A global air-touch adaptive display conference integrated machine

Abstract

The application discloses a full-domain air-touch self-adaptive display conference integrated machine, and relates to the technical field of intelligent instruments.The technical problem of insufficient gesture recognition accuracy and stability and easy misjudgment is solved.The application realizes accurate segmentation of continuous gestures by time stamp marking and time sequence deconstruction of continuous gesture action flow, combined with comprehensive judgment of acceleration threshold and direction change rate, overcomes the defects of fuzzy action boundary demarcation and inability to accurately extract short actions in the prior art, and further sets an intention prediction mechanism based on double condition joint determination of hand staying time and motion speed before gesture recognition, thereby avoiding misjudgment of the system on non-intention actions such as unintentional swiping, and reducing the false touch rate of air keys through infrared transmitting and receiving to detect the shielding distance and finger staying time of a pipe array, combined with a triple condition determination mechanism of continuous multi-frame stability verification.

Description

Technical Field

[0001] This invention relates to the field of intelligent equipment technology, specifically to a full-area air-touch adaptive display conference all-in-one machine. Background Technology

[0002] With the normalization of hybrid office and digital teaching, smart conference all-in-one machines that integrate functions such as high-definition display, multi-touch, video conferencing, and wireless screen projection have become the core interactive terminals in scenarios such as corporate conference rooms, education and training, and digital exhibition halls.

[0003] However, existing touch-screen conference all-in-one machines lack an effective user intent judgment mechanism in actual use. When the sensor is too sensitive, the system is prone to misjudging the user's non-intent actions. For air button operations, existing technology lacks targeted anti-mistouch detection methods. A brief pause or quick swipe of the finger in the button area can easily be misidentified as a valid operation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a full-area air-touch adaptive display conference all-in-one machine, which solves the problems of insufficient accuracy and stability of gesture recognition and the ease with which misjudgments can occur.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a full-area air-touch adaptive display conference all-in-one machine, comprising:

[0006] The multimodal sensing module is used to acquire the user's continuous gesture action stream data, and realize the temporal deconstruction of the continuous gesture action stream by timestamp marking. Then, it identifies the action change points and divides the continuous gesture action stream into individual independent touch action units to obtain the corresponding gesture actions.

[0007] The effective gesture recognition module is used to perform intent prediction before gesture recognition. This includes recording the dwell time of the hand in the sensing area and tracking the palm to calculate the movement speed, dynamically calculating the adaptive dwell time threshold and speed threshold, calculating the intent confidence score based on the dwell time and threshold, and activating gesture recognition according to the confidence level. After intent prediction, the gesture action is sequentially verified by direction vector angle constraint, key point length ratio constraint and slope fluctuation value constraint. If all three are passed, the gesture is determined to be valid.

[0008] The coordinate mapping module is used to map the hand coordinates of valid gestures to the screen coordinates of the conference all-in-one machine, and to map valid gestures to system events. Based on the system events and screen coordinates, the conference all-in-one machine is controlled to perform corresponding operations.

[0009] As a further aspect of the present invention, the specific steps of the action change point include:

[0010] Calculate the movement speed and acceleration of the preset reference point of the hand. When the movement speed accelerates from below the preset speed threshold to above the preset start threshold, mark the position as the starting point of the action. When the movement speed decelerates from above the preset speed threshold to below the preset stop threshold, mark the position as the ending point of the action.

[0011] As a further aspect of the present invention, the step of determining the user's gesture intention based on the dwell time and movement speed of the user's hand within the sensing area specifically includes:

[0012] When the hand first enters the sensing area, the entry time t is recorded. 进 When the hand leaves the sensing area, the time t of departure is recorded. 离 Calculate the dwell time T 停 =t 离 -t 进 Using the palm as the key point, the system tracks the position change of the hand within the sensing area. Let the coordinates of the palm in the nth frame be (x... n y n , z n Then the palm coordinates of the (n-1)th frame are (x n-1 y n-1 , z n-1 ), obtain the inter-frame time interval Δt, then the instantaneous velocity .

[0013] As a further aspect of the present invention, the method for dynamically calculating the adaptive dwell time threshold and speed threshold is as follows:

[0014] According to formula T th =T base +ΔT user +ΔT scene Calculate the dwell time threshold T th T base Based on the time threshold, ΔT user ΔT is the amount of user adaptive adjustment. scene For scene adaptation adjustment, according to formula V th =V base +ΔV user +ΔV scene Calculate the motion velocity threshold V th， Where V base Based on the velocity threshold, ΔV user ΔV is the amount of user-adaptive adjustment. scene Adjustments for scene adaptability.

[0015] As a further aspect of the present invention, the step of calculating the intent confidence score based on dwell time and a threshold and activating gesture recognition according to the confidence level includes:

[0016] According to formula C base =w T ×f T (T 停 T th )+w V ×f V (V, V) th ), where f T (T 停 T th ) is the membership function for dwell time, f V (V, V) th Velocity membership function, w T and w V These are the corresponding weight coefficients, based on the obtained basic intent confidence score, combined with auxiliary feature correction, according to formula C. final =min(1, max(0, C) base +C aux The final confidence score C is calculated. final C aux Indicates the contribution value of auxiliary features;

[0017] When the final confidence score C final When the confidence score is ≥0.75, it indicates a high-confidence intent, and gesture recognition is immediately activated. When the final confidence score C... final A confidence level between 0.4 and 0.75 is considered moderate. After extending the observation window by 200ms, the confidence level is reassessed. When the final confidence score is C... final If the confidence level is less than 0.4, it is considered low and should be ignored.

[0018] As a further aspect of the present invention, the direction vector angle constraint verification specifically includes:

[0019] The displacement vector from fingertip to palm in the start and end frames of the gesture action is calculated as the gesture direction vector. The horizontal rightward direction of the screen is used as the reference direction vector. The angle between the two is calculated. If the angle exceeds the range of [15°, 165°], the gesture is deemed invalid.

[0020] The keypoint length ratio constraint is verified by calculating the Euclidean distance D between the fingertip of the middle finger and the palm. a And the Euclidean distance D between the tips of the index and little fingers. b Calculate the ratio Z=D a / D b If Z exceeds the range [0.3, 2.5], the gesture is deemed invalid.

[0021] The slope fluctuation value constraint verification is as follows: obtain the palm trajectory point sequence during the gesture execution, and calculate the standard deviation of the slope of adjacent frames as the slope fluctuation value after median filtering. If the slope fluctuation value is greater than 0.3, the gesture is determined to be invalid.

[0022] As a further aspect of the present invention, it also includes an anti-misoperation module for judging the anti-mistouch operation of the air button: the occlusion distance D1 of each air button area is obtained by using an infrared transceiver array, and the dwell time of the finger above each button is detected. When the occlusion distance D1 is less than the preset trigger threshold Dth and the dwell time is greater than or equal to the preset time threshold, it is determined to be a valid press; otherwise, it is determined to be a mistouch.

[0023] As a further aspect of the present invention, the specific method for determining the anti-misoperation module is as follows:

[0024] In the anti-misoperation module, when there is no obstruction, the obstruction distance is recorded as the maximum obstruction distance Dmax; for each infrared transceiver pair associated with the air button area, it is detected whether the obstruction distance D1 is continuously lower than Dth and whether the dwell time reaches the preset time threshold. If both conditions are met, it is determined to be a valid press.

[0025] As a further aspect of the present invention, it also includes a motion amplitude optimization module, which is used to obtain the actual movement distance of the finger and the movable range of the virtual control on the screen, determine the number of intervals according to the movable range and divide the maximum single-trip movement distance of the finger into multiple intervals, determine the target interval number where the actual movement distance of the finger is located, calculate the movement distance of the control according to the formula and control the movement of the control.

[0026] As a further aspect of the present invention, the specific processing procedure of the motion amplitude optimization module is as follows:

[0027] Recognize the target air gesture, obtain the actual finger movement distance D2 corresponding to the gesture, and determine the current movable range R of the virtual control on the screen. Then, based on the movable range R of the virtual control and the basic control precision P... base The number of intervals N is dynamically calculated, and N = min(10, max(3, R / P)). base Based on the number of intervals N, the maximum single-trip distance D2max of the finger is divided into N consecutive finger distance intervals.

[0028] First, the area [0, D2max] is uniformly divided into N equal-length intervals, each with a length L0 = D2max / N. Based on the operational characteristics of the virtual control, an interval density adjustment factor α is introduced, where 0.5 ≤ α ≤ 2.0. Each interval is assigned a number i, i = 1, 2, ..., N, in order from the start to the end. The boundary positions are determined by a nonlinear mapping function B. i =D2max×(i / N)α Calculate boundary B i ;

[0029] Compare the actual finger movement distance D2 with the boundary B i When comparing, when B k-1 ≤D2≤B k When the finger actually moves, the distance D2 is determined to belong to the k-th finger distance interval. If D2 = D2max, it is determined to belong to the last interval. Based on the determined finger distance interval k and the total number of intervals N, the movement distance DL of the virtual control on the screen is calculated. Specifically, DL = (k / N) × R.

[0030] This invention provides a full-area, air-touch adaptive display conference all-in-one machine. Compared with existing technologies, it has the following advantages:

[0031] This invention detects action change points by timestamping and deconstructing continuous gesture action streams, and combining acceleration thresholds and direction change rates for comprehensive judgment, thus achieving accurate segmentation of continuous gestures. On this basis, it models the spatial topological relationship between key points of the hand through graph structure modeling, and uses graph convolutional networks to handle the dynamic changes in the connection relationship between nodes. Even when the hand is occluded or the structure is deformed, it can still maintain the continuity and stability of structural judgment, overcoming the defects of vague action boundary delineation and inaccurate extraction of short actions in the prior art.

[0032] This invention establishes an intent prediction mechanism based on a dual condition of hand dwell time and movement speed before entering gesture recognition. The gesture recognition mode is activated only when the hand dwell time in the sensing area reaches a preset dwell threshold and the movement speed does not exceed a preset speed threshold. This avoids the system misjudging unintentional actions such as unintentional swipes. For air button operations, the invention uses an infrared transceiver array to detect the occlusion distance and finger dwell time, combined with a triple condition judgment mechanism of continuous multi-frame stability verification, which reduces the false touch rate of air buttons. Attached Figure Description

[0033] Figure 1 This invention provides a block diagram of a conference all-in-one machine system.

[0034] Figure 2 The following is a flowchart of the operation of the conference all-in-one machine for this invention. Detailed Implementation

[0035] 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.

[0036] Example 1

[0037] Please see Figure 1 and Figure 2 This application provides a full-area air-touch adaptive display conference all-in-one machine, including:

[0038] The multimodal sensing module is used to acquire the user's continuous gesture action stream data, and realize the temporal deconstruction of the continuous gesture action stream by timestamp marking. Then, it identifies the action change points and divides the continuous gesture action stream into individual independent touch action units to obtain the corresponding gesture actions.

[0039] The effective gesture recognition module is used to determine the user's intention based on the dwell time and movement speed of the user's hand in the sensing area. After the intention is predicted, the obtained gesture is sequentially verified by the direction vector angle constraint, the length ratio between key points constraint, and the slope fluctuation value constraint. When all three are verified, the gesture is determined to be effective.

[0040] The coordinate mapping module is used to map the hand coordinates of valid gestures to the screen coordinates of the conference all-in-one machine, and to map valid gestures to system events. Based on the system events and screen coordinates, the conference all-in-one machine is controlled to perform corresponding operations.

[0041] Example 2

[0042] As a second embodiment of the present invention, it is implemented based on the first embodiment, but differs from the first embodiment in the following ways:

[0043] The multimodal sensing module collects user's air gesture data through multimodal sensors, including a depth camera and an infrared sensor array. The depth camera collects the three-dimensional spatial coordinates of key points on the user's hand, and the infrared sensor array collects the distribution data of infrared signal intensity reflected from the hand. Based on the obtained air gesture data, the module acquires continuous gesture action stream data and timestamps each frame in the continuous gesture action stream data. The timestamps are used to deconstruct the temporal sequence of the continuous gesture action stream data and detect action change points in the continuous gesture action stream data. Specifically, the acceleration of key points on the hand is calculated. When the speed accelerates from a standstill to exceed the start threshold, it is marked as the start of the action. When the speed slows down and falls below the stop threshold, it is marked as the end of the action. This is used to identify action change points. Based on the identified action change points, the temporally deconstructed continuous gesture action stream is divided into individual independent touch action units. Each action unit corresponds to one complete air touch operation, resulting in the corresponding gesture action.

[0044] The effective gesture recognition module first determines the user's gesture intention before entering the air gesture recognition area. When the hand first enters the sensing area, the entry time t is recorded. 进 When the hand leaves the sensing area, the time t of departure is recorded. 离 Calculate the dwell time T 停 =t 离 -t 进 Using the palm as the key point, the system tracks the position change of the hand within the sensing area. Let the coordinates of the palm in the nth frame be (x... n y n , z n Then the palm coordinates of the (n-1)th frame are (x n-1 y n-1 , z n-1 ), obtain the inter-frame time interval Δt, then the instantaneous velocity ;

[0045] Next, according to formula T th =T base +ΔT user +ΔT scene Calculate the dwell time threshold T th T base Based on the time threshold, ΔT user ΔT is the amount of user adaptive adjustment. scene For scene adaptability adjustment, according to formula V th =V base +ΔV user +ΔV scene Calculate the motion velocity threshold V th， Where V base Based on the velocity threshold, ΔV user ΔV is the amount of user-adaptive adjustment. scene Adjustments for scene adaptability;

[0046] Next, the user's basic intent confidence score is calculated, specifically C. base =w T ×f T (T 停 T th )+w V ×f V (V, V) th ), where f T (T 停 T th ) is the membership function for dwell time, when T 停 <0.5×T th When T, the overall value is 0. 停 >T th When the whole is 1, f V (V, V)th Velocity membership function, when V < 0.5 × V th When V > V, the overall value is 0. th At that time, the whole is 1, w T and w V These are the corresponding weight coefficients, based on the obtained basic intent confidence score, combined with auxiliary feature correction, according to formula C. final =min(1, max(0, C) base +C aux The final confidence score C is calculated. final C aux This represents the contribution value of auxiliary features. Specific auxiliary features include abrupt changes in trajectory curvature and changes in hand orientation angle. When there is an abrupt change in trajectory curvature or a significant change in hand orientation angle, C... aux Take a positive value; when the hand trajectory is chaotic, C aux Take the negative value and base it on the final confidence score C. final To categorize intentions into different levels:

[0047] When the final confidence score C final When the confidence score is ≥0.75, it indicates a high-confidence intent, and gesture recognition is immediately activated. When the final confidence score C... final A confidence level between 0.4 and 0.75 is considered moderate. After extending the observation window by 200ms, the confidence level is reassessed. When the final confidence score is C... final If the confidence level is less than 0.4, it is considered low and should be ignored.

[0048] After the intention is predicted, the obtained gesture action is constrained and verified, specifically including the direction vector angle constraint, the length ratio between key points constraint, and the slope fluctuation value constraint.

[0049] The specific verification process for the directional vector angle constraint is as follows: obtain the start frame and end frame of the gesture action, calculate the displacement vector with the fingertip and palm as key points in the start frame and end frame to obtain the directional vector of the gesture action, and take the horizontal rightward direction of the screen as the reference directional vector, calculate the angle between the gesture directional vector and the reference directional vector, and compare it with the preset angle range, which is [15°, 165°]. Specifically, when the angle is less than 15°, the gesture direction is almost coincident with the reference direction, which may be a natural hand shake rather than an intentional gesture. When the angle is greater than 165°, the gesture direction is almost opposite to the reference direction, which may also be an unintentional gesture. If the angle exceeds the preset angle range, the gesture action is determined to be invalid.

[0050] The verification process for the length ratio constraint between key points is as follows: Select the fingertip of the middle finger, the palm, the fingertip of the index finger, and the fingertip of the little finger as key points. Calculate the Euclidean distance Da between the fingertip of the middle finger and the palm, and the Euclidean distance Db between the fingertip of the index finger and the fingertip of the little finger, and calculate the ratio between the two to obtain the length ratio Z = Da / Db. Compare this ratio with a preset range of 0.3 to 2.5. If the length ratio Z exceeds the preset range, the gesture is deemed invalid.

[0051] The verification process for slope fluctuation value constraint is as follows: First, obtain the trajectory point sequence with the palm as the key point in several consecutive sampling frames during the execution of the gesture action. Preprocess the trajectory point sequence and remove outliers by median filtering. Then, calculate the slope value sequence between adjacent frames and calculate the standard deviation of the slope value sequence to obtain the slope fluctuation value. Calculate and compare it with the slope fluctuation value threshold, and the normal value is 0.3. When the slope fluctuation value exceeds the threshold, the gesture is determined to be invalid.

[0052] The coordinate mapping module determines a gesture to be valid only if all the above constraint verifications pass. Based on the valid gesture, it maps the hand coordinates to the screen coordinates of the all-in-one device, specifically the horizontal screen coordinate X. 横 =(X hand -X min ) / (X max -X min )×W screen The screen's vertical coordinate Y 纵 =(Y hand -Y min ) / (Y max -Y min )×H screen , where X hand and Y hand Let X be the coordinates of the hand within the sensing area. min X max Y min Y max W represents the boundary coordinates of the sensing area. screen and H screen The width and height of the screen are defined, and valid gestures are mapped to system events, including page turning, zooming, rotating, selecting, confirming, canceling, volume adjustment, brightness adjustment, etc.

[0053] Example 3

[0054] As a third embodiment of the present invention, it is implemented based on the valid gesture action determined in the second embodiment, and the difference from the second embodiment is that this embodiment performs anti-accidental touch judgment based on the air button operation obtained from the valid gesture recognition.

[0055] The anti-misoperation module uses an infrared transceiver array to obtain the occlusion distance of each air button area and detects the dwell time of the finger above each button. When the occlusion distance is less than a preset trigger threshold and the dwell time is greater than or equal to a preset time threshold, it is determined to be a valid press; otherwise, it is determined to be a misoperation. The specific processing procedure is as follows:

[0056] Specifically, an infrared transceiver array is used for scanning to obtain the blocking distance D1 of each infrared transceiver pair. The blocking distance represents the distance between the blocking object and the infrared emitting tube when the intensity of the infrared signal received by the infrared receiving tube is attenuated, calculated based on the degree of attenuation. When there is no blocking object, the blocking distance of each infrared transceiver pair is recorded as the maximum blocking distance Dmax. When a finger enters the sensing area and stays above a certain button area, the blocking distance of the corresponding infrared transceiver pair decreases from the maximum blocking distance Dmax to the blocking distance D1.

[0057] For each infrared transceiver pair associated with an air button area, it is detected whether the obstruction distance D1 is continuously lower than the preset trigger threshold Dth, and at the same time, it is detected whether the dwell time of the finger above the button exceeds the preset time threshold. When the obstruction distance D1 above a button is less than the preset trigger threshold Dth and the duration is greater than or equal to the preset time threshold, it is determined to be a valid press. If the obstruction distance D1 is greater than or equal to Dth or the dwell time is less than the preset time threshold, it is determined to be a false touch and no operation is triggered.

[0058] Example 4

[0059] As a fourth embodiment of the present invention, it is implemented based on the second embodiment, and the difference from the second embodiment is that the action amplitude scaling mapping processing is performed based on the air gesture operation obtained by effective gesture recognition.

[0060] The motion amplitude optimization module identifies the type of air gesture performed by the user, determines the virtual control object corresponding to the gesture, and obtains the actual finger movement distance D2, the maximum single-journey finger movement distance D2max, and the current movable range R of the virtual control.

[0061] The method for obtaining the actual finger movement distance D2 is as follows: during the execution of the current air gesture, the fingertip of the index finger or the palm is used as the tracking key point, and the Euclidean distance of the key point from the starting position of the gesture to the current frame position in three-dimensional space is calculated as the actual finger movement distance.

[0062] The maximum single-stroke movement distance D2max of the finger is obtained by: based on the spatial range of the air gesture sensing area and the effective spatial boundary of the current user's hand activity, the maximum effective single-direction movement distance that the finger can complete in a single gesture is preset. This parameter can be dynamically calculated by the system based on information such as the user's arm length and sensing area boundary collected by the depth camera, or it can be manually calibrated by the user in the settings interface.

[0063] The current movable range R of a virtual control is obtained as follows: determine the movable range of the target virtual control in the current screen display state; for slider controls, R is the length of the slider's track (in pixels); for window controls, R is the maximum draggable distance of the window in the corresponding direction (in pixels); for scrollbar controls, R is the total scrollable distance of the scrollbar.

[0064] Next, based on the movable range R of the virtual control and the basic control precision P... base The number of intervals N is dynamically calculated, and N = min(10, max(3, R / P)). base Based on the number of intervals N, the maximum single-trip finger movement distance D2max is divided into N consecutive finger distance intervals. A dynamic non-linear division strategy is used here, as detailed below:

[0065] First, the area [0, D2max] is uniformly divided into N equal-length intervals, each with a length L0 = D2max / N. Based on the operational characteristics of the virtual control, an interval density adjustment factor α is introduced, where 0.5 ≤ α ≤ 2.0. For controls requiring fine-tuning, α is greater than 1; for controls requiring rapid, large-range movement, α is less than 1. Each interval is assigned a number i, i = 1, 2, ..., N, in order from the start to the end. The boundary positions are determined by a non-linear mapping function B. i =D2max×(i / N) α Calculate boundary B i Specifically, when α=1, the division is uniform; when α>1, the starting intervals are denser; when α<1, the ending intervals are denser.

[0066] Compare the actual finger movement distance D2 with the boundary B i When comparing, when B k-1 ≤D2≤B k When the finger actually moves, the distance D2 is determined to belong to the k-th finger distance interval. If D2 = D2max, it is determined to belong to the last interval. Based on the determined finger distance interval k and the total number of intervals N, the movement distance DL of the virtual control on the screen is calculated. Specifically, DL = (k / N) × R.

[0067] Example 5

[0068] As a fifth embodiment of the present invention, the focus is on combining the implementation processes of all the above embodiments.

[0069] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0070] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A full-area air-touch adaptive display conference all-in-one machine, characterized in that, include: The multimodal sensing module is used to acquire the user's continuous gesture action stream data, and realize the temporal deconstruction of the continuous gesture action stream by timestamp marking. Then, it identifies the action change points and divides the continuous gesture action stream into individual independent touch action units to obtain the corresponding gesture actions. The effective gesture recognition module is used to perform intent prediction before gesture recognition. This includes recording the dwell time of the hand in the sensing area and tracking the palm to calculate the movement speed, dynamically calculating the adaptive dwell time threshold and speed threshold, calculating the intent confidence score based on the dwell time and threshold, and activating gesture recognition according to the confidence level. After intent prediction, the gesture action is sequentially verified by direction vector angle constraint, key point length ratio constraint and slope fluctuation value constraint. If all three are passed, the gesture is determined to be valid. The coordinate mapping module is used to map the hand coordinates of valid gestures to the screen coordinates of the conference all-in-one machine, and to map valid gestures to system events. Based on the system events and screen coordinates, the conference all-in-one machine is controlled to perform corresponding operations.

2. The all-in-one conference machine with omnidirectional touch control and adaptive display according to claim 1, characterized in that, The specific steps for the action change point include: Calculate the movement speed and acceleration of the preset reference point of the hand. When the movement speed accelerates from below the preset speed threshold to above the preset start threshold, mark the position as the starting point of the action. When the movement speed decelerates from above the preset speed threshold to below the preset stop threshold, mark the position as the ending point of the action.

3. The all-in-one conference machine with full-area air-touch adaptive display according to claim 1, characterized in that, The step of determining the user's gesture intention based on the dwell time and movement speed of the user's hand in the sensing area specifically includes: When the hand first enters the sensing area, the entry time t is recorded. 进 When the hand leaves the sensing area, the time t of departure is recorded. 离 Calculate the dwell time T 停 =t 离 -t 进 Using the palm as the key point, the system tracks the position change of the hand within the sensing area. Let the coordinates of the palm in the nth frame be (x... n y n , z n Then the palm coordinates of the (n-1)th frame are (x n-1 y n-1 , z n-1 ), obtain the inter-frame time interval Δt, then the instantaneous velocity .

4. The all-in-one conference machine with full-area air-touch adaptive display according to claim 1, characterized in that, The method for dynamically calculating the adaptive dwell time threshold and speed threshold is as follows: According to formula T th =T base +ΔT user +ΔT scene Calculate the dwell time threshold T th T base Based on the time threshold, ΔT user ΔT is the amount of user adaptive adjustment. scene For scene adaptability adjustment, according to formula V th =V base +ΔV user +ΔV scene Calculate the motion velocity threshold V th， Where V base Based on the velocity threshold, ΔV user ΔV is the amount of user-adaptive adjustment. scene Adjustments for scene adaptability.

5. The all-in-one conference machine with omnidirectional touch control and adaptive display according to claim 1, characterized in that, The process of calculating the intent confidence score based on dwell time and a threshold, and activating gesture recognition according to the confidence level, includes: According to formula C base =w T ×f T (T 停 T th )+w V ×f V (V, V) th ), where f T (T 停 T th ) is the membership function for dwell time, f V (V, V) th Velocity membership function, w T and w V These are the corresponding weight coefficients, based on the obtained basic intent confidence score, combined with auxiliary feature correction, according to formula C. final =min(1, max(0, C) base +C aux The final confidence score C is calculated. final C aux Indicates the contribution value of auxiliary features; When the final confidence score C final When the confidence score is ≥0.75, it indicates a high-confidence intent, and gesture recognition is immediately activated. When the final confidence score C... final A confidence level between 0.4 and 0.75 is considered moderate. After extending the observation window by 200ms, the confidence level is reassessed. When the final confidence score is C... final If the confidence level is less than 0.4, it is considered low and should be ignored.

6. The all-in-one conference machine with full-area air-touch adaptive display according to claim 1, characterized in that, The verification of the direction vector angle constraint specifically includes: The displacement vector from fingertip to palm in the start and end frames of the gesture action is calculated as the gesture direction vector. The horizontal rightward direction of the screen is used as the reference direction vector. The angle between the two is calculated. If the angle exceeds the range of [15°, 165°], the gesture is deemed invalid. The keypoint length ratio constraint is verified by calculating the Euclidean distance D between the fingertip of the middle finger and the palm. a And the Euclidean distance D between the tips of the index and little fingers. b Calculate the ratio Z=D a / D b If Z exceeds the range [0.3, 2.5], the gesture is deemed invalid. The slope fluctuation value constraint verification is as follows: obtain the palm trajectory point sequence during the gesture execution, and calculate the standard deviation of the slope of adjacent frames as the slope fluctuation value after median filtering. If the slope fluctuation value is greater than 0.3, the gesture is determined to be invalid.

7. The all-in-one conference machine with full-area air-touch adaptive display according to claim 1, characterized in that, It also includes an anti-misoperation module, which is used to prevent accidental touches during air button operations: the occlusion distance D1 of each air button area is obtained by using an infrared transceiver array, and the dwell time of the finger above each button is detected. When the occlusion distance D1 is less than the preset trigger threshold Dth and the dwell time is greater than or equal to the preset time threshold, it is determined to be a valid press; otherwise, it is determined to be a mis-touch.

8. A full-area air-touch adaptive display conference all-in-one machine according to claim 7, characterized in that, The specific method for preventing accidental touches in the anti-misoperation module is as follows: In the anti-misoperation module, when there is no obstruction, the obstruction distance is recorded as the maximum obstruction distance Dmax; for each infrared transceiver pair associated with the air button area, it is detected whether the obstruction distance D1 is continuously lower than Dth and whether the dwell time reaches the preset time threshold. If both conditions are met, it is determined to be a valid press.

9. A full-area air-touch adaptive display conference all-in-one machine according to claim 1, characterized in that, It also includes a motion amplitude optimization module, which is used to obtain the actual movement distance of the finger and the movable range of the virtual control on the screen, determine the number of intervals based on the movable range, divide the maximum single-trip movement distance of the finger into multiple intervals, determine the target interval number where the actual movement distance of the finger is located, calculate the movement distance of the control according to the formula, and control the movement of the control.

10. A full-area air-touch adaptive display conference all-in-one machine according to claim 9, characterized in that, The specific processing procedure of the motion amplitude optimization module is as follows: Recognize the target air gesture, obtain the actual finger movement distance D2 corresponding to the gesture, and determine the current movable range R of the virtual control on the screen. Then, based on the movable range R of the virtual control and the basic control precision P... base The number of intervals N is dynamically calculated, and N = min(10, max(3, R / P)). base Based on the number of intervals N, the maximum single-trip distance D2max of the finger is divided into N consecutive finger distance intervals. First, the area [0, D2max] is uniformly divided into N equal-length intervals, each with a length L0 = D2max / N. Based on the operational characteristics of the virtual control, an interval density adjustment factor α is introduced, where 0.5 ≤ α ≤ 2.

0. Each interval is assigned a number i, i = 1, 2, ..., N, in order from the start to the end. The boundary positions are determined by a nonlinear mapping function B. i =D2max×(i / N) α Calculate boundary B i ; Compare the actual finger movement distance D2 with the boundary B i When comparing, when B k-1 ≤D2≤B k When the finger actually moves, the distance D2 is determined to belong to the k-th finger distance interval. If D2 = D2max, it is determined to belong to the last interval. Based on the determined finger distance interval k and the total number of intervals N, the movement distance DL of the virtual control on the screen is calculated. Specifically, DL = (k / N) × R.

Images