensing touch panel and initializing the acceleration sensor; Acquire a timestamp of the vibration of the force-sensing touch panel, collect an original signal of the force-sensing touch panel based on the acceleration sensor, perform denoising on the original signal, filter out low-frequency noise and high-frequency noise of the original signal, and obtain a denoised acceleration signal; Performing dynamic range adjustment on the denoised acceleration signal to eliminate amplitude deviation in the denoised acceleration signal and obtain an acceleration signal of uniform magnitude; The uniform magnitude acceleration signal is time-aligned with the timestamp to obtain the time-aligned acceleration signal.

3. The vibration detection method of a notebook computer force touch pad according to claim 1, characterized in that: The step of performing time domain analysis processing on the acceleration signal and converting the acceleration signal into a frequency domain signal to obtain the amplitude and phase information of each frequency component specifically includes: Performing frame processing on the acceleration signal according to a preset time window length and a preset overlap rate to obtain a plurality of time window signals; Performing time domain analysis and processing on each of the time window signals to obtain vibration characteristic data within each time window; Performing fast Fourier transform on the vibration characteristic data in each time window to obtain frequency components of each frequency band; A modular operation is performed on the complex value of each frequency component to obtain the amplitude of each frequency component, and a complex angle of each frequency component is calculated to obtain the phase information of each frequency component.

4. The vibration detection method of a notebook computer force touch pad according to claim 1, characterized in that: The step of performing multi-level decomposition on the amplitude and phase information of each frequency component and extracting the multi-scale time-frequency features in the acceleration signal to obtain the multi-scale feature matrix of the acceleration signal specifically includes: Performing wavelet transform processing on the amplitude and phase information of each frequency component to obtain multi-level wavelet coefficients of the frequency domain signal; Performing inverse transformation on the wavelet coefficients to reconstruct the time-frequency diagram of the acceleration signal at each scale; Refining the time-frequency graph by thresholding and peak detection to extract characteristic components at multiple scales; The characteristic components at each scale are standardized to obtain time-frequency characteristic data; The time-frequency feature data is divided into multiple time windows according to time and frequency, and the feature data in each time window is organized into a multi-scale feature matrix according to scale.

5. The vibration detection method of a notebook computer force touch pad according to claim 1, characterized in that: The step of converting all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data specifically includes: Performing column mean calculation on all feature data of the multi-scale feature matrix to obtain the mean of each column of feature data; Performing a mean removal process on each column of the multi-scale feature matrix according to the mean value to obtain a mean removal feature matrix; Calculating the column variance of the de-meaned feature matrix to obtain the variance of each column of feature data of the de-meaned feature matrix; Standardizing the de-meaned feature matrix according to the variance to obtain a standardized feature data matrix with unit variance; The standardized characteristic data matrix is ​​normalized, and outlier detection is performed on the normalized standardized characteristic data matrix to obtain standardized acceleration characteristic data.

6. The vibration detection method of a notebook computer force touch pad according to claim 1, characterized in that: The step of performing classification prediction processing on the dimension-reduced feature matrix to obtain a classification result and a confidence score, and generating a test report in combination with the classification result and the confidence score specifically includes: Normalizing each row of the reduced-dimensional feature matrix, calculating the eigenvalue range of the reduced-dimensional feature matrix by column and performing a preset scaling to obtain a normalized reduced-dimensional feature matrix; Passing the normalized reduced-dimensional feature matrix as input data to a support vector machine classifier, performing nonlinear mapping processing on the input data according to a kernel function of the support vector machine classifier, generating a classification decision boundary, and outputting a preliminary classification result; According to the decision function distance formula of the support vector machine classifier, the distance value from the input data to the classification decision boundary is calculated, and a confidence distribution is generated for the input data to obtain a confidence score matrix corresponding to the input data; For input data whose confidence in the preliminary classification results is lower than a preset threshold, a random forest classifier is used for re-testing, and a revised classification result is generated by a majority voting method; Divide the sample data according to the modified classification result and the confidence score matrix corresponding to the modified classification result, calculate the confidence mean, variance and coverage of the sample data of each category, obtain the comprehensive confidence eigenvalue of each category, and perform correlation analysis on the confidence eigenvalue of each category with the dimension reduction feature matrix to generate a comprehensive feature analysis table including category distribution and eigenvector relationship; According to the comprehensive feature analysis table, the quantity distribution, confidence range and feature vector direction of the classified samples are statistically analyzed by category to generate a multidimensional performance evaluation index of the vibration characteristics, and the multidimensional performance evaluation index is compared with the preset detection standards item by item to determine whether the vibration performance of the force-sensing touch panel meets the requirements, and a detection report is generated, wherein the detection report includes the classified sample distribution statistics, the confidence analysis chart, the abnormal point feature list and the final pass / fail result.

7. A vibration detection device for a force-sensitive touch pad of a notebook computer, characterized in that: include: An acquisition module, used for acquiring the original signal of the force-sensing touch panel, filtering the original signal to obtain the acceleration signal of the force-sensing touch panel, wherein the acceleration signal includes acceleration data of three axes: X, Y, and Z; An analysis module, used to perform time domain analysis on the acceleration signal and convert the acceleration signal into a frequency domain signal to obtain the amplitude and phase information of each frequency component; A decomposition module performs multi-level decomposition on the amplitude and phase information of each frequency component, and extracts multi-scale time-frequency features in the acceleration signal to obtain a multi-scale feature matrix of the acceleration signal; A conversion module, used for converting all feature data of the multi-scale feature matrix into standard data with zero mean and unit variance to obtain standardized acceleration feature data; The calculation module is used to calculate the covariance matrix of the normalized acceleration feature data, sort the calculated multiple eigenvalues, select the eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting results, and establish a dimension reduction feature matrix according to the eigenvectors, specifically including: According to the normalized acceleration feature data, the covariance between the dimensions of the acceleration feature data is calculated to obtain a covariance matrix; Performing eigendecomposition on the covariance matrix to obtain eigenvalues ​​of the covariance matrix and eigenvectors corresponding to the eigenvalues; Sorting the eigenvalues ​​in descending order, and selecting eigenvectors corresponding to the eigenvalues ​​with preset values ​​before the sorting result; The feature vectors of the pre-set values ​​are organized into a dimension reduction matrix by columns, and an acceleration feature data matrix is ​​established for the acceleration feature data, and matrix multiplication is performed on the dimension reduction matrix and the acceleration feature data matrix to obtain the dimension reduction feature matrix; The generation module is used to perform classification prediction processing on the reduced-dimensional feature matrix to obtain a classification result and a confidence score, and generate a detection report in combination with the classification result and the confidence score.

8. A computer device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and running on the processor, and the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

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 method according to any one of claims 1 to 6 is implemented.

Images