An infrared sensing application method

Abstract

This invention discloses an infrared sensing application method. It relates to the field of infrared sensing technology and solves the technical problems of signal drift and failure under complex working conditions, as well as serious accidental touch issues in dynamic environments. This invention dynamically adjusts the pressure sensing threshold and capacitance judgment threshold based on real-time data such as the root mean square value of the accelerometer and the rate of change of suspension travel. When a bumpy road section is detected, the trigger threshold is automatically raised, fundamentally suppressing accidental touches caused by vehicle vibration. This solves the problem that traditional fixed thresholds cannot adapt to dynamic road conditions. By calculating the dynamic correlation coefficient between the finger movement speed vector and the vehicle bump acceleration vector, it can effectively distinguish between active touch and passive vibration, accurately shielding accidental touches caused by bumps. Combined with touch area analysis and a ghost hand signal recognition model based on CNN / GBDT, it completely solves the interference problems of large-area foreign objects, small noise, and irregular ghost signals.

Description

Technical Field

[0001] This invention relates to the field of infrared sensing technology, specifically to an infrared sensing application method. Background Technology

[0002] With the rapid development of automotive intelligence and connectivity, in-vehicle touchscreen displays have become the core interactive interface of smart cockpits. Modern cars generally adopt large-size, high-resolution, and even curved capacitive touchscreens to provide a smooth visual experience and rich functional interactions.

[0003] Traditional capacitive touchscreens typically use fixed sensitivity thresholds and static filtering algorithms. When a vehicle is traveling on bumpy roads, the vibrations of the vehicle body are transmitted to the screen, causing unexpected relative displacement or pressure fluctuations between the finger and the screen. These fluctuations can easily be misinterpreted by the system as swipes or clicks. In addition, the inertial swaying of the driver's body during emergency maneuvers or sharp turns often causes the arm or clothing to accidentally touch the edge of the screen, resulting in the "ghost hand" effect or misoperation, posing a significant safety hazard.

[0004] The in-vehicle environment is characterized by high temperature, high humidity, and strong electromagnetic interference. Existing technologies for handling baseline drift mostly use a simple fixed-window averaging method, which cannot adapt to capacitance drift caused by temperature changes after long-term operation. In rainy or high-humidity environments, the water film on the surface of fingers can cause abnormal increases in capacitance signals, and existing single capacitance detection mechanisms cannot distinguish between real touch and water droplet interference, leading to touch failure or false triggering. At the same time, electromagnetic interference (EMI) generated by in-vehicle electronic devices can also reduce the signal-to-noise ratio and affect touch accuracy. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an infrared sensing application method that solves the serious problems of signal drift and failure under complex working conditions and accidental touch in dynamic environments.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an infrared sensing application method, which specifically includes the following steps: Step 1: Collect the user's original touch operation signal and perform baseline drift correction and spatial domain filtering to obtain preprocessed touch frame data. At the same time, obtain the vehicle's current driving status parameters and environmental status parameters, and dynamically adjust the threshold parameters for touch judgment. Step 2: Generate candidate touch points based on the adjusted touch judgment threshold and extract primary touch feature parameters. Perform adaptive Kalman filter smoothing on the position coordinate sequence of the same touch point in multiple consecutive frames to obtain a robust touch movement trajectory. Then, extract the trajectory feature parameters and input them into the pre-trained lightweight gesture recognition model to output the gesture classification result. Step 3: Align the extracted touch movement trajectory feature parameters with the vehicle accelerometer data in the time domain, calculate the dynamic correlation coefficient between the finger movement speed vector and the vehicle bump acceleration vector, and identify accidental touches. At the same time, based on the touch area parameter and the ghost hand signal recognition model, determine and block large-area foreign object accidental touches, small foreign object accidental touches and ghost hand signals respectively. Step 4: After determining that the current touch operation is not a mis-touch event, construct a driving risk level assessment model based on driving status parameters, determine the functional area that is currently allowed to be touched, generate the corresponding control command, and dynamically generate a tactile feedback signal based on the touch event type while executing the control command.

[0007] As a further aspect of the present invention, the method for obtaining the preprocessed touch frame data is as follows: Subtract the dynamically updated background capacitance reference value from the capacitance change of each node in the current frame. The background capacitance reference value is calculated in real time based on the non-touch state signals of the past N frames using an exponentially weighted moving average method, and its update formula is as follows: ,in This is the current background capacitance reference value. This is the capacitance measurement value in the previous frame when the device was not in touch mode. As a smoothing factor, This is the baseline value for the background capacitance of the previous frame; Spatial domain filtering is performed on the corrected touch frame data, and a 3×3 Gaussian filter kernel is used to perform convolution operation on the data of each node to obtain preprocessed touch frame data.

[0008] As a further aspect of the present invention, the method for dynamically adjusting the threshold parameter for touch detection is as follows: When it is determined that the vehicle is currently traveling on a bumpy road based on the root mean square value of the accelerometer data or the fluctuation range of the suspension travel rate, the pressure sensing threshold on which the touch judgment depends is increased by a first preset multiple. When the humidity sensor data inside the vehicle determines that the humidity on the surface of the user's finger exceeds the preset humidity threshold, the threshold for determining the change in capacitance is increased by a second preset multiple. When the electromagnetic interference intensity detection circuit determines that there is a strong interference source exceeding the preset intensity threshold, the common-mode noise suppression filter is automatically activated.

[0009] As a further aspect of the present invention, the primary touch feature parameters include: Position coordinates: the centroid coordinates of the connected region ; Touch area: The total number of valid touch nodes within the connected area multiplied by the physical area corresponding to a single node; Pressure estimate: The sum of capacitance changes of all nodes in the connected region, mapped to a dimensionless pressure value by a calibration function; Touch duration: The cumulative number of frames from the first touch to the current frame multiplied by the frame interval.

[0010] As a further aspect of the present invention, the method for performing adaptive Kalman filtering smoothing on the position coordinate sequence of the same touch point in multiple consecutive frames is as follows: Establish a discrete Kalman filter model, with the state vector as follows: The observation vector is Process noise covariance matrix and measurement noise covariance matrix Based on the current vehicle driving status parameters, the system dynamically adjusts the measurement noise covariance matrix. When a bumpy road section is detected, the measurement noise covariance matrix is ​​increased. When a high-speed, straight road section is detected, the measurement noise covariance matrix is ​​reduced. For each new frame, a Kalman prediction and update step is performed, outputting the smoothed touch point position coordinates and velocity estimate. .

[0011] As a further aspect of the present invention, the specific process of outputting the gesture classification result is as follows: Trajectory feature parameters include: Touch start position and touch end position ; instantaneous speed average speed Instantaneous acceleration ; Trajectory curvature: the curvature at each point And extract the maximum and average curvature, where This represents the Y-axis acceleration component of the object at point k. This represents the X-axis acceleration component of the object at point k; Rate of change of trajectory direction: direction angle , cumulative sum of directional changes ; trajectory length Straightness ; For multi-touch, additional multi-point feature parameters are extracted, including the initial distance, final distance, and relative distance change rate between two touch points; the rotation angle change of the line connecting the two touch points; and the change of the convex hull area formed by the multiple points. The extracted trajectory feature parameters are input into a pre-trained lightweight gesture recognition model to classify the current touch operation. The classification results include: single-point click, double-click, long press, swipe, drag, pinch zoom, and can be extended to advanced gestures such as rotation and multi-finger flick.

[0012] As a further solution of the present invention, the specific method for calculating the dynamic correlation coefficient between the finger movement speed vector and the vehicle bump acceleration vector and identifying mis-touch situations is as follows: The calculation formula for the correlation coefficient is , where is the projection component of the finger movement speed vector of the i-th frame in the vehicle reference coordinate system, is the corresponding vehicle bump acceleration vector, and the correlation coefficient determination threshold is dynamically adjusted based on the driving state parameters of the current vehicle , and the calculation formula is , where is the reference correlation coefficient threshold, k is the adjustment coefficient, is the standard fluctuation amplitude of the suspension stroke change rate; If the calculated dynamic correlation coefficient exceeds the dynamically adjusted correlation coefficient determination threshold , it is determined that the current touch event is an unconscious swipe caused by road bumps, and the touch event is marked as a bump mis-touch and blocked, and does not enter the subsequent gesture recognition and control instruction generation process.

[0013] As a further solution of the present invention, the method for determining and shielding large-area foreign object mis-touch, small foreign object mis-touch and ghost hand signals based on the touch area parameter and the ghost hand signal recognition model is as follows: Construct a ghost hand signal recognition model based on a convolutional neural network or a gradient boosting decision tree. The input of the recognition model is the original touch frame data of continuous N frames, and the output is the probability value that the current touch signal is determined to be a ghost hand signal; The ghost hand signal recognition model extracts discriminant features including the spatial distribution feature of the touch signal, the time-domain stability feature of the touch signal, and the frequency component feature of the touch signal; If the probability of the ghost hand signal output by the recognition model exceeds the preset ghost hand determination threshold, it is determined that the current touch event is a ghost hand signal, and the corresponding touch event is blocked.

[0014] As a further solution of the present invention, the method for constructing a driving risk level assessment model based on driving state parameters is as follows: When the vehicle speed V ≤ Vlow and the gear is in P gear or N gear, it is determined as a stationary / parking scenario, and the risk level is level 0. When Vlow < V ≤ Vmid and the lateral acceleration < When it is in this condition, it is determined as a low-speed stable driving scenario, and the risk level is level 1. When Vmid < V ≤ Vhigh and the lateral acceleration < When it is in this condition, it is determined as a medium-speed stable driving scenario, and the risk level is level 2. When V > Vhigh, or the lateral acceleration ≥ , or the longitudinal acceleration ≥ When it is in this condition, it is determined as a high-speed or dynamic driving scenario, and the risk level is level 3.

[0015] As a further solution of the present invention, the method for determining the currently allowed touch function area is as follows: All interaction functions of the in-vehicle touch screen are pre-divided into multiple function categories, and each function category is associated with a minimum risk level threshold for allowed operations. The core driving safety function corresponds to a minimum risk level threshold of level 0, the high-frequency driving assistance function corresponds to a minimum risk level threshold of level 1, navigation and traffic information, audio correspond to a minimum risk level threshold of level 2, video, vehicle settings and personalization correspond to a minimum risk threshold of level 0; When the driving risk level increases, the touch area corresponding to the function category with a risk level lower than the minimum risk level threshold of this function category is prohibited.

[0016] The present invention provides an infrared sensing application method. Compared with the prior art, it has the following beneficial effects: By adopting the exponentially weighted moving average method to dynamically update the background capacitance reference value and introducing a common-mode noise suppression filter, the present invention effectively overcomes high-temperature drift, high-humidity water film interference and strong electromagnetic interference, ensures high-precision touch response even in extreme environments, and uses adaptive Kalman filtering to dynamically adjust the measurement noise covariance matrix according to road conditions, increases the noise tolerance to smooth the jitter trajectory on bumpy roads, and reduces the noise on straight roads to improve accuracy, achieving the best smoothing effect of trajectory data.

[0017] According to real-time data such as the root mean square value of the accelerometer and the change rate of the suspension stroke, the present invention dynamically adjusts the pressure sensing threshold and the capacitance determination threshold, automatically increases the trigger threshold when detecting a bumpy road section, suppresses false touches caused by vehicle body vibration from the root, solves the problem that traditional fixed thresholds cannot adapt to dynamic road conditions, and can effectively distinguish active touches from passive vibrations by calculating the dynamic correlation coefficient between the finger movement speed vector and the vehicle bump acceleration vector, accurately shielding bumpy false touches; combined with touch area analysis and a ghost hand signal recognition model based on CNN / GBDT, it completely solves the interference problems of large-area foreign objects, small noises and irregular ghost signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1This is a flowchart of the infrared sensing application method of the present invention. Detailed Implementation

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

[0020] First Embodiment Please see Figure 1 This application provides an infrared sensing application method, which specifically includes the following steps: Step 1: Collect the user's original touch operation signal and perform baseline drift correction and spatial domain filtering to obtain preprocessed touch frame data. At the same time, obtain the vehicle's current driving status parameters and environmental status parameters, and dynamically adjust the threshold parameters for touch judgment. Step 2: Generate candidate touch points based on the adjusted touch judgment threshold and extract primary touch feature parameters. Perform adaptive Kalman filter smoothing on the position coordinate sequence of the same touch point in multiple consecutive frames to obtain a robust touch movement trajectory. Then, extract the trajectory feature parameters and input them into the pre-trained lightweight gesture recognition model to output the gesture classification result. Step 3: Align the extracted touch movement trajectory feature parameters with the vehicle accelerometer data in the time domain, calculate the dynamic correlation coefficient between the finger movement speed vector and the vehicle bump acceleration vector, and identify accidental touches. At the same time, based on the touch area parameter and the ghost hand signal recognition model, determine and block large-area foreign object accidental touches, small foreign object accidental touches and ghost hand signals respectively. Step 4: After determining that the current touch operation is not a mis-touch event, construct a driving risk level assessment model based on driving status parameters, determine the functional area that is currently allowed to be touched, generate the corresponding control command, and dynamically generate a tactile feedback signal based on the touch event type while executing the control command.

[0021] Second Embodiment As a second embodiment of the present invention, it is implemented based on the first embodiment, and the difference from the first embodiment is as follows: By using a capacitive sensing array embedded in the touch detection layer of the touch screen, the original signals of the user's touch operation are collected in real time at a scanning frequency of not less than 120Hz. The capacitive sensing array is a matrix electrode layout, including multiple driving electrodes arranged at equal intervals along a first direction and multiple sensing electrodes arranged at equal intervals along a second direction. The driving electrodes and sensing electrodes form mutual capacitance nodes at the intersection. By scanning each mutual capacitance node row by row, the original touch frame data containing the capacitance change of each node is obtained. Baseline drift correction is performed on the original touch frame data. Specifically, the capacitance change of each node in the current frame is subtracted from the dynamically updated background capacitance reference value. The background capacitance reference value is calculated in real time based on the non-touch state signals of the past N frames using an exponentially weighted moving average method, and its update formula is as follows: ,in This is the current background capacitance reference value. This is the capacitance measurement value in the previous frame when the device was not in touch mode. This is a smoothing factor, with a value range of 0.05-0.2. This is the baseline value for the background capacitance of the previous frame; Spatial domain filtering is performed on the corrected touch frame data. A Gaussian filter kernel of size 3×3 is used to perform convolution operation on the data of each node to remove isolated noise points and random spike interference, so as to obtain preprocessed touch frame data. Simultaneously, the vehicle's current driving status parameters and environmental status parameters are acquired. Driving status parameters include vehicle speed, steering wheel angle, three-axis accelerometer data, and suspension travel rate of change. Environmental status parameters include in-vehicle temperature, ambient light intensity, in-vehicle relative humidity, and electromagnetic interference intensity. Based on these driving and environmental status parameters, the threshold parameters for touch detection are dynamically adjusted. The specific method of dynamic adjustment is as follows: When it is determined that the vehicle is currently traveling on a bumpy road based on the root mean square value of the accelerometer data or the fluctuation range of the suspension travel rate, the pressure sensing threshold on which the touch judgment depends is increased by a first preset multiple. The specific value of the first preset multiple is set by the operator. When the humidity sensor data inside the vehicle determines that the humidity on the surface of the user's finger exceeds the preset humidity threshold, the threshold for determining the change in capacitance is increased by a second preset factor, and the specific value of the second preset factor is set by the operator. When the electromagnetic interference intensity detection circuit determines that there is a strong interference source that exceeds the preset intensity threshold, the common-mode noise suppression filter is automatically activated. The filter adopts a differential signal cancellation structure or an adaptive notch filter algorithm.

[0022] The specific implementation process in step two is as follows: Based on the adjusted touch detection threshold, the preprocessed touch frame data is binarized. Specifically, the capacitance change of each mutual capacitance node is compared with the current dynamic threshold. Nodes with capacitance changes greater than or equal to the threshold are considered valid touch nodes and assigned a value of 1; otherwise, they are assigned a value of 0, forming a binary touch image. Connectivity analysis is then performed on the binary touch image, using 8-neighbor or 4-neighbor connectivity rules to label each connected region. Each connected region corresponds to a candidate touch point. For each candidate touch point, the following primary touch feature parameters are extracted: Position coordinates: the centroid coordinates of the connected region The calculation formula is: ,in Weights for changes in node capacitance; Touch area: The total number of valid touch nodes within the connected area multiplied by the physical area corresponding to a single node; Pressure estimate: The sum of capacitance changes of all nodes in the connected region, mapped to a dimensionless pressure value by a calibration function; Touch duration: The cumulative number of frames from the first touch to the current frame multiplied by the frame interval time; If multiple candidate touch points are detected, each is assigned a temporary ID and its state is maintained. Simultaneously, the position coordinate sequence of the same touch point in multiple consecutive frames is smoothed using adaptive Kalman filtering to obtain a robust touch movement trajectory. The specific acquisition method is as follows: Establish a discrete Kalman filter model, with the state vector as follows: The observation vector is Process noise covariance matrix and measurement noise covariance matrix Based on the current vehicle driving status parameters, the system dynamically adjusts the measurement noise covariance matrix. When a bumpy road section is detected, the measurement noise covariance matrix is ​​increased. When a high-speed, straight road section is detected, the measurement noise covariance matrix is ​​reduced. For each new frame, a Kalman prediction and update step is performed, outputting the smoothed touch point position coordinates and velocity estimate. ; Based on the smoothed touch movement trajectory, trajectory feature parameters for gesture recognition are extracted, and these trajectory feature parameters include: Touch start position and touch end position ; instantaneous speed average speed Instantaneous acceleration ; Trajectory curvature: the curvature at each point And extract the maximum and average curvature, where This represents the Y-axis acceleration component of the object at point k. This represents the X-axis acceleration component of the object at point k; Rate of change of trajectory direction: direction angle , cumulative sum of directional changes ; trajectory length Straightness ; For multi-touch, additional multi-point feature parameters are extracted, including the initial distance, final distance, and relative distance change rate between two touch points; the rotation angle change of the line connecting the two touch points; and the change of the convex hull area formed by the multiple points. The extracted trajectory feature parameters are input into a pre-trained lightweight gesture recognition model to classify the current touch operation. The lightweight gesture recognition model adopts a temporal convolutional network or gated recurrent unit architecture, and the model parameter size is controlled within 500KB to meet the requirements of automotive-grade real-time inference. The input of the model is a fixed-length feature sequence, and the output is the probability distribution of various gestures. The classification results include at least: single click, double click, long press (touch duration exceeds 500ms and displacement is less than the threshold), swipe, drag, pinch to zoom, and can be extended to advanced gestures such as rotation and multi-finger swipe.

[0023] Third Embodiment As a third embodiment of the present invention, it is implemented based on the second embodiment, and the difference from the second embodiment is as follows: The specific implementation process of step three is as follows: the extracted touch movement trajectory feature parameters are aligned in the time domain with the vehicle accelerometer data synchronously obtained from the vehicle CAN bus. The time domain alignment includes linear interpolation synchronization based on the timestamp of the touch event and the sampling timestamp of the acceleration data to eliminate the time offset error between the two types of sensors. On the aligned data frame, calculate the dynamic correlation coefficient between the finger movement velocity vector and the vehicle bump acceleration vector. And the formula for calculating the correlation coefficient is: ,in Let be the projection component of the finger movement velocity vector in the vehicle reference coordinate system in the i-th frame. For the corresponding vehicle bump acceleration vector, the correlation coefficient judgment threshold is dynamically adjusted based on the current vehicle driving state parameters. And the calculation formula is ,in The threshold value is the baseline correlation coefficient, and k is the adjustment coefficient. This represents the standard fluctuation range of the suspension travel change rate; If the calculated dynamic correlation coefficient Exceed the correlation coefficient determination threshold after dynamic adjustment , it is determined that the current touch event is an unconscious slide caused by road bumps, mark the touch event as a bump mis-touch and shield it, and do not enter the subsequent gesture recognition and control instruction generation process; Extract the touch area parameters of each touch point in the preliminary touch event. The touch area parameter is the total number of touch-effective nodes in the connected area multiplied by the physical area corresponding to a single node. Compare the touch area parameter with the preset upper and lower threshold values of the finger touch area. If the touch area > the upper threshold, it is determined that the current touch is not from a finger, marked as a large-area foreign object mis-touch and shielded; if the touch area < the lower threshold, it is determined as a small foreign object or electrical noise interference, marked as a small foreign object mis-touch and shielded; Apply the ghost hand signal recognition model to the continuous multi-frame touch data for pseudo-touch determination. Construct a ghost hand signal recognition model based on a convolutional neural network or a gradient boosting decision tree. The input of the recognition model is the raw touch frame data of continuous N frames, and the output is the probability value that the current touch signal is determined to be a ghost hand signal. And the ghost hand signal recognition model extracts at least one of the following discriminant features: (a) Spatial distribution feature of the touch signal - The capacitance change amount of a real finger touch on the mutual capacitance nodes usually shows a Gaussian distribution that decreases from the touch point to the surrounding area, while the capacitance change amount distribution of the ghost hand signal shows an irregular flat distribution or a local mutation distribution; (b) Temporal stability feature of the touch signal - The capacitance change amount of a real finger touch changes smoothly between consecutive frames, while the inter-frame fluctuation amplitude of the ghost hand signal is significantly larger; (c) Frequency component feature of the touch signal - The main frequency components of a real finger touch are concentrated in the low-frequency band, while there are specific noise frequency components in the ghost hand signal; If the probability of the ghost hand signal output by the recognition model exceeds the preset ghost hand determination threshold, it is determined that the current touch event is a ghost hand signal, and the corresponding touch event is shielded.

[0024] The specific implementation process of Step Four is as follows: After determining that the current touch operation does not belong to a mis-touch event, determine the currently allowed touch function area according to the driving state parameters, and generate corresponding control instructions. The determination method of the currently allowed touch function area is as follows: Based on the driving state parameters, construct a driving risk level assessment model, and output the driving risk level at the current moment. When the vehicle speed V ≤ Vlow and the gear is in P gear or N gear, it is determined as a stationary / parking scenario, and the risk level is level 0. When Vlow < V ≤ Vmid and the lateral acceleration < When it is in this situation, it is determined as a low-speed stable driving scenario, and the risk level is level 1. When Vmid < V ≤ Vhigh and the lateral acceleration < When it is in this situation, it is determined as a medium-speed stable driving scenario, and the risk level is level 2. When V > Vhigh, or the lateral acceleration ≥ , or the longitudinal acceleration ≥ When it is in this situation, it is determined as a high-speed or dynamic driving scenario, and the risk level is level 3; All interaction functions of the in-vehicle touch screen are pre-divided into multiple function categories, and each function category is associated with a minimum risk level threshold for allowing operations . The minimum risk level threshold corresponding to the core driving safety function is level 0, the minimum risk level threshold corresponding to the high-frequency driving assistance function is level 1, the minimum risk level thresholds corresponding to navigation and traffic information, and audio are level 2, and the minimum risk thresholds corresponding to video, vehicle settings and personalization are level 0; When the driving risk level increases, prohibit > the touch areas corresponding to the function categories, and display visual prompts in the corresponding interface areas. For the function areas that switch from the allowed state to the prohibited state, send a characteristic vibration prompt that the function is unavailable to the user through the tactile feedback module, While executing the control instruction on the touch screen, generate the corresponding tactile feedback signal, and provide tactile feedback to the user through the linear vibration motor arranged on the back of the display screen or the piezoelectric ceramic actuator arranged on the frame of the display screen. The intensity, frequency and duration of the tactile feedback signal are dynamically adjusted according to the type of touch event. The click operation corresponds to a short single-pulse vibration, the long-press operation corresponds to continuous vibration, and the swipe operation corresponds to a vibration sequence with a gradually changing frequency.

[0025] The Fourth Embodiment As the fourth embodiment of the present invention, the focus is on combining the implementation processes of all the above embodiments.

[0026] Some of the data in the above formula are numerically calculated after removing their dimensions, and the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0027] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for applying infrared sensing, characterized in that, The method specifically includes the following steps: Step 1: Collect the user's original touch operation signal and perform baseline drift correction and spatial domain filtering to obtain preprocessed touch frame data. At the same time, obtain the vehicle's current driving status parameters and environmental status parameters, and dynamically adjust the threshold parameters for touch judgment. Step 2: Generate candidate touch points based on the adjusted touch judgment threshold and extract primary touch feature parameters. Perform adaptive Kalman filter smoothing on the position coordinate sequence of the same touch point in multiple consecutive frames to obtain a robust touch movement trajectory. Then, extract the trajectory feature parameters and input them into the pre-trained lightweight gesture recognition model to output the gesture classification result. Step 3: Align the extracted touch movement trajectory feature parameters with the vehicle accelerometer data in the time domain, calculate the dynamic correlation coefficient between the finger movement speed vector and the vehicle bump acceleration vector, and identify accidental touches. At the same time, based on the touch area parameter and the ghost hand signal recognition model, determine and block large-area foreign object accidental touches, small foreign object accidental touches and ghost hand signals respectively. Step 4: After determining that the current touch operation is not a mis-touch event, construct a driving risk level assessment model based on driving status parameters, determine the functional area that is currently allowed to be touched, generate the corresponding control command, and dynamically generate a tactile feedback signal based on the touch event type while executing the control command.

2. The infrared sensing application method according to claim 1, characterized in that, The method for obtaining the preprocessed touch frame data is as follows: Subtract the dynamically updated background capacitance reference value from the capacitance change of each node in the current frame. The background capacitance reference value is calculated in real time based on the non-touch state signals of the past N frames using an exponentially weighted moving average method, and its update formula is as follows: ,in This is the current background capacitance reference value. This is the capacitance measurement value in the previous frame when the device was not in touch mode. As a smoothing factor, This is the baseline value for the background capacitance of the previous frame; Spatial domain filtering is performed on the corrected touch frame data, and a 3×3 Gaussian filter kernel is used to perform convolution operation on the data of each node to obtain preprocessed touch frame data.

3. The infrared sensing application method according to claim 1, characterized in that, The method for dynamically adjusting the threshold parameter for touch detection is as follows: When it is determined that the vehicle is currently traveling on a bumpy road based on the root mean square value of the accelerometer data or the fluctuation range of the suspension travel rate, the pressure sensing threshold on which the touch judgment depends is increased by a first preset multiple. When the humidity sensor data inside the vehicle determines that the humidity on the surface of the user's finger exceeds the preset humidity threshold, the threshold for determining the change in capacitance is increased by a second preset multiple. When the electromagnetic interference intensity detection circuit determines that there is a strong interference source exceeding the preset intensity threshold, the common-mode noise suppression filter is automatically activated.

4. The infrared sensing application method according to claim 1, characterized in that, The primary touch feature parameters include: Position coordinates: the centroid coordinates of the connected region ; Touch area: The total number of valid touch nodes within the connected area multiplied by the physical area corresponding to a single node; Pressure estimate: The sum of capacitance changes of all nodes in the connected region, mapped to a dimensionless pressure value by a calibration function; Touch duration: The cumulative number of frames from the first touch to the current frame multiplied by the frame interval.

5. The infrared sensing application method according to claim 1, characterized in that, The method for adaptive Kalman filtering smoothing of the position coordinate sequence of the same touch point in multiple consecutive frames is as follows: Establish a discrete Kalman filter model, with the state vector as follows: The observation vector is Process noise covariance matrix and measurement noise covariance matrix Based on the current vehicle driving status parameters, the system dynamically adjusts the measurement noise covariance matrix. When a bumpy road section is detected, the measurement noise covariance matrix is ​​increased. When a high-speed, straight road section is detected, the measurement noise covariance matrix is ​​reduced. For each new frame, a Kalman prediction and update step is performed, outputting the smoothed touch point position coordinates and velocity estimate. .

6. The infrared sensing application method according to claim 1, characterized in that, The specific process for outputting the gesture classification result is as follows: Trajectory feature parameters include: Touch start position and touch end position ; instantaneous speed average speed Instantaneous acceleration ; Trajectory curvature: the curvature at each point And extract the maximum and average curvature, where This represents the Y-axis acceleration component of the object at point k. This represents the X-axis acceleration component of the object at point k; Rate of change of trajectory direction: direction angle , cumulative sum of directional changes ; trajectory length Straightness ; For multi-touch, additional multi-point feature parameters are extracted, including the initial distance, final distance, and relative distance change rate between two touch points; the rotation angle change of the line connecting the two touch points; and the change of the convex hull area formed by the multiple points. The extracted trajectory feature parameters are input into a pre-trained lightweight gesture recognition model to classify the current touch operation. The classification results include: single-point click, double-click, long press, swipe, drag, pinch to zoom, and can be extended to advanced gestures such as rotation and multi-finger swipe.

7. The infrared sensing application method according to claim 1, characterized in that, The specific method for calculating the dynamic correlation coefficient between the finger movement speed vector and the vehicle bump acceleration vector and identifying accidental touches is as follows: The formula for calculating the correlation coefficient is: ,in Let be the projection component of the finger movement velocity vector in the vehicle reference coordinate system in the i-th frame. For the corresponding vehicle bump acceleration vector, the correlation coefficient judgment threshold is dynamically adjusted based on the current vehicle driving state parameters. And the calculation formula is ,in The threshold value is the baseline correlation coefficient, and k is the adjustment coefficient. This represents the standard fluctuation range of the suspension travel change rate; If the calculated dynamic correlation coefficient The threshold for determining the correlation coefficient exceeds the dynamically adjusted value. If the current touch event is determined to be an unintentional swipe caused by road bumps, the touch event is marked as a bumpy accidental touch and blocked, and will not proceed to the subsequent gesture recognition and control command generation process.

8. The infrared sensing application method according to claim 1, characterized in that, The method for determining and blocking accidental touches by large-area foreign objects, accidental touches by small foreign objects, and "ghost hand" signals based on touch area parameters and ghost hand signal recognition models is as follows: Construct a ghost hand signal recognition model based on convolutional neural network or gradient boosting decision tree. The input of the recognition model is the raw touch frame data of N consecutive frames, and the output is the probability value of the current touch signal being judged as a ghost hand signal. The ghost hand signal recognition model extracts discriminative features including the spatial distribution features of touch signals, the temporal stability features of touch signals, and the frequency component features of touch signals; If the probability of the ghost hand signal output by the recognition model exceeds the preset ghost hand judgment threshold, the current touch event is determined to be a ghost hand signal, and the corresponding touch event is blocked.

9. The infrared sensing application method according to claim 1, characterized in that, The method for constructing a driving risk level assessment model based on driving state parameters is as follows: When the vehicle speed V ≤ Vlow and the gear is in P or N, it is determined as a stationary / parking scenario, and the risk level is level 0. When Vlow < V ≤ Vmid and the lateral acceleration < is such that, it is determined as a low-speed stable driving scenario, and the risk level is level 1. When Vmid < V ≤ Vhigh and the lateral acceleration < is such that, it is determined as a medium-speed stable driving scenario, and the risk level is level 2. When V > Vhigh, or the lateral acceleration ≥ , or the longitudinal acceleration ≥ is such that, it is determined as a high-speed or dynamic driving scenario, and the risk level is level 3.

10. The infrared sensing application method according to claim 1, characterized in that, The method for determining the currently allowed touch area is as follows: All interactive functions of the in-vehicle touch screen are pre-divided into multiple function categories. Each function category is associated with a minimum risk level threshold that allows operation. The minimum risk level threshold for core driving safety functions is 0, the minimum risk level threshold for high-frequency driving assistance functions is 1, the minimum risk level threshold for navigation and traffic information and audio is 2, and the minimum risk level threshold for video, vehicle settings and personalization is 0. When the driving risk level increases, the touch area corresponding to the function category whose risk level is lower than the minimum risk level threshold of that function category is prohibited.

Images