An adaptive touch recognition method and system based on a dynamic threshold adjustment function

Abstract

The present application relates to a kind of adaptive touch identification method and system based on dynamic threshold adjustment function, by real-time reading environmental humidity data, and based on temperature and humidity data, user historical touch pressure data and the environmental compensation parameter set by dynamic threshold adjustment function dynamic adjustment touch sensitivity threshold;In touch sensitivity threshold dynamic adjustment state, the original signal of touch behavior is collected, the original signal is filtered and features are extracted, high-speed decision tree is used to filter false trigger data quickly, and then based on the user SVM decision function of online learning, the accurate identification of user touch operation behavior is carried out, which can not only intelligently distinguish the operation habits of different users, but also use historical automatic adjustment to identify sensitivity, provide personalized touch experience, actual test shows that the recognition accuracy of the scheme remains above 95%, much higher than the 80% of traditional fixed parameter algorithm, solve the problem of low accuracy of traditional algorithm in these scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of touch recognition, more particularly, to an adaptive touch recognition method and system based on a dynamic threshold adjustment function. BACKGROUND

[0002] Traditional touch systems face significant technical bottlenecks in complex environments and diverse user habits, specifically manifested in recognition accuracy decline and response delay problems, which seriously restrict the improvement of human-computer interaction experience. The following analyzes from four dimensions of environmental interference, user difference, behavior ambiguity and response delay:

[0003] 1. Environmental interference: Capacitance characteristic drift and noise superposition;

[0004] The stability of traditional capacitive touch screens in complex environments faces severe challenges. Changes in environmental temperature and humidity will cause the capacitance characteristics of the screen to drift, and the signal baseline will deviate from the normal threshold. For example, in a high temperature and high humidity environment, the capacitance value changes significantly, and the system has difficulty in accurately recognizing effective touch signals. At the same time, the noise signal introduced by static electricity interference is superimposed with the real touch signal, further reducing the signal-to-noise ratio. Environmental interference not only affects the detection accuracy of single-point touch, but also exacerbates the complexity of signal analysis in multi-point touch scenarios, making it difficult for the system to distinguish between effective touch points and noise points.

[0005] 2. User difference: Generalization dilemma of recognition adaptation;

[0006] Different users have significantly different operation habits, such as finger dryness, pressing force, operation speed, etc. The dryness of the finger directly affects the capacitive coupling effect of the touch screen: a wet finger may produce stronger signal strength, while a dry finger results in weak signal. The difference in pressing force and operation speed further increases the dynamic range of touch signals, making it difficult for fixed threshold detection algorithms to accommodate all user operation modes. For example, gentle touch may not trigger an effective response, while fast operation is easily mistaken for sliding or accidental touch. This user difference leads to obvious shortcomings in the generalization of traditional touch systems, making it difficult to meet the needs of diverse user groups.

[0007] 3. Behavior ambiguity: Difficulty in analyzing operation intent;

[0008] The behavior ambiguity of touch operation is another big technical pain point. Fast swipe and false touch may present similar characteristics in the signal level, such as similar signal fluctuations due to fine-tuning of finger position during fast swipe. Long press and dwell operation overlap in the time dimension, and the system needs to accurately determine the duration of the touch signal to distinguish the two behaviors. In addition, in the multi-point touch scene, the signals of different touch points may interfere with each other, making it difficult for the system to accurately identify the position and state of each touch point. Behavior ambiguity not only reduces the accuracy of operation recognition, but also causes misoperation in complex interaction scenarios, affecting user experience.

[0009] 4. Response delay: core bottleneck of hand followability;

[0010] The whole pipeline from original signal collection to application response is long, which is the core reason for the response lag of traditional touch systems. In the signal collection stage, the sensor needs to extract effective touch signals from the complex environment, which may take a long time due to environmental noise or user differences. In the signal processing stage, the system needs to perform filtering, noise reduction, feature extraction, etc. on the collected signals, and the computational complexity of these steps is high, especially in the multi-point touch scene, the processing time may increase significantly. In the application response stage, the system needs to convert the processed signal into specific interaction instructions and pass them to the application layer, which may introduce additional delay due to insufficient system architecture or software optimization. Response delay directly affects the hand followability of operation, making users feel that touch operation is not smooth enough, reducing the immersion of interaction.

[0011] In summary, traditional touch systems face the core pain points of reduced recognition accuracy and response lag in complex environments and diverse user habits. These problems not only restrict the application range of touch technology, but also highlight its limitations in fields such as intelligent terminals, industrial control, medical devices, etc. that require high interaction precision and real-time performance. Therefore, developing a new type of touch system that can effectively cope with environmental interference, adapt to user differences, analyze behavior ambiguity, and reduce response delay has become a key problem to be solved in the current technical field. SUMMARY

[0012] The technical problem to be solved by the present application is to provide an adaptive touch recognition method based on a dynamic threshold adjustment function that can effectively cope with environmental interference, adapt to user differences, analyze behavior ambiguity, and reduce response delay, and an adaptive touch recognition system based on a dynamic threshold adjustment function that implements the method, in order to solve the above-mentioned defects of the prior art.

[0013] The technical scheme adopted by the present application to solve its technical problem is:

[0014] A kind of adaptive touch recognition method based on dynamic threshold adjustment function is constructed, comprising the following steps:

[0015] Step one: real-time reading of environmental temperature and humidity data, and based on temperature and humidity data, user historical touch pressure data and set environmental compensation parameters, dynamically adjusting touch sensitivity threshold through dynamic threshold adjustment function;

[0016] Step two: after detecting the original touch interrupt, obtaining the original signal of the coordinates, pressure and timestamp data of the touch behavior collected under the dynamic adjustment state of the touch sensitivity threshold, filtering the original signal and extracting features;

[0017] Step three: using a lightweight cascade behavior classifier to perform two-level processing on the extracted features: the first level uses a high-speed decision tree to quickly filter false trigger data, and the second level uses an online learning-based user SVM decision function to accurately identify user touch operation behavior;

[0018] Step four: distributing the successfully identified accurate touch event.

[0019] The adaptive touch recognition method based on the dynamic threshold adjustment function, wherein the step four further comprises:

[0020] Calculating the confidence of the output result, if the result does not meet the confidence standard or whether there is a user correction operation, then go to the next step, otherwise distribute the successfully identified accurate touch event and end;

[0021] The method further comprises:

[0022] Step five: adjusting the parameters of the dynamic threshold adjustment function and the user SVM decision function according to the feedback and historical data.

[0023] The adaptive touch recognition method based on the dynamic threshold adjustment function, wherein the adjustment of the dynamic threshold adjustment function in step five comprises adjusting the environmental compensation parameter;

[0024] The adjustment of the user SVM decision function in step five comprises optimizing the weight of the lightweight cascade behavior classifier.

[0025] The adaptive touch recognition method based on the dynamic threshold adjustment function, wherein the

[0026] The dynamic threshold adjustment function comprises:

[0027] Let The touch sensitivity threshold is determined by the following formula:

[0028] ;

[0029] Wherein: The system preset basic sensitivity threshold is is the average pressure value of the last N valid touch events of the current user, used to learn the user's pressing force habit;

[0030] is the environmental compensation factor, which is a function of temperature T and humidity H:

[0031] ;

[0032] and is the reference environmental condition, and k1, k2 are compensation coefficients;

[0033] α, β, γ are weight coefficients, and α+β+γ=1, used to balance the contribution degrees of the basic value, user habit and environmental influence.

[0034] The adaptive touch recognition method based on the dynamic threshold adjustment function, wherein the

[0035] The user SVM decision function includes the formula:

[0036] ;

[0037] wherein, is the feature vector, is the kernel function, is the normal vector, is the bias term, and λ is the learning rate; is the user preference vector, which is a weighted average of all deviation vectors in the recent period, wherein the deviation vector is calculated using the formula:

[0038] ;

[0039] is the user-corrected feature vector, is the feature vector recognized by the system.

[0040] The adaptive touch recognition method based on the dynamic threshold adjustment function, wherein the filtering processing and feature extraction on the original signal include:

[0041] The adaptive Kalman filter preprocessing is used on the original signal, and the filter parameters are dynamically adjusted according to the signal variance during processing, so that the smoothing is strengthened when the signal is stable, and the rapid tracking is realized when the mutation occurs;

[0042] The signal after the adaptive Kalman filter preprocessing is extracted.

[0043] The adaptive touch recognition method based on the dynamic threshold adjustment function, wherein the feature extraction includes:

[0044] Extract time dimension features: touch duration, press / lift speed, operation frequency;

[0045] Extract spatial dimension features: touch trajectory length, curvature, direction change;

[0046] Extract pressure dimension features: pressure peak, pressure change rate, average pressure.

[0047] The adaptive touch recognition method based on the dynamic threshold adjustment function, wherein the method further comprises:

[0048] Based on the Markov chain model, the recent touch sequence is analyzed, the possible next operation of the user is predicted, and the corresponding processing resource is prepared in advance.

[0049] An adaptive touch recognition system based on a dynamic threshold adjustment function is used to implement the adaptive touch recognition method based on the dynamic threshold adjustment function as described above, wherein the system comprises a data preprocessing module, a feature extraction module, a behavior recognition module, and an adaptive adjustment module.

[0050] The adaptive adjustment module is used to read the environmental temperature and humidity data in real time, and based on the temperature and humidity data, the user's historical touch pressure data and the set environmental compensation parameters, the touch sensitivity threshold is dynamically adjusted by the dynamic threshold adjustment function.

[0051] The data preprocessing module is used to detect the original touch interrupt, and obtain the original signal containing the touch behavior coordinates, pressure and timestamp data collected under the dynamic adjustment state of the touch sensitivity threshold.

[0052] The feature extraction module is used to filter and extract features from the original signal.

[0053] The behavior recognition module is used to adopt a lightweight cascading behavior classifier to perform two-level processing on the extracted features: the first level uses a high-speed decision tree to quickly filter false trigger data, and the second level uses an online learning user SVM decision function to accurately identify the user touch operation behavior; calculate the confidence of the output result, if the result does not meet the confidence standard or whether there is a user correction operation, go to the next step, otherwise distribute the successfully identified accurate touch event and end;

[0054] The adaptive adjustment module is also used to adjust the parameters of the dynamic threshold adjustment function and the user SVM decision function according to the feedback and historical data.

[0055] The adaptive touch recognition system based on the dynamic threshold adjustment function, wherein the system further comprises a prediction optimization module.

[0056] The prediction optimization module is configured to analyze a recent touch sequence based on a Markov chain model, predict a possible next operation of the user, and prepare corresponding processing resources in advance.

[0057] The application has the advantages that the application can intelligently distinguish the operation habits of different users, automatically adjust the recognition sensitivity according to the use history, and provide personalized touch experience. Actual tests show that the recognition accuracy of the algorithm is maintained at more than 95% in various complex environments, which is much higher than the 80% of the traditional fixed parameter algorithm. The application breaks through the limitations of the traditional fixed parameter touch recognition, dynamically adjusts the recognition parameters by analyzing the touch data and environmental factors in real time, and significantly improves the accuracy and adaptability of touch recognition. The algorithm is particularly suitable for complex environments and different user habits, and solves the problem of low accuracy of traditional algorithms in these scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the present application will be further described below with reference to the drawings and embodiments. The drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor:

[0059] Figure 1 is a flow chart of an adaptive touch recognition method based on a dynamic threshold adjustment function of a preferred embodiment of the present application;

[0060] Figure 2 is a principle block diagram of an adaptive touch recognition system based on a dynamic threshold adjustment function of a preferred embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0062] The adaptive touch recognition method based on a dynamic threshold adjustment function of a preferred embodiment of the present application, as shown in Figure 1 includes the following steps:

[0063] S01: Real-time reading of environmental temperature and humidity data, and dynamic adjustment of touch sensitivity threshold based on temperature and humidity data, user historical touch pressure data and set environmental compensation parameters through a dynamic threshold adjustment function;

[0064] A specific dynamic threshold adjustment function includes:

[0065] Let be the touch sensitivity threshold, determined by the following formula:

[0066] ;

[0067] Wherein: is the system preset basic sensitivity threshold, is the average pressure value of the current user's last N valid touch events, used to learn the user's pressing force habit;

[0068] is the environmental compensation factor, the physical meaning is: the threshold offset caused by the current environment relative to the standard environment, which is a function of temperature T and humidity H, used to offset the influence of environmental changes on the touch screen capacitance characteristics, the calculation uses:

[0069] ;

[0070] and are the reference environmental temperature and humidity conditions calibrated in the laboratory, respectively, and k1, k2 are predefined compensation coefficients, respectively representing the compensation amount required for the threshold to change by one degree of temperature and one unit of humidity;

[0071] α, β, γ are weight coefficients, and α+β+γ=1, used to balance the contribution of the basic value, user habit and environmental influence, which ensures that the final calculated dynamic threshold is a reasonable weighted average of the three parts, rather than unlimited superposition;

[0072] The model combines the fixed reference, personalized history and real-time environment, and through configurable weights, finally outputs an optimal recognition threshold suitable for the current user in the current environment. This enables the system to automatically lower the threshold for light touch users, increase the threshold for heavy pressure users, and compensate in cold and dry environments, always maintaining high recognition rate.

[0073] S02: After detecting the original touch interrupt, acquire the original signal of the coordinates, pressure and timestamp data containing touch behavior collected under the touch sensitivity threshold dynamic adjustment state, filter the original signal and extract features;

[0074] The original touch refers to the actual behavior of the user's touch. Since the touch sensitivity threshold is constantly changing over time, the moment when the corresponding touch behavior occurs will have different thresholds. Based on this method, the accuracy of the initial data can be ensured when acquiring the initial data;

[0075] Wherein, the filtering processing and feature extraction on the original signal comprises:

[0076] The adaptive Kalman filter is used for preprocessing the original signal, and the filter parameters are dynamically adjusted according to the signal variance during processing, and the smoothing is strengthened when the signal is stable, and the rapid tracking is realized when the mutation occurs;

[0077] The signal after the adaptive Kalman filter preprocessing is extracted for feature extraction;

[0078] The adaptive Kalman filter refers to the use of measurement data for filtering, and the system dynamics is constantly judged by the filter itself to determine whether the system dynamics has changed, and the model parameters and noise statistical characteristics are estimated and corrected to improve the filter design and reduce the actual error of the filter. This filtering method combines system identification and filtering estimation organically.

[0079] Of course, it can be understood that other filtering methods can be used for equivalent replacement at this place, and the scheme obtained based on the replacement also belongs to the protection scope of the application.

[0080] Wherein, the feature extraction comprises:

[0081] Extracting time dimension features: touch duration, press / lift speed, operation frequency;

[0082] Extracting spatial dimension features: touch trajectory length, curvature, direction change;

[0083] Extracting pressure dimension features: pressure peak, pressure change rate, average pressure;

[0084] These feature values can accurately reflect the touch behavior from three dimensions.

[0085] S03: A lightweight cascaded behavior classifier is used to process the extracted features in two stages: the first stage uses a high-speed decision tree to quickly filter false trigger data, and the second stage uses an online learning user SVM decision function to accurately identify user touch operation behavior;

[0086] The lightweight cascaded behavior classifier is a computer vision model specially designed for efficient detection, usually based on a cascaded classifier architecture, which significantly reduces computational overhead while ensuring detection accuracy through a multi-stage screening mechanism;

[0087] Wherein, the decision tree is a decision analysis method for evaluating project risk and judging its feasibility by constructing a decision tree to obtain the probability that the expected value of net present value is greater than or equal to zero on the basis of the probability of various situations, which is a kind of graphical method for directly using probability analysis; here, the preferred first level adopts a high-speed decision tree, which can filter out more than 90% of obvious non-intentional touches (such as palm mis-touches, etc.), and is the key to improving response speed and further improving accuracy;

[0088] And the user SVM decision function includes the formula:

[0089] ;

[0090] Wherein, is a feature vector, is a kernel function, is a normal vector, is a bias term, and λ is a learning rate; is a user preference vector, which is a weighted average of all bias vectors in the recent period, wherein the bias vector is calculated using the formula:

[0091] ;

[0092] is a user-corrected feature vector, is a system-recognized feature vector;

[0093] The decision-making process of a standard support vector machine in classifying a touch feature vector (including coordinates, pressure, speed, acceleration, etc.) can be described as: calculating the position of the feature vector in the high-dimensional space after mapping through the kernel function , and the geometric relationship between the separation hyperplane (defined by the normal vector and the bias term ) obtained from the training data. A classification result (e.g. +1 for "single click" and -1 for "slide") is output by the sign function.

[0094] Description of adaptive decision function:

[0095] To make the system adapt to the user's individual habits, we introduce a "user preference vector" into the standard decision function. The improved decision-making process is as follows:

[0096] Generation of user preference vector : When the system makes a recognition error (e.g. misjudges "long press" as "single click") and is actively corrected by the user (e.g. using the "undo" function), the system will record this event. The vector is obtained by calculating the weighted average of the difference between the "correct operation feature vector" and the "previously misjudged feature vector" in a series of such correction behaviors of the user. The difference vector essentially describes the typical deviation direction between the user's specific operation habits and the system's general model.

[0097] Adjustment of the decision-making process: When making a classification decision, we do not directly input the original feature vector into the decision function, but first add a scaled user preference vector λ· to obtain a "personalized corrected" feature vector +λ· Here λ is a "learning rate" parameter that controls the strength of the user's preference on the original features.

[0098] Final decision: Substitute this corrected feature vector +λ· into the standard SVM decision function for calculation. This means that before the decision hyperplane is actually determined, the data points in the feature space have been slightly translated or adjusted in the direction of reducing similar errors according to the user's historical correction records.

[0099] The core of this improvement is to convert the user's explicit feedback (correction behavior) into implicit fine-tuning of the classification feature space. By introducing and λ, the system no longer adheres to a static, general classification boundary trained based on all user data, but allows this boundary to be flexibly adapted to the operation characteristics of individual users, thereby achieving personalized recognition capability that gets more accurate with use.

[0100] S04: Calculate the confidence of the output result. If the result does not meet the confidence standard or there is no user correction operation, go to the next step, otherwise distribute the successfully recognized accurate touch event and end;

[0101] S05: Adjust the parameters of the dynamic threshold adjustment function and the user SVM decision function according to the feedback and historical data.

[0102] Among them, the main adjustment of the dynamic threshold adjustment function is to adjust the environmental compensation parameter;

[0103] The adjustment of the user SVM decision function mainly optimizes the weight of the light-weight cascade behavior classifier;

[0104] In implementation, an independent feature configuration file can be created for each user and stored in a secure partition;

[0105] A feedback loop is established: after the user uses the "undo" operation or manually calibrates in the settings, the system takes the original feature vector of this touch and the "correct" label as new samples, incrementally updates the SVM model and the user feature mean, and realizes closed-loop optimization.

[0106] The application can intelligently distinguish the operation habits of different users, automatically adjust the recognition sensitivity according to the use history, and provide personalized touch experience. Actual tests show that the recognition accuracy of the algorithm remains above 95% in various complex environments, which is much higher than the 80% of traditional fixed parameter algorithms.

[0107] The present solution breaks through the limitations of traditional fixed parameter touch recognition, dynamically adjusts the recognition parameters through real-time analysis of touch data and environmental factors, and significantly improves the accuracy and adaptability of touch recognition. The algorithm is particularly suitable for complex environments and different user habits, solving the low accuracy problem of traditional algorithms in these scenarios.

[0108] Through actual testing, the application of the present solution has the following advantages:

[0109] 1. High robustness: in an environment of -10°C to 50°C and humidity of 20% to 80%, the click recognition accuracy is stable at above 99.5%.

[0110] 2. Personalized adaptation: the system can learn and adapt to new user habits within 100 operations, reducing the false touch rate by 70%.

[0111] 3. Low latency: from touch event occurrence to UI response, the end-to-end latency is stable within 8ms, achieving the experience of "pointing where to shoot".

[0112] 4. High energy efficiency: the algorithm itself consumes less than 5mW, which has negligible impact on the total system power consumption.

[0113] In a preferred embodiment, the method further comprises:

[0114] S06: Based on the Markov chain model, analyze the recent touch sequence, predict the user's possible next operation, and prepare the corresponding processing resources in advance.

[0115] An adaptive touch recognition system based on a dynamic threshold adjustment function is used to implement the adaptive touch recognition method based on the dynamic threshold adjustment function as described above, as shown in Figure 2 The system includes a data preprocessing module 100, a feature extraction module 101, a behavior recognition module 102, and an adaptive adjustment module 103.

[0116] The adaptive adjustment module 103 is configured to read environmental temperature and humidity data in real time, and dynamically adjust the touch sensitivity threshold based on the temperature and humidity data, historical touch pressure data of the user, and the set environmental compensation parameter through a dynamic threshold adjustment function.

[0117] The data preprocessing module 100 is configured to, after detecting the original touch interrupt, acquire original signals of coordinates, pressure, and timestamp data of touch behaviors collected in a state of dynamic adjustment of the touch sensitivity threshold.

[0118] The feature extraction module 101 is configured to filter and extract features from the original signals.

[0119] The behavior recognition module 102 is configured to perform two-stage processing on the extracted features by using a lightweight cascaded behavior classifier: the first stage uses a high-speed decision tree to quickly filter false trigger data, and the second stage uses an online learning-based user SVM decision function to accurately recognize the user touch operation behavior; the confidence of the output result is calculated, and if the result does not meet the confidence standard or whether there is a user correction operation, the next step is entered, otherwise the accurately recognized precise touch event is distributed and the process is ended.

[0120] The adaptive adjustment module 103 is further configured to adjust the parameters of the dynamic threshold adjustment function and the user SVM decision function according to the feedback and historical data.

[0121] The method of the application can intelligently distinguish the operation habits of different users, automatically adjust the recognition sensitivity according to the use history, and provide personalized touch experience. Actual tests show that the recognition accuracy of the algorithm remains above 95% in various complex environments, which is much higher than the 80% of the traditional fixed parameter algorithm.

[0122] The scheme breaks through the limitations of traditional fixed parameter touch recognition, dynamically adjusts the recognition parameters by analyzing touch data and environmental factors in real time, and significantly improves the accuracy and adaptability of touch recognition. The algorithm is particularly suitable for complex environments and different user habits, and solves the problem of low accuracy of traditional algorithms in these scenarios.

[0123] Preferably, the system further comprises a prediction optimization module 104.

[0124] The prediction optimization module 104 is configured to analyze the recent touch sequence based on a Markov chain model, predict the user's possible next operation, and prepare the corresponding processing resources in advance.

[0125] It should be understood that those skilled in the art can make improvements or changes according to the above description, and all these improvements and changes shall fall within the protection scope of the appended claims of the present application.

Claims

1. An adaptive touch recognition method based on a dynamic threshold adjustment function, characterized in that, The method comprises the following steps: Step 1: Real-time reading of environmental temperature and humidity data, and dynamic adjustment of touch sensitivity threshold based on temperature and humidity data, user historical touch pressure data and set environmental compensation parameters through a dynamic threshold adjustment function; Step 2: After detecting the original touch interrupt, obtaining the original signal of coordinate, pressure and timestamp data of touch behavior collected under the dynamic adjustment state of touch sensitivity threshold, filtering the original signal and extracting features; Step 3: Two-stage processing of the extracted features using a lightweight cascading behavior classifier: the first stage uses a high-speed decision tree to quickly filter false trigger data, and the second stage uses an online learning-based user SVM decision function to accurately identify user touch operation behavior; Step 4: Distributing the accurately identified touch event; The dynamic threshold adjustment function comprises: Set The touch sensitivity threshold is determined by the following equation: ； wherein: a basic sensitivity threshold preset for the system, an average pressure value of the last N valid touch events of the current user, used to learn the user's pressing force habit; The environmental compensation factor, is a function of temperature T and humidity H: ； and are reference ambient conditions, and k1, k2 are compensation factors. α, β and γ are weight coefficients, and α+β+γ=1, used to balance the contribution of the base value, user habits and environmental influence; The user SVM decision function comprises the formula: ； wherein, is a feature vector, is a kernel function, is a normal vector, is a bias term, and λ is a learning rate; is a user preference vector, and is a weighted average of all bias vectors in the recent past, wherein bias vector is calculated using the formula: ； a feature vector corrected for the user, a feature vector identified for the system.

2. The adaptive touch recognition method based on a dynamic threshold adjustment function according to claim 1, characterized in that, The step 4 further comprises: Calculating the confidence of the output result, if the result does not meet the confidence standard or whether there is a user correction operation, then proceed to the next step, otherwise distribute the accurately identified touch event and end; The method further comprises: Step 5: Adjusting the parameters of the dynamic threshold adjustment function and the user SVM decision function according to feedback and historical data. 3.The adaptive touch recognition method based on dynamic threshold adjustment function according to claim 2, characterized in that, The adjustment of the dynamic threshold adjustment function in step 5 comprises adjusting the environmental compensation parameters; The adjustment of the user SVM decision function in step 5 comprises optimizing the weight of the lightweight cascading behavior classifier.

4. The adaptive touch recognition method based on a dynamic threshold adjustment function according to claim 1, characterized in that, The filtering and feature extraction of the original signal comprises: Adaptive Kalman filter preprocessing is performed on the original signal, and the filter parameters are dynamically adjusted according to the signal variance during processing, with strong smoothing when the signal is stable and fast tracking when there is a sudden change; Feature extraction is performed on the signal after adaptive Kalman filter preprocessing.

5. The adaptive touch recognition method based on dynamic threshold adjustment function according to claim 4, characterized in that, The feature extraction comprises: Extracting time dimension features: touch duration, press / lift speed, operation frequency; Extracting spatial dimension features: touch trajectory length, curvature, direction change; Extracting pressure dimension features: pressure peak value, pressure change rate, average pressure.

6. The adaptive touch recognition method based on dynamic threshold adjustment function according to claim 1, wherein, The method further comprises: Based on the Markov chain model, the recent touch sequence is analyzed, the user's possible next operation is predicted, and the corresponding processing resources are prepared in advance.

7. An adaptive touch recognition system based on a dynamic threshold adjustment function for implementing the adaptive touch recognition method based on a dynamic threshold adjustment function according to any one of claims 1 to 6, characterized in that, The system comprises a data preprocessing module, a feature extraction module, a behavior recognition module and an adaptive adjustment module; The adaptive adjustment module is used to read environmental temperature and humidity data in real time, and dynamically adjust the touch sensitivity threshold based on temperature and humidity data, user historical touch pressure data and set environmental compensation parameters through a dynamic threshold adjustment function; The data preprocessing module is used to obtain the original signal of coordinate, pressure and timestamp data of touch behavior collected under the dynamic adjustment state of touch sensitivity threshold after detecting the original touch interrupt; The feature extraction module is used to filter and extract features from the original signal; The feature extraction module is used to filter and extract features from the original signal; The behavior recognition module is configured to perform two-stage processing on the extracted features by using a light-weight cascade behavior classifier: in the first stage, a high-speed decision tree is used to quickly filter out false trigger data; and in the second stage, a user SVM decision function based on online learning is used to accurately recognize the user touch operation behavior; the confidence of the output result is calculated, and if the result does not meet the confidence standard or whether there is a user correction operation, the next step is entered, otherwise the accurately recognized precise touch event is distributed and the process is ended. The adaptive adjustment module is further configured to adjust the parameters of the dynamic threshold adjustment function and the user SVM decision function according to the feedback and historical data.

8. The adaptive touch recognition system based on dynamic threshold adjustment function according to claim 7, wherein, The system further comprises a prediction optimization module. The prediction optimization module is configured to analyze the recent touch sequence based on a Markov chain model, predict the possible next operation of the user, and prepare the corresponding processing resources in advance.

Images