A touch screen false touch prevention control method and system based on multi-sensor fusion

Abstract

The application relates to the technical field of human-computer interaction, and discloses a touch screen false touch control method and system based on multi-sensor fusion. The method acquires touch data and inertial data, and performs comprehensive analysis from three dimensions: based on an original capacitance matrix, divergence characteristics of a gradient vector field are analyzed, and an electric field characteristic index representing an approach mode is output; an amplitude square coherence function of a whole device vibration signal and a touch point micro-displacement signal in a preset physiological frequency band is calculated, and a coupling coefficient representing a physical connection state is output; a direction state probability distribution of a touch point moving track is counted, and Shannon entropy is calculated, and a track entropy value representing operation intention order is output. The method further calculates a false touch confidence of the touch point according to the electric field characteristic index, the coupling coefficient and the track entropy value, and generates a dynamic inhibition mask based on the confidence, so that a false touch event with high confidence is intercepted at the bottom. Through multi-dimensional feature fusion, the application can significantly improve recognition accuracy and robustness.

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, specifically to a touchscreen anti-mistouch control method and system based on multi-sensor fusion. Background Technology

[0002] Touchscreens have become the mainstream input interface for electronic devices such as smartphones, tablets, and in-vehicle terminals. The quality of the user interaction experience largely depends on the touchscreen's responsiveness to valid operation commands and its ability to prevent unintentional touches.

[0003] However, in actual use, accidental touches often affect the user experience. This phenomenon occurs frequently, especially when users hold the device, as non-operational parts such as the palm and knuckles may unintentionally touch the screen surface. This problem is becoming increasingly prominent, particularly with the design trend of increasingly narrow screen bezels.

[0004] To address this issue, existing anti-mistouch technologies typically rely on a single physical characteristic for judgment. For example, a common method is based on the size of the contact area, assuming that touch points exceeding a certain threshold are accidental touches originating from the palm or similar objects. While this method can filter out some obvious large-area contacts, its discrimination capability is limited. It struggles to distinguish between intentional finger presses with similar pressure areas and unintentional palm contact, and may also incorrectly identify valid multi-finger gestures as accidental touches, or fail to recognize accidental touches originating from smaller areas such as knuckles.

[0005] To compensate for the above shortcomings, some solutions have added criteria such as contact duration or the stillness of the contact point position. However, these methods also have limitations. When a user holds the device, their hand may create a contact point that remains in a relatively fixed position for a long time, which is similar to the behavior of a user intentionally long-pressing a screen icon, making it difficult to distinguish. At the same time, slight movements of the hand on the device may also be incorrectly interpreted as intentional sliding operations.

[0006] The fundamental limitation of existing technologies in preventing accidental touches lies in their analytical dimensions, which are confined to the two-dimensional data generated by the touchscreen itself, lacking a deep understanding of the background of touch events. Specifically, these technologies cannot effectively distinguish the physical origin of the touch point: whether it comes from a finger that is separate from the device's movement and actively operates, or from a hand that is mechanically coupled to the device and moves with it. This inherent ambiguity forces existing technologies to compromise between recognition sensitivity and accuracy, making it difficult to simultaneously avoid the loss of effective operations and interference from accidental touches. Summary of the Invention

[0007] The technical problem to be solved by this invention is that existing touch screen anti-mistouch technologies have the problem of single recognition dimension and low accuracy when distinguishing between intentional operation and unintentional contact in complex scenarios; relying solely on single physical characteristics such as contact area or contact duration makes it difficult to accurately identify large-area, long-term static or dynamic mis-touches generated by the palm, knuckles, etc. when holding the device, thus affecting the user experience.

[0008] To address the aforementioned technical problems, this invention provides a touchscreen anti-accidental touch control method and system based on multi-sensor fusion.

[0009] The first aspect of this invention provides a touchscreen anti-mistouch control method based on multi-sensor fusion. This method integrates electric field physical characteristics, biomechanical coupling characteristics, and spatiotemporal trajectory intention characteristics to establish a multi-dimensional comprehensive decision model, so as to achieve accurate discrimination of the nature of touch events.

[0010] In one implementation, the method first acquires touch data and inertial data. The touch data includes the original capacitance matrix and touch point coordinate data, while the inertial data includes the device's acceleration and angular velocity information.

[0011] To distinguish the nature of touch points from their initial physical contact, this method calculates the divergence characteristics of the gradient vector field in the touch point region based on the original capacitance matrix in the touch data, outputting an electric field characteristic index representing the proximity mode. Specifically, by calculating the capacitance gradient vector and divergence characteristics of each sensing node in the capacitance matrix, the geometric shape of the electric field distribution can be quantified. When a small-area conductor such as a fingertip approaches the screen perpendicularly, a local peak is formed in the capacitance value concentration, and its divergence characteristics show strong centripetal convergence; while when a large-area conductor such as a palm contacts the screen tangentially, the resulting capacitance distribution is relatively flat, and the divergence characteristics are not significant. By combining the divergence characteristics and the magnitude of the capacitance gradient vector within the touch point region, this electric field characteristic index can be generated to characterize the "perpendicularity" of the touch point's approach to the screen.

[0012] To distinguish the source of touch points based on their physical connection state, this method calculates the amplitude-squared coherence function of the overall device vibration signal and the micro-displacement signal of the touch point within a preset physiological frequency band, based on touch point coordinate data and inertial data from the touch data. It then outputs a coupling coefficient characterizing the physical connection state between the touch point and the device. When a person holds a device, the inherent physiological vibrations of their muscle system (typically in the 8–12 Hz frequency band) are transmitted to the device and the hand touching the screen through a rigid connection. Therefore, the micro-displacement signal of the touch point generated by the hand has a high linear correlation with the overall device vibration signal within this frequency band. This method calculates the amplitude-squared coherence function between the two and integrates this function within the preset physiological frequency band, provided that the energy of the overall device vibration signal is higher than a preset static noise threshold, to obtain the coupling coefficient. A high coupling coefficient indicates a rigid connection between the touch point and the device, suggesting accidental touch due to gripping; conversely, a low coupling coefficient indicates external operation independent of the device. The amplitude squared coherence function is calculated based on the autopower spectral density of the overall vibration signal of the equipment, the autopower spectral density of the contact micro-displacement signal, and the cross-power spectral density between the two signals, ensuring the accuracy of the analysis.

[0013] To differentiate touchpoint behaviors based on operational intent, this method uses touchpoint coordinate data to statistically analyze the probability distribution of directional states in the touchpoint's movement trajectory and calculates Shannon entropy, outputting a trajectory entropy value that characterizes the orderliness of the operational intent. Intentional swipes or clicks typically exhibit a clear directionality in their trajectory within a local time window, with a concentrated distribution of directional states, exhibiting a low-entropy state. Unintentional micro-movements or jitters, on the other hand, have random directional directions and a uniform distribution of directional states, exhibiting a high-entropy state. This method encodes the touchpoint's movement trajectory into a directional chain code sequence, statistically analyzes the probability of occurrence of each directional state within a preset time window, and calculates the trajectory entropy value accordingly. In one implementation, this trajectory entropy value can also be converted into an intent verification coefficient using a preset S-shaped mapping function for subsequent fusion calculations.

[0014] After obtaining the feature indicators of the above three dimensions, the method calculates the false contact confidence level of the contact point based on the electric field feature indicator, coupling coefficient, and trajectory entropy value. In one embodiment, the false contact confidence level is obtained by weighted linear combination of the electric field feature indicator, coupling coefficient, and an indicator negatively correlated with the trajectory entropy value. This multi-factor fusion model can comprehensively evaluate various attributes of the contact point and arrive at a more reliable judgment than a single feature.

[0015] Finally, the method generates a dynamic suppression mask based on the false touch confidence level and intercepts touch events located within this mask. When the false touch confidence level of a touch point is higher than a preset suppression threshold, the system identifies the original capacitance profile of the touch point and uses a convex hull algorithm to envelop it, generating a dynamic suppression mask. This method generates a mask that accurately fits the actual shape of the false touch area. Touch events located within the mask are intercepted and discarded by the underlying driver. To improve the fault tolerance of the interaction, in one implementation, when the intercepted touch event is a movement event and its trajectory continuously moves out of the dynamic suppression mask, and the movement distance reaches a preset departure threshold, the system will release the interception of the touch event, making it a valid operation.

[0016] A second aspect of the present invention provides a touchscreen anti-mistouch control system based on multi-sensor fusion. This system operates in an electronic device and is used to execute the methods of any of the above embodiments. The system includes a central processing unit, which internally configures:

[0017] The data synchronization acquisition module is used to acquire touch data and inertial data;

[0018] The electric field feature analysis module is used to calculate the divergence characteristics of the gradient vector field of the touch area based on the original capacitance matrix in the touch data, and output the electric field feature index characterizing the proximity mode.

[0019] The biomechanical coupling analysis module is used to calculate the amplitude square coherence function of the overall vibration signal of the device and the micro-displacement signal of the touch point in the preset physiological frequency band based on the touch point coordinate data and the inertial data in the touch data, and output the coupling coefficient characterizing the physical connection state between the touch point and the device.

[0020] The trajectory entropy verification module is used to statistically analyze the probability distribution of the direction state of the contact point movement trajectory based on the contact point coordinate data and calculate the Shannon entropy, and output the trajectory entropy value that represents the orderliness of the operation intention.

[0021] The multi-dimensional decision control module is used to calculate the false touch confidence level of the touch point based on the electric field characteristic index, the coupling coefficient and the trajectory entropy value, and generate a dynamic suppression mask based on the false touch confidence level, and send instructions to the underlying driver to intercept touch events located within the dynamic suppression mask.

[0022] This invention provides a touchscreen anti-mistouch control method and system based on multi-sensor fusion. It has the following beneficial effects:

[0023] 1. This invention makes a comprehensive judgment by integrating three orthogonal dimensions of physical meaning: electric field morphology, biomechanical coupling, and motion trajectory intention. This greatly improves the accuracy and robustness of mis-touch recognition in complex scenarios (especially single-handed holding scenarios) and effectively reduces the false judgment rate.

[0024] 2. This invention uses a convex hull algorithm based on the original capacitor profile to generate a suppression mask, which can accurately and dynamically match the actual shape and size of the accidental touch area, avoiding interaction problems caused by the suppression range being too large or too small.

[0025] 3. This invention intercepts high-confidence accidental touch events at the operating system level, preventing invalid data from entering upper-layer applications and ensuring system smoothness and real-time response to user operations. Attached Figure Description

[0026] Figure 1 This is a structural block diagram of a touchscreen anti-mistouch control system based on multi-sensor fusion according to an embodiment of the present invention;

[0027] Figure 2 This is a flowchart of a touchscreen anti-accidental touch control method based on multi-sensor fusion according to an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of protection of the present invention.

[0029] See attached document Figure 1 , Figure 1 This is a structural block diagram of a touchscreen anti-mistouch control system based on multi-sensor fusion according to an embodiment of the present invention. The present invention provides a touchscreen anti-mistouch control system based on multi-sensor fusion, which operates in an electronic device, including but not limited to smartphones, tablets, or in-vehicle touch terminals. At the hardware level, the system includes a touch sensing unit, an inertial measurement unit, and a central processing unit. The touch sensing unit detects the capacitance signal of an external conductor and outputs touch data, while the inertial measurement unit detects the device's motion posture and outputs inertial data.

[0030] The central processing unit is equipped with a logic processing module, which includes a data synchronization acquisition module, an electric field characteristic analysis module, a biomechanical coupling analysis module, a trajectory entropy verification module, and a multi-dimensional decision control module.

[0031] The data synchronization acquisition module connects the touch sensing unit and the inertial measurement unit to acquire raw capacitance matrix data, touch point coordinate data, and six-axis inertial data including acceleration and angular velocity, respectively. The data synchronization acquisition module is equipped with time alignment logic to synchronize heterogeneous signals based on a unified timestamp reference, and performs interpolation upsampling processing on touch point coordinate data with lower sampling rates, outputting time-synchronized touch sequences and inertial sequences.

[0032] The electric field feature analysis module, connected to the data synchronization acquisition module, is used to receive the raw capacitance matrix data. The electric field feature analysis module is configured to calculate the capacitance gradient vector of the contact area, quantify the convergence degree of the electric field distribution based on the divergence characteristics of the gradient vector field, and output electric field feature indices characterizing the proximity mode.

[0033] The biomechanical coupling analysis module, connected to the data synchronization acquisition module, receives synchronized contact coordinate data and inertial data. The biomechanical coupling analysis module is configured to perform frequency domain analysis, calculating the amplitude squared coherence function of the overall equipment vibration signal and the contact micro-displacement signal within a preset physiological frequency band, and outputting the coupling coefficient characterizing the physical connection state between the contact and the equipment.

[0034] The trajectory entropy verification module, connected to the data synchronization acquisition module, is used to receive touch point coordinate data. The trajectory entropy verification module is configured to encode the direction of the touch point's movement trajectory, statistically analyze the probability distribution of the direction state, calculate the Shannon entropy, and output a trajectory entropy value representing the ordered nature of the operation intention.

[0035] The multi-dimensional decision control module is connected to the electric field feature analysis module, the biomechanical coupling analysis module, and the trajectory entropy verification module, respectively. The multi-dimensional decision control module calculates the false touch confidence level of the touch point based on the electric field feature indicators, coupling coefficient, and trajectory entropy value, and generates a dynamic suppression mask accordingly, sending touch interception commands for specific areas to the underlying driver.

[0036] See attached document Figure 2 , Figure 2 This is a flowchart of a touchscreen anti-mistouch control method based on multi-sensor fusion according to an embodiment of the present invention. The method is executed through the aforementioned system and specifically includes the following steps:

[0037] Step S100: Perform synchronous acquisition and preprocessing of multi-source heterogeneous signals.

[0038] In step S100, the system simultaneously initiates capacitive touch scanning and inertial measurement scanning. The data synchronization acquisition module acquires the touch point centroid coordinate sequence and the original capacitance image output by the touchscreen at a first sampling rate, and acquires the device acceleration vector and angular velocity vector output by the inertial measurement unit at a second sampling rate. The second sampling rate is greater than the first sampling rate. The data synchronization acquisition module aligns the two data streams using timestamps and improves the temporal resolution of the touch point centroid coordinate sequence to be consistent with the inertial data using a Lagrange interpolation algorithm, forming a synchronization signal sequence under the same time reference.

[0039] Step S200: Perform proximity feature extraction based on electric field gradient flow.

[0040] In step S200, when an object is detected approaching or touching the screen, the electric field feature analysis module constructs an electric field gradient flow field based on the original capacitance image. This step analyzes the divergence characteristics of the gradient flow in the local neighborhood of the touch point by calculating the gradient vector of each node in the capacitance matrix. When the gradient flow exhibits a centripetal convergent distribution, it is determined to be a vertical touch feature; when the gradient flow exhibits a unidirectional parallel or diffuse distribution, it is determined to be a tangential contact feature. Based on this, the electric field feature analysis module outputs quantified electric field feature indicators.

[0041] Step S300: Perform rigid coupling analysis based on biomechanical tremor coherence.

[0042] In step S300, the biomechanical coupling analysis module extracts the overall angular velocity magnitude signal of the device and the micro-displacement velocity signals of each contact point in the screen coordinate system. This step uses a power spectral density estimation method to calculate the cross-power spectrum and the individual self-power spectrum of the angular velocity magnitude signal and the micro-displacement velocity signal. Based on the spectral density calculation results, the amplitude squared coherence function of the two signals within the human physiological tremor frequency band is obtained. This step calculates the average coherence within this frequency band through integration to obtain the coupling coefficient. The coupling coefficient is used to characterize whether there is a mechanical rigid connection between the external conductor corresponding to the contact point and the device.

[0043] Step S400: Perform intent depth verification based on spatiotemporal trajectory entropy.

[0044] In step S400, for contacts identified as non-rigidly connected, the trajectory entropy verification module discretizes their motion trajectory within a preset time window. This step calculates the motion direction angles of adjacent sampling points and maps continuous direction angles to discrete direction status codes. The trajectory entropy verification module statistically analyzes the frequency of occurrence of each direction status code within the time window and calculates the information entropy of the trajectory sequence. The trajectory entropy value is used to distinguish between intentional operations with clear directionality and unintentional micro-movements with randomness.

[0045] Step S500: Perform multi-dimensional fusion decision and dynamic mask generation.

[0046] In step S500, the multi-dimensional decision control module performs a weighted calculation on the feature values ​​output from steps S200, S300, and S400 to obtain a false touch confidence score for each touch point. The system selects touch points with false touch confidence scores exceeding a preset threshold as reference points and uses a geometric algorithm to generate an irregular polygonal region covering the reference points and their neighborhood. This region is defined as a dynamic suppression mask. The multi-dimensional decision control module marks touch events located within the dynamic suppression mask as invalid events and intercepts them, while marking touch events located outside the dynamic suppression mask as valid events and reporting them to the application layer.

[0047] To further explain the implementation principle of this technical solution, the following section will combine specific mathematical models and signal processing logic to discuss the above. Figure 2 Each key step in the process shown is explained in detail.

[0048] For step S100, which involves the synchronous acquisition and preprocessing of multi-source heterogeneous signals, the data synchronization acquisition module executes specific signal processing logic. This module eliminates the asynchrony in sampling frequency between the touch sensing unit and the inertial measurement unit by establishing a shared time reference.

[0049] Step S100 specifically includes the following steps:

[0050] Step S101: Define and acquire multidimensional raw signals.

[0051] The system establishes a unified time reference axis and defines time. The original capacitor matrix of the capacitive touchscreen is This matrix consists of mutual capacitance or self-capacitance values ​​collected by the touchscreen sensing channels, and has a dimension of [missing information]. ,in Represents the number of row channels. This represents the number of column channels. After processing and coordinate calculation by the analog front-end (AFE) of the touch chip, it outputs the set of touch points recognized in the current frame. ,in This represents the total number of currently detected touch points. For any given number of touch points in the set... Contact points Its state vector The definition is as follows:

[0052] ;

[0053] In the formula, and These represent the x and y coordinates of the touch point's centroid in the screen coordinate system, respectively. This indicates the effective contact area of ​​the contact point.

[0054] Meanwhile, the inertial measurement unit operates at an independent high-frequency sampling rate, outputting the device's motion state vector. :

[0055] ;

[0056] In the formula, , , These represent the linear acceleration components along the X, Y, and Z axes of the equipment body, respectively. , , These represent the angular velocity components of the rotation around the X, Y, and Z axes of the device body, respectively. All of these signals include microsecond-level timestamps generated by the system's underlying hardware clock.

[0057] Step S102: Perform timestamp alignment of heterogeneous data streams.

[0058] Due to the frequency of touch reporting (Typically 60Hz to 120Hz) and the sampling frequency of the inertial measurement unit. There is a difference between the two clock sources (typically 100Hz to 400Hz or higher), and their clock sources may drift. The data synchronization acquisition module maintains two independent circular buffers, one for storing touch data frames and the other for storing inertial data frames. When the system acquires a new inertial data frame, it records its timestamp. At that time, the module performs a backtracking retrieval in the touch data buffer and locks the time window. The internal touch data sequence, in which This is the historical time length required for correlation analysis. If there is a transmission delay in the touch data, the system performs minor compensation on the latest touch coordinates based on linear extrapolation to ensure that the data frames entering the calculation process are physically aligned with the time axis of the inertial data.

[0059] Step S103: Perform interpolation upsampling of the contact coordinate sequence.

[0060] Regarding touch reporting frequency Below the inertial sampling frequency In this case, an interpolation algorithm is used to construct a continuous contact motion trajectory function, and resampling is performed according to the timestamp sequence of the inertial data. In this embodiment, to balance computational efficiency and curve smoothness, a four-point Lagrange interpolation method (i.e., order-wise) is used. ).

[0061] Let the sequence of the four most recent discrete contact coordinates retrieved before the current time be . For any inertial data sampling time (in (The interpolated x-coordinate) The calculation formula is:

[0062] ;

[0063] in, Here, the expression for the Lagrange basis function is:

[0064] ;

[0065] Similarly, for the ordinate Perform the same operation to obtain This step ensures that each frame of inertial data has a corresponding, high-precision contact coordinate to match, thereby eliminating phase jitter.

[0066] Step S104: Calculate the micro-displacement velocity signal and the equipment vibration signal.

[0067] Synchronous coordinate sequence generated by interpolation System calculation contacts The instantaneous micro-displacement velocity. To suppress quantization noise caused by digital sampling, this embodiment uses a first-order difference combined with an exponentially weighted moving average filter for calculation. Define time... Original speed of the contact point With filtered speed as follows:

[0068] ;

[0069] ;

[0070] In the formula, The sampling interval for inertial data (i.e. ); This is the filtering smoothing coefficient, typically ranging from [0.1, 0.3], used to filter out high-frequency noise above the physiological tremor frequency band.

[0071] Simultaneously, the system extracts the angular velocity magnitude signal reflecting the overall vibration intensity of the equipment from the inertial data. Since physiological tremor in handheld devices mainly manifests as minute rotational shaking, the angular velocity modulus can effectively characterize the tremor features of the hand.

[0072] ;

[0073] After processing steps S101 to S104, the system outputs two sets of time-synchronized signal sequences with the same sampling rate. and The signal sequence is stored in a sliding window of a preset length, serving as the input source for the subsequent biomechanical coupling analysis in step S300.

[0074] For step S200, which involves performing proximity feature extraction based on electric field gradient flow, the electric field feature analysis module performs specific field feature analysis based on the original capacitance data during the hovering phase or initial contact stage when an external conductor is detected approaching the touch panel but before a stable touch event is triggered. This process aims to utilize the spatial distribution of capacitance signals to predict the proximity pattern of the external conductor before or at the moment of physical contact, thereby distinguishing between vertical click operations and large-area tangential accidental touches.

[0075] Step S200 specifically includes the following steps:

[0076] Step S201: Construct the gradient vector field of the original capacitance matrix.

[0077] The electric field feature analysis module acquires the raw capacitance image frame output by the touch sensing unit. This raw capacitance image frame is composed of an array of capacitance values ​​at the intersections of the sensing channels. Arbitrary sensing nodes are defined within the raw capacitance matrix. The capacitance value at that location is ,in For row index, This is the column index. To accurately capture the drastic change and direction of the electric field, the system uses the discrete difference method to calculate the rate of change of capacitance at this node in the row and column directions, constructing a capacitance gradient vector. This gradient vector represents the direction and rate of change of the electric field intensity in its spatial distribution, and its calculation formula is as follows:

[0078] ;

[0079] In the formula, and These represent the discrete first-order differences of the capacitance value in the row and column directions, respectively. By performing the above calculations on all nodes in the matrix and ignoring boundary points, the system generates a two-dimensional gradient vector field with the same dimension as the original capacitance matrix.

[0080] Step S202: Calculate the divergence characteristics of the gradient flow field.

[0081] To quantify the geometric shape of the electric field distribution, the system further calculates the divergence of the gradient vector field. In this embodiment, the discrete Laplace operator is used to calculate the nodes. Second-order differential characteristic at ... . The specific discrete calculation formula is used to characterize the source and sink properties of gradient flow in local space as follows:

[0082] ;

[0083] The above formula is implemented using a four-neighbor convolution kernel. When the value is less than the first preset negative threshold, it indicates that the gradient vector around that point exhibits a centripetal convergent distribution, corresponding to the local maximum electric field distribution formed when a small-area conductor (such as a fingertip) approaches the screen; when When the value is close to zero or positive, it indicates that the gradient vector around the point is parallel or divergent, corresponding to the gentle electric field distribution formed when a large-area conductor (such as the edge of a palm) is close to the screen.

[0084] The first preset negative threshold is dynamically set based on the statistical characteristics of the current ambient noise. Specifically, the system calculates the standard deviation of the background noise when the screen is idle. ,Will Set as:

[0085] ;

[0086] In the formula, The sensitivity coefficient is preferably set within the range of [3.0, 5.0]. This ensures that a signal feature is considered valid only when the convergence intensity of the local electric field exceeds 3 to 5 times the standard deviation of the background noise.

[0087] Step S203: Quantize the proximity mode and output electric field characteristic indicators.

[0088] For each potential touchpoint The system defines its region of interest in the original capacitance matrix. This region, with the integer index corresponding to the centroid coordinates of the contact point as its geometric center, covers a set of sensing nodes of a preset size (e.g., 5×5). The electric field feature analysis module... By combining gradient magnitude and divergence characteristics within the range, a normalized electric field characteristic index is calculated. This is used to numerically characterize the "verticality" of proximity patterns. The calculation formula is as follows:

[0089] ;

[0090] In the formula, This represents the total number of nodes within the region of interest. The L2 norm (modulus) of the gradient vector is represented by the gradient vector. To prevent extremely small positive numbers with a denominator of zero.

[0091] The calculation formula shows that for a valid finger tap, due to the significant capacitance peak in the central region, its divergence... Negative numbers with large absolute values ​​(i.e., below the first preset negative threshold) Furthermore, the gradient magnitude is relatively large, resulting in... It shows a positive high value; while for accidental touches by the palm, due to the gentle charge distribution, the divergence is low. The absolute value is small, making The result will be a low or negative value. The system will calculate... The cache serves as the first-level input feature for the subsequent multi-dimensional decision control module to determine the confidence level of false touches.

[0092] For step S300, which involves performing a rigid coupling analysis based on biomechanical vibration coherence, the biomechanical coupling analysis module executes specific frequency domain signal processing logic. This step utilizes the inherent physiological vibration characteristics of the human musculoskeletal system to determine whether a rigid transmission path exists between the external conductor corresponding to the contact and the device casing by calculating the synchronization index of the overall vibration signal of the device and the micro-displacement signal of the contact in the frequency domain.

[0093] The specific implementation process of step S300 includes the following steps:

[0094] Step S301: Construct a time-domain analysis frame and perform detrending and windowing processing.

[0095] The biomechanical coupling analysis module reads data of length from the output queue of the data synchronization acquisition module. The data segment. To ensure that the time-domain data contains a sufficient number of physiological tremor cycles to meet the accuracy requirements of frequency-domain analysis, the time length of the data segment is configured to be no less than 0.5 seconds. For example, when the sampling frequency... At 200Hz, the number of data points The preferred configuration is 128 to 256 points.

[0096] The system extracts the angular velocity magnitude signal sequence of the equipment respectively. With contact micro-displacement velocity signal sequence ,in To eliminate the DC component and non-periodic extremely low-frequency trend term caused by slow changes in device attitude in the signal, the system first performs detrending processing on both sets of signals. Specifically, the mean of the signal segment is calculated and subtracted to generate a zero-mean signal. :

[0097] ;

[0098] ;

[0099] Subsequently, to suppress spectral leakage caused by data truncation during the Discrete Fourier Transform, the system uses a Hanning or Hamming window to weight the zero-mean signal. The window function is defined. Vibration signals of equipment after adding windows With contact micro-displacement signal The calculation is as follows:

[0100] ;

[0101] ;

[0102] ;

[0103] The output of this step and This is the standard time-domain frame used for subsequent frequency-domain analysis.

[0104] Step S302: Estimate the self-power spectral density and cross-power spectral density based on the Welch method.

[0105] This embodiment employs the Welch power spectral density estimation method, dividing long data frames into multiple overlapping short data segments for averaging to reduce the impact of random noise on spectral estimation. The biomechanical coupling analysis module then processes the aforementioned data frames of length... Signal segmentation There are overlapping segments, each with a length of [number]. The overlap rate is 50%. Among them, the segment length... The selection must meet the frequency resolution requirements. The resolution should be less than a preset resolution threshold (e.g., 1 Hz) to ensure that the characteristic peaks of physiological tremor in the 8 Hz to 12 Hz frequency band can be distinguished.

[0106] For the Data segments The system calculates the discrete Fourier transforms of the equipment vibration signal and the contact micro-displacement signal, respectively. Definition and The first Segment signal at frequency Spectral coefficients at:

[0107] ;

[0108] ;

[0109] In the formula The value is an imaginary unit. Based on the spectral coefficients of each segment, the system calculates the average corrected periodogram, and then obtains the autopower spectral density of the equipment vibration signal. And the self-power spectral density of the contact micro-displacement signal. :

[0110] ;

[0111] ;

[0112] Simultaneously, the cross-power spectral density between the two sets of signals is calculated. The cross-power spectral density reflects the average of the complex conjugate product of two signals at a specific frequency, preserving phase information.

[0113] ;

[0114] In the formula, express Complex conjugate; The normalization coefficient is used to compensate for energy loss caused by the window function, and its calculation formula is as follows: Through the above calculations, the system obtains a value characterizing the vibration energy distribution of the equipment in the frequency domain. Characterizing the distribution of micro-motion energy at the contact point and characterizing the correlation between the two These three sets of spectral density functions form the computational basis for determining rigid coupling relationships.

[0115] After obtaining the power spectral density and cross-power spectral density of the device vibration signal and the contact micro-displacement signal, for step S300, the biomechanical coupling analysis module further performs the construction of the frequency domain amplitude square coherence function and the characteristic integration within a specific frequency band to quantitatively evaluate the physical connection tightness between the contact and the device.

[0116] Step S300 also includes the following steps:

[0117] Step S303: Construct the frequency domain amplitude squared coherence function.

[0118] Based on the self-power spectral density calculated in step S302 , and cross-power spectral density Biomechanical coupling analysis module calculates contact points With the overall movement of the equipment at frequency The amplitude squared coherence function (MSC) at that point. This is used to normalize the linear correlation between two signals in the frequency domain, and its numerical range is strictly defined within the interval [0,1]. The calculation formula is as follows:

[0119] ;

[0120] In the formula, Represents the magnitude of the cross-power spectral density; To prevent extremely small positive numbers with a denominator of zero, the preferred value is 10. −10 Up to 10 −6 Magnitude.

[0121] In this computational model, when When the value approaches 1, it represents the frequency At this point, a significant linear causal relationship exists between the micro-displacement signal of the contact and the vibration signal of the device. This corresponds to the rigid transmission characteristics in the gripping state, that is, the displacement of the contact is generated by the overall vibration of the device through the rigid connection between the hand bones and muscle tissue.

[0122] Conversely, when When the value approaches 0, it indicates that the two are uncorrelated at that frequency, corresponding to the independent motion characteristics in the non-grip state. That is, the displacement of the touch point is mainly generated by the active muscle force of the finger clicking in the air, and is not modulated by the vibration of the device.

[0123] Step S304: Perform rigid coupling coefficient integration within the physiological tremor frequency band.

[0124] To suppress interference from user-initiated actions (low-frequency domain) and environmental noise (high-frequency domain) on the judgment, the biomechanical coupling analysis module is configured to extract features of the coherence function only within the physiological tremor frequency band of the human musculoskeletal system. The system presets the lower limit frequency of the physiological tremor frequency band as follows: The upper limit frequency is Based on the micro-tremor characteristics of human neuromuscular systems, this embodiment sets... =8Hz, =12Hz.

[0125] Before calculating the coupling coefficient, the system first performs an effective vibration energy determination. This involves calculating the total vibration energy of the equipment within the specified frequency band. :

[0126] ;

[0127] like Less than the preset static noise threshold This indicates that the device is in a static or stable position, and there is no excitation source sufficient to trigger physiological tremor resonance. In this case, the coupling coefficient can be directly set to... .

[0128] like Then, weighted integral calculation within the frequency band is performed. Considering the difference in signal-to-noise ratio at different frequency points, this embodiment uses the power-weighted average method to calculate the rigid coupling coefficient. This method uses the self-power spectral density of equipment vibration as a weighting factor, and the calculation formula is as follows:

[0129] ;

[0130] In the formula, Represents discrete frequency points within a frequency band; and Corresponding frequencies and Index value in the discrete spectrum.

[0131] Through the above weighted calculation, the system can adaptively enhance the weight of the main flutter frequency (i.e., the frequency point with the most concentrated energy) in the final score and suppress random coherent noise at non-flutter frequencies, thereby obtaining highly robust coupling characteristics. The calculated... The data is directly output to the multi-dimensional decision control module as a key quantitative basis for determining whether the contact point is a case of accidental touch by the holding hand.

[0132] For step S400, which involves performing intent depth verification based on spatiotemporal trajectory entropy, the intent verification module performs dynamic orderliness analysis on the movement trajectory of the touch point on the screen surface. This step uses statistical entropy values ​​to quantify the dispersion of the touch point's movement direction distribution, thereby distinguishing between "intentional sliding operations" with clear directional guidance and "unintentional accidental touch interference" with random directional distribution.

[0133] Typically, a user's intentional swipes exhibit high directional consistency within a local time window, resulting in a low-entropy state; while touches on the edge of the palm or unconscious knuckles are often accompanied by slight, randomly diverging tremors, resulting in a high-entropy state.

[0134] Step S400 specifically includes the following steps:

[0135] Step S401: Perform direction chain coding of the contact trajectory.

[0136] Intent verification module obtains touch points The system uses a set of coordinate points on a continuous time series. To eliminate the influence of different screen resolutions and scrolling speeds on the calculation, the system first maps the touch point trajectory to a discrete sequence of direction chain codes. The touch point at time [time value missing] is defined. The position is To prevent quantization noise caused by excessively high sampling frequency leading to small single-frame displacement, the system selects a time interval. Corresponding historical moment position Perform differential calculations. The system calculates the displacement vector. and its direction angle :

[0137] ;

[0138] In the formula, It is a two-parameter arctangent function with an output range of (−π,π).

[0139] Subsequently, the system will continuously adjust the azimuth angle. Quantified as A discrete directional sector. In this embodiment, it is preferable to set... This means that 8-neighborhood direction coding is used, with each sector covering an angle range of 45°. The quantized direction chain code... The calculation formula is as follows:

[0140] ;

[0141] In the formula, This indicates the floor function. This indicates the modulo operation. This formula ensures that the center of each quantization sector is aligned with the standard orientation axis (such as 0 degrees, 45 degrees, 90 degrees) by introducing a 1 / 2 phase offset.

[0142] Through this step, the continuous contact trajectory is transformed into a set of integers. Composed of time series .

[0143] Step S402: Calculate the probability distribution of the trajectory direction state.

[0144] To evaluate the orderliness of the trajectory within a local time range, the intent verification module sets a length of... A sliding observation window (e.g., containing 10 to 20 non-zero displacement points). Within the chain code sequence in the current window, the system statistically analyzes each discrete direction. Frequency of occurrence ,in .

[0145] Based on the frequency statistics, calculate the probability of each directional state occurring within the current window. :

[0146] ;

[0147] The probability distribution This reflects the concentration of the contact point's movement direction. If the trajectory is a straight line, the probability will be concentrated in one or two specific directions; if the trajectory is a random perturbation, the probability will tend to be evenly distributed in all directions.

[0148] Step S403: Calculate the trajectory Shannon entropy and generate a discriminant index.

[0149] Based on the probability distribution obtained from statistics The system uses the Shannon entropy model to calculate the spatiotemporal entropy value of the trajectory. To facilitate the setting of a universal decision threshold, this embodiment normalizes the entropy value, limiting its range to [0,1]. Normalized trajectory entropy. The calculation formula is as follows:

[0150] ;

[0151] In the formula, To prevent the division of logarithms into extremely small positive numbers that are undefined; denominator For the system in The maximum entropy that can be achieved when all directions are completely uniformly distributed.

[0152] The normalized trajectory entropy It can be directly used as a quantitative indicator of intent depth. When Less than the preset orderliness threshold (For example When the contact movement exhibits significant directionality, it is classified as "high intent confidence"; when... Greater than When this occurs, it indicates that the contact point movement exhibits disordered random motion characteristics or stationary shaking characteristics, and is judged as "low intention confidence".

[0153] Step S404: Output the intent verification coefficient.

[0154] The system is based on calculations. Generate the final intent verification coefficient. To achieve a soft handover of the decision result and avoid system response jumps caused by entropy fluctuations near the threshold, this embodiment uses a sigmoid mapping function to convert the entropy value into a gain coefficient:

[0155] ;

[0156] In the formula, This is the slope control factor, with a preferred value range of [10, 20]. This parameter determines the width of the transition band. The larger the value, the closer the judgment is to the binary hard threshold; The smaller the value, the smoother the transition. (Output) The closer the value is to 1, the more orderly the trajectory and the clearer the operation intention; the closer the value is to 0, the more chaotic the trajectory and the higher the possibility of accidental touch. This coefficient will serve as one of the weighting factors of the multi-dimensional decision control module to adjust the final touch response strategy.

[0157] For step S500, which is to perform comprehensive decision-making and suppression mask generation based on multi-dimensional features, the multi-dimensional decision control module fuses the feature indicators output by the aforementioned three modules to obtain the final false touch confidence of the touch point, and dynamically generates the suppression region based on this confidence.

[0158] Step S500 specifically includes the following steps:

[0159] Step S501: Construct a multi-factor false touch confidence scoring model.

[0160] The multi-dimensional decision control module receives electric field characteristic indices from the electric field characteristic analysis module. Rigid coupling coefficients from the biomechanical coupling analysis module And intent verification coefficients from the intent verification module The system calculates the contact points using a weighted linear combination model. Overall false touch confidence :

[0161] ;

[0162] In the formula, , , These are the weight coefficients for the corresponding features, satisfying the normalization condition. In the preferred one-handed grip scenario, since accidental touches are the primary issue, the weighting is optimally configured as follows: For example, it can be set , , In non-grip scenarios such as two-handed gaming or flat-handed operation, weighting It can be reduced to enhance. and Contribution to the detection of accidental touches.

[0163] Electric field characteristic index The higher the value, the more it tends to be vertically close; this item is related to... Positive correlation.

[0164] rigid coupling coefficient The higher the value, the stronger the correlation with device vibration, and the more likely it is to be accidentally touched by the user's hand. This item is related to... Positive correlation.

[0165] Intent verification coefficient The closer the value is to 1, the higher the orderliness of the trajectory, and the less likely it is to be a false trigger. Therefore, using... Item, making it with Negative correlation.

[0166] This model unifies three physically independent features into a single quantifiable index. .

[0167] Step S502: Dynamic suppression mask generation based on convex hull algorithm.

[0168] For each false touch confidence level Higher than the preset suppression threshold Contacts The multidimensional decision control module performs the generation of a dynamic suppression mask.

[0169] First, the system identifies the contact point. The original capacitance profile, i.e., the capacitance difference value after background subtraction and filtering is greater than the second preset threshold. Connected regions Second preset threshold Set as the standard deviation of background noise Times, preferred Subsequently, the module uses the minimum convex hull algorithm on the point set. Perform envelope calculation to obtain the convex hull. .

[0170] The convex hull can be calculated using either the Graham scan method or the Monoton Chain algorithm. Using the convex hull... It can effectively encompass the minute jitter of the contact area and its edges, and has high computational efficiency. The final generated suppression mask area... That is, the convex hull. Extend outward by a safe distance The area behind. This safe distance. It is preferable to set it in the range of 5mm to 10mm to allow slight fluctuations at the edge of the contact area when the contact is stationary.

[0171] Step S503, underlying interception strategy within the mask area.

[0172] Dynamic suppression mask area After generation, the multidimensional decision control module pushes the information of this region to the underlying input event processing module of the operating system.

[0173] At the underlying level, the input event handling module handles all new raw touch input events. Perform geometric position verification. If the event... coordinates Located in any activated inhibition mask The interior, that is The system will then execute the following strategy:

[0174] 1. For the Down event of a touch event: directly intercept and discard the event, do not dispatch it to the upper-level application or window manager, and prevent accidental activation of the application or control.

[0175] 2. For Move events (movement) of touch events: If the Move event follows the previously intercepted event, it will continue to be intercepted and discarded. If the trajectory of the Move event continues to move out of the mask area... And the movement distance reaches the preset escape threshold. (Preferably set in the range of 10mm to 20mm), the system then releases the suppression of the contact and dispatches the point as a new valid contact (Down event) to the upper layer.

[0176] 3. For the Up event (lift) of a touch event: Intercept and discard, and then mask the event. The status is set to pending clearing, waiting for the next valid touch event to occur or for the timer to expire before being cleared.

[0177] This interception, performed at the lowest level of the input event stream (operating system kernel or HAL layer), avoids sending high-confidence accidental touch data to the upper-level driver or application layer for processing, ensuring the highest response efficiency.

[0178] Example:

[0179] This application example illustrates a scenario where a user, holding a mobile device with one hand, accidentally touches the edge of the screen with the heel of their palm or the inside edge of their thumb and attempts to perform an intentional click. The system processes and determines this action in real time according to steps S100 to S500 of the present invention.

[0180] Scene setting:

[0181] Device status: The mobile device is held in one hand with the screen facing up.

[0182] Input 1 (accidental touch source): The edge of the palm heel (a large-area conductor) of the holding hand comes into contact with the lower left corner of the screen, creating a low-speed, high-coupling contact. .

[0183] Input 2 (Valid Operation): The user's other hand (non-holding hand) taps the center of the screen with its fingertip, creating a high-speed, low-coupling touch point. .

[0184] Physical state: The holding hand is maintaining the posture of the device, and its muscle tremors are transmitted through the device casing. The inertial sensor continuously detects weak vibration signals of 8 to 12 Hz.

[0185] The implementation process is as follows:

[0186] Step S100: Synchronous Acquisition and Preprocessing of Multi-Source Heterogeneous Signals

[0187] The system collects and aligns touch and inertial data using a unified time base. Because the touch frequency is lower than the inertial frequency, the system tracks the coordinate sequences of the two touch points. Four-point Lagrange interpolation is performed to resample the contact trajectory onto the time axis of the inertial sensor. Subsequently, the system calculates the micro-displacement velocities of the two contacts. and overall vibration speed of the equipment This yields a time-synchronized signal sequence.

[0188] Step S200: Proximity feature extraction based on electric field gradient flow

[0189] The electric field characteristic analysis module calculates the divergence characteristics of the gradient flow based on the original capacitance matrix. .

[0190] For contacts (Heel of the palm): Due to the large contact area and dispersed pressure, the charge distribution is gentle, and the local dispersion is low. The absolute value is small, and most nodes The calculated electric field characteristic parameters Presenting low values ​​(e.g.) =0.1).

[0191] For contacts (Fingertip): Small contact area, high and concentrated capacitance peaks, dramatic gradient changes, and localized divergence. The absolute value is large, and most nodes The calculated electric field characteristic parameters High values ​​(e.g.) =0.8).

[0192] Step S300: Rigid coupling analysis based on biomechanical tremor coherence

[0193] The biomechanical coupling analysis module calculates the spectral density based on S302. , and Construct the amplitude squared coherence function .

[0194] For contacts (Holding Hand): A rigid connection exists between the palm and the device casing, allowing physiological vibrations (8–12 Hz) from the holding hand to be efficiently transmitted to the device and contact traces. Therefore, within the physiological vibration frequency band [8 Hz, 12 Hz], the coherence function... Approaching 1. The rigid coupling coefficient obtained by weighted integral calculation. High values ​​(e.g.) =0.9).

[0195] For contacts (Non-holding hand): Clicking with fingertips suspended in the air results in an instantaneous and elastic connection to the device, leading to extremely low vibration transmission efficiency. Within the physiological tremor frequency band, the coherence function... Approaching 0. The calculated rigid coupling coefficient Presenting low values ​​(e.g.) =0.1).

[0196] Step S400: Intent depth verification based on spatiotemporal trajectory entropy

[0197] The intent verification module analyzes the minute displacement trajectory of the touch point. Calculate its normalized trajectory entropy .

[0198] For contacts (Heel of the palm): The heel of the palm is usually still or accompanied by unconscious, minute tremors, with highly random trajectories and chain code sequences. Uniform distribution. Calculated trajectory entropy. High values ​​(e.g.) =0.8), the intent verification coefficient calculated using the sigmoid mapping function. Presenting low values ​​(e.g.) =0.2).

[0199] For contacts (Fingertip): The user performed a clear click operation, and the micro-displacement of the finger before and after the Down event showed high directional consistency. The calculated trajectory entropy... Presenting low values ​​(e.g.) =0.3), Intent verification coefficient High values ​​(e.g.) =0.9).

[0200] Step S500: Multidimensional Decision Control and Dynamic Inhibition

[0201] The multi-dimensional decision control module weights and fuses various feature indicators, employing an optimal weighting method. =0.5, =0.25, =0.25 for calculation:

[0202] 1. Calculate the confidence level of false triggers :

[0203] (Accidental touch with the heel of the palm):

[0204] ;

[0205] ;

[0206] (Effective Operation):

[0207] ;

[0208] ;

[0209] 2. Generate a suppression mask:

[0210] Set a preset suppression threshold =0.5.

[0211] =0.675>0.5, the system determines This is a high-confidence false trigger. The multi-dimensional decision control module immediately... The original capacitance region Generate dynamic convex hull and expand the safe distance outward. Generate suppression mask .

[0212] =0.275 < 0.5, the system determines For effective operation, no suppression mask is generated.

[0213] 3. Low-level event interception:

[0214] All from The original touch Down / Move events within the coordinate range, due to It is directly intercepted and discarded by the underlying input module of the operating system.

[0215] All from The Down / Move events are normally dispatched to the upper-layer application.

[0216] This embodiment has been successfully implemented:

[0217] Accidental touch suppression: prevent accidental touch from the heel of the palm. The confidence level is calculated to be high, and a suppression mask is immediately generated at the bottom layer to prevent the Down event of the accidental touch event from entering the application layer, effectively preventing the application from being accidentally activated.

[0218] Valid access: Tap with your fingertip The confidence level is calculated to be low, ensuring that the valid operation can be received and processed normally by the system.

[0219] The results show that the present invention achieves accurate real-time identification and efficient low-level suppression of accidental touches in complex scenarios by combining a comprehensive multi-factor model that integrates electric field geometric features, human biomechanical features, and motion intention features.

[0220] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A touchscreen anti-mistouch control method based on multi-sensor fusion, characterized in that, Includes the following steps: Acquire touch data and inertial data; Based on the original capacitance matrix in the touch data, the divergence characteristics of the gradient vector field in the touch area are calculated, and an electric field characteristic index representing the proximity mode is output. Based on the touch point coordinate data and the inertial data in the touch data, the amplitude square coherence function of the overall vibration signal of the device and the micro displacement signal of the touch point in the preset physiological frequency band is calculated, and the coupling coefficient representing the physical connection state between the touch point and the device is output. Based on the contact point coordinate data, the probability distribution of the direction state of the contact point movement trajectory is statistically analyzed and the Shannon entropy is calculated. The trajectory entropy value, which represents the orderliness of the operation intention, is output. The confidence level of false contact is calculated based on the electric field characteristic index, the coupling coefficient, and the trajectory entropy value. A dynamic suppression mask is generated based on the false touch confidence level, and touch events located within the dynamic suppression mask are intercepted. The step of outputting the electric field characteristic index of the proximity mode includes: The capacitance gradient vector of each sensing node is calculated based on the original capacitance matrix. The divergence characteristics of each sensing node are calculated based on the original capacitance matrix; Within the contact area, the electric field characteristic index is calculated by combining the divergence characteristics and the magnitude of the capacitance gradient vector. The step of outputting the coupling coefficient characterizing the physical connection state between the contact and the device includes: Under the condition that the energy of the overall vibration signal of the device is higher than the preset static noise threshold, the amplitude square coherence function is integrated within the preset physiological frequency band to obtain the coupling coefficient; The amplitude squared coherence function is calculated based on the autopower spectral density of the overall vibration signal of the device, the autopower spectral density of the micro-displacement signal of the contact point, and the cross-power spectral density between the two signals. The step of generating the dynamic suppression mask specifically includes: When the false touch confidence level is higher than the preset suppression threshold, the original capacitance profile of the touch point is identified. The original capacitor profile is enveloped using a convex hull algorithm to generate the dynamic suppression mask.

2. The touchscreen anti-mistouch control method based on multi-sensor fusion according to claim 1, characterized in that, The step of outputting the trajectory entropy value, which characterizes the ordered nature of the operational intent, includes: The movement trajectory of the contact point is encoded into a direction chain code sequence; The probability of occurrence of each directional state in the directional chain code sequence is statistically analyzed within a preset time window. The trajectory entropy value is calculated based on the probability of occurrence.

3. The touchscreen anti-mistouch control method based on multi-sensor fusion according to claim 2, characterized in that, After calculating the trajectory entropy value, the process also includes: The trajectory entropy value is converted into an intent verification coefficient using a preset S-shaped mapping function; The false touch confidence level is calculated based on the intent verification coefficient.

4. The touchscreen anti-mistouch control method based on multi-sensor fusion according to claim 1, characterized in that, The specific steps for calculating the false touch confidence level of the contact point are as follows: The false touch confidence level is obtained by weighted linear combination of the electric field characteristic index, the coupling coefficient, and an index negatively correlated with the trajectory entropy value.

5. The touchscreen anti-mistouch control method based on multi-sensor fusion according to claim 1, characterized in that, The step of intercepting touch events located within the dynamic suppression mask further includes: When the touch event is a movement event and its movement trajectory continuously moves out of the dynamic suppression mask, and the movement distance reaches a preset escape threshold, the interception of the touch event is released.

6. A touchscreen anti-mistouch control system based on multi-sensor fusion, used to execute the touchscreen anti-mistouch control method based on multi-sensor fusion as described in any one of claims 1-5, characterized in that, Includes a central processing unit, wherein the central processing unit is internally configured with: The data synchronization acquisition module is used to acquire touch data and inertial data; The electric field feature analysis module is used to calculate the divergence characteristics of the gradient vector field of the touch area based on the original capacitance matrix in the touch data, and output the electric field feature index characterizing the proximity mode. The biomechanical coupling analysis module is used to calculate the amplitude square coherence function of the overall vibration signal of the device and the micro-displacement signal of the touch point in the preset physiological frequency band based on the touch point coordinate data and the inertial data in the touch data, and output the coupling coefficient characterizing the physical connection state between the touch point and the device. The trajectory entropy verification module is used to statistically analyze the probability distribution of the direction state of the contact point movement trajectory based on the contact point coordinate data and calculate the Shannon entropy, and output the trajectory entropy value that represents the orderliness of the operation intention. The multi-dimensional decision control module is used to calculate the false touch confidence level of the touch point based on the electric field characteristic index, the coupling coefficient and the trajectory entropy value, and generate a dynamic suppression mask based on the false touch confidence level, and send instructions to the underlying driver to intercept touch events located within the dynamic suppression mask.

Images