An aerial gesture control method and system based on attitude compensation

Abstract

This invention discloses an aerial gesture control method and system based on posture compensation, belonging to the field of human-computer interaction and robot control technology. The invention uses a hardware synchronization mechanism to synchronously acquire the arm's baseline posture data from the first IMU module at the forearm and the composite posture data from the second IMU module at the fingertips. After coordinate system alignment and posture calculation, adaptive Kalman filtering is used to eliminate arm swing interference in real time, extracting pure fingertip waving motion data. Valid gestures are identified through dynamic thresholding, mapped to robot volume adjustment or on / off control commands, and wirelessly transmitted for execution. This invention completely eliminates arm swing interference, achieving precise fingertip control. It has advantages such as strong anti-interference capability, good real-time performance, and wide adaptability, and can be widely applied to non-contact human-computer interaction control of various robots.

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, and in particular to an aerial gesture control method and system based on posture compensation. Background Technology

[0002] With the rapid popularization of service robots and industrial collaborative robots, non-contact natural human-robot interaction has become a core requirement for robot on-site operation. Air gesture control, due to its characteristics of requiring no physical peripherals and being intuitive to operate, has become a research hotspot in this field. Existing air gesture control solutions are mainly divided into two categories: The first type is a gesture recognition solution based on computer vision. This solution relies on depth cameras or monocular cameras to collect gesture images. It is easily affected by ambient light, gesture occlusion, and shooting distance. In industrial sites where robots operate and complex outdoor scenarios, the recognition accuracy drops sharply, and the edge computing latency is generally higher than 50ms, which cannot meet the low latency requirements of real-time robot control.

[0003] The second category is gesture recognition solutions based on wearable IMUs. Existing solutions mostly use a single IMU installed in the palm, which can only recognize large-scale movements of the entire arm and cannot separate macro-posture interference from natural arm swings and body swaying from fine fingertip waving movements, resulting in a false touch rate of over 30%. A few dual-IMU solutions have three core unavoidable defects: 1. No hardware-level time synchronization mechanism for dual IMUs is set up, resulting in sampling time differences between the two data streams, leading to cumulative errors in posture decoupling compensation; 2. The fixed frequency threshold filtering scheme cannot adapt to the frequency differences of arm swings of different users, resulting in incomplete interference elimination; 3. No dynamic zero-point calibration mechanism is set up, resulting in zero-point drift after long-term use and a continuous decrease in control accuracy.

[0004] The aforementioned defects make existing solutions unsuitable for control scenarios such as robot volume adjustment and start / stop switches, which require distinguishing between precise active movements and unconscious arm swings. They are prone to misoperation due to unconscious arm swings, posing serious safety hazards and failing to meet the high reliability control requirements of industrial and consumer robots. Summary of the Invention

[0005] To address the aforementioned core deficiencies of existing technologies, the present invention aims to provide an aerial gesture control method and system based on posture compensation. Through three core innovations—hardware-level dual IMU synchronous acquisition, adaptive frequency domain filtering decoupling, and dynamic threshold calibration—it completely eliminates interference caused by arm swinging and body swaying, accurately extracts fine fingertip waving movements, and achieves highly reliable, low-latency, and contactless precise control of robot volume and on / off functions. It also possesses strong user adaptability and adaptability to all scenarios.

[0006] The attitude-compensated aerial gesture control system of the present invention includes a wearable sensing unit, a main control processing unit, a wireless communication unit, and a robot execution unit. Wearable sensing unit: includes a first IMU module and a second IMU module; the first IMU module is fixed to the user's forearm near the wrist by a strap, and its coordinate system is fully aligned with the rigid body coordinate system of the arm, and is used to collect the motion posture data of the rigid body of the arm in real time as a reference for posture compensation; the second IMU module is fixed to the user's index fingertip by a flexible finger sleeve, and its coordinate system is aligned with the direction of fingertip movement, and is used to collect composite motion posture data including arm movement superimposed with fingertip waving in real time. Main control processing unit: electrically connected to the wearable sensing unit, with built-in hardware synchronization trigger module, data preprocessing module, attitude calculation module, adaptive attitude compensation module, gesture recognition and command mapping module, used to perform corresponding data processing and algorithm calculation; Wireless communication unit: electrically connected to the main control processing unit, used to wirelessly transmit the mapped control commands to the robot execution unit; Robot execution unit: It has a built-in wireless receiver module, volume control module and power switch control module, which are used to receive control commands and perform corresponding volume adjustment or on / off operation.

[0007] Core steps of control method S1 Synchronous Data Acquisition and Preprocessing: Through a hardware synchronous triggering mechanism, the first IMU module and the second IMU module are synchronously driven to acquire arm reference posture data and fingertip composite posture data in real time at the same sampling frequency; the two data channels are preprocessed by sliding mean filtering, power-on static zero bias calibration, and temperature drift compensation to eliminate random noise, sensor zero bias error and temperature drift error. S2 Coordinate System Alignment and Attitude Calculation: Predefined geodetic coordinate system, arm rigid body coordinate system, and fingertip coordinate system are used to complete the initial zero-point calibration upon power-on. Based on the pre-processed two-channel synchronous data, the first attitude quaternion of the first IMU module relative to the geodetic coordinate system and the second attitude quaternion of the second IMU module relative to the geodetic coordinate system are calculated using the quaternion-gradient descent method. Through coordinate system transformation, the second attitude quaternion is transformed to the arm rigid body coordinate system to obtain the initial relative attitude quaternion of the fingertip relative to the arm and the initial relative attitude data. S3 Adaptive Attitude Decoupling and Interference Compensation: Based on the arm reference attitude data, the motion frequency range of arm swing is identified in real time, and an adaptive Kalman filter model is constructed. Using the arm reference attitude data as the observation value, the initial relative attitude data is predicted and corrected. The rigid body motion interference components caused by arm swing and body swaying in the corresponding frequency range are adaptively filtered out, while the high-frequency motion components of active fingertip waving are retained, and the pure waving motion data of fingertip relative to arm is extracted. S4 Dynamic Threshold Valid Gesture Recognition and Command Mapping: Predefined basic action thresholds and command mapping rules for fingertip waving gestures, dynamically adjusting action thresholds based on user historical action data; threshold judgment and feature matching are performed on pure waving action data to identify valid fingertip waving gestures and map them to corresponding robot control commands; among which, the command mapping rules include: waving the fingertip left and right in the horizontal direction corresponds to robot volume increase and volume decrease commands, and waving the fingertip forward and backward in the vertical direction corresponds to robot power on / off / start / stop switch commands; S5 command transmission and execution: The mapped control commands are sent to the robot execution unit through the wireless communication unit. After receiving the commands, the robot execution unit completes the corresponding control operation.

[0008] Preferred Implementation In step S1, the hardware synchronization triggering mechanism is as follows: the main control processing unit synchronously outputs a trigger signal to the external interrupt pin of the two IMU modules through the same GPIO pin, so as to realize the microsecond-level synchronous acquisition of the two data and eliminate the compensation cumulative error caused by the sampling time difference.

[0009] In step S2, the specific steps for power-on initial zero-point calibration are as follows: the user keeps their arm and fingertips still for 3s~5s, the main control processing unit collects two channels of static data during this time period, calculates the initial relative attitude quaternion, and uses it as the reference zero point for subsequent attitude compensation.

[0010] In step S3, the arm reference posture data is analyzed in the frequency domain by Fast Fourier Transform (FFT) to identify the peak frequency of the current arm swing in real time and adaptively update the filter cutoff frequency threshold; the motion frequency range of the arm swing is 0.5Hz~2Hz, and the frequency range of the high-frequency motion component of the active fingertip waving is 2Hz~8Hz.

[0011] In step S4, the criteria for determining a valid gesture are: the duration of a pure waving motion is within the range of 50ms to 500ms, and the attitude increment and angular velocity corresponding to the motion both exceed the current dynamically adjusted motion threshold, further filtering out jitter noise and unconscious movements.

[0012] Beneficial effects Compared with the prior art, the present invention has the following outstanding substantive features and beneficial effects: Complete interference elimination and a significant improvement in control precision and anti-interference capability: This invention eliminates the cumulative error caused by the time difference between the sampling of two data streams by using hardware-level dual IMUs for microsecond-level synchronous acquisition; through adaptive Kalman filtering, it can adapt to the arm swing frequency of different users in real time, completely separating and eliminating rigid body motion interference from natural arm swing and body swaying. The fingertip action recognition accuracy is ≥99.5%, and the false touch rate is ≤0.2%, which completely solves the core pain points of high false touch rate and insufficient control precision in existing technologies.

[0013] Strong environmental adaptability and unrestricted adaptation to all scenarios: Adopting a wearable IMU sensing solution, it is completely unaffected by ambient light, gesture occlusion, and operating distance, and can be adapted to various robot operating environments such as indoor and outdoor industrial sites, home scenarios, and outdoor scenarios. Compared with vision solutions, it has an overwhelming advantage in environmental adaptability.

[0014] Excellent real-time performance with no lag in control: It adopts lightweight quaternion attitude calculation and adaptive Kalman filtering algorithm, with a single frame data processing latency of ≤8ms, which is far lower than the latency of more than 50ms in existing solutions. The superimposed end-to-end wireless transmission latency is ≤20ms, which fully meets the real-time control requirements of the robot.

[0015] Highly adaptable and user-friendly: Through a dynamic threshold calibration mechanism, the gesture judgment threshold can be automatically adjusted based on the user's action habits, eliminating the need for manual calibration by the user. It is suitable for users of different ages and with different operating habits, making it extremely easy to use. At the same time, through power-on static calibration and temperature drift compensation, the zero-point drift problem in long-term use has been solved, and the control stability has been greatly improved.

[0016] Intuitive operation and safe and reliable: horizontal fingertip waving is mapped to continuous volume adjustment, and forward and backward waving is mapped to discrete on / off control, which is in line with the user's natural operating habits and requires no complicated learning; through the relative motion judgment mechanism, it completely eliminates the misoperation caused by unconscious arm swinging, which greatly improves the safety and reliability of robot control.

[0017] With a simple hardware structure and low industrialization cost, it requires only two low-cost MEMSMU modules, a general-purpose microcontroller and a wireless communication unit. It is easy to wear and does not require complex supporting equipment. It can be quickly adapted to the control systems of various service robots and industrial collaborative robots, and has extremely high industrialization and promotion value. Attached Figure Description

[0018] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a block diagram of the overall architecture of the air gesture control system based on attitude compensation in an embodiment of the present invention; Figure 2This is a flowchart illustrating the air gesture control method based on attitude compensation in an embodiment of the present invention. Figure 3 This is a schematic diagram showing the installation positions of the first IMU module and the second IMU module in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the definition and instruction mapping of fingertip waving gestures in an embodiment of the present invention; Figure 5 This is a schematic diagram of the algorithm flow for adaptive Kalman filtering in an embodiment of the present invention.

[0019] In the diagram: Wearable sensing unit 1, first IMU module 11, second IMU module 12, main control processing unit 2, wireless communication unit 3, robot execution unit 4. Detailed Implementation

[0020] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0021] Example 1: High-precision industrial-grade example Reference Figure 1 The attitude-compensated aerial gesture control system of this embodiment includes a wearable sensing unit 1, a main control processing unit 2, a wireless communication unit 3, and a robot execution unit 4.

[0022] The wearable sensing unit 1 includes a first IMU module 11 and a second IMU module 12. In this embodiment, both the first IMU module 11 and the second IMU module 12 adopt a 9-axis MEMSMU sensor, which integrates a 3-axis accelerometer, a 3-axis gyroscope, and a 3-axis magnetometer. The measurement ranges are ±16g, ±2000° / s, and ±4900μT, respectively, and the data output frequency is 200Hz, which meets the requirements for high-precision real-time acquisition.

[0023] Reference Figure 3 The first IMU module 11 is fixed to the user's forearm near the wrist by a nylon strap. Its coordinate system is fully aligned with the rigid body coordinate system of the forearm. It is used to collect motion posture data of the rigid body of the arm in real time as a reference for posture compensation. The second IMU module 12 is fixed to the tip of the user's index finger by a flexible silicone finger sleeve. Its coordinate system is aligned with the movement direction of the fingertip. It is used to collect composite motion posture data including arm swing and fingertip waving in real time.

[0024] The main control processing unit 2 uses an STM32F407 microcontroller with a built-in ARM Cortex-M4 core and a main frequency of 168MHz, which has sufficient computing power to process data from two IMUs. The main control processing unit 2 is electrically connected to the first IMU module 11 and the second IMU module 12 via an I2C bus, and synchronously outputs a trigger signal to the external interrupt pin of the two IMU modules through the same GPIO pin to achieve microsecond-level synchronous acquisition of the two data streams. The main control processing unit 2 has built-in hardware synchronization trigger module, data preprocessing module, attitude calculation module, adaptive attitude compensation module, and gesture recognition and command mapping module, which execute the corresponding data processing steps respectively.

[0025] The wireless communication unit 3 uses a Bluetooth 5.0 module and is electrically connected to the main control processing unit 2 via a UART serial port. The communication distance can reach 10m, which meets the distance requirements for on-site control of industrial robots. The end-to-end transmission delay is ≤20ms. It is used to wirelessly transmit the control commands mapped by the main control processing unit 2 to the robot execution unit 4.

[0026] Robot execution unit 4 is the main control terminal of the industrial collaborative robot. It has a built-in Bluetooth receiver module, volume control module and power switch control module, which are used to receive control commands transmitted wirelessly and execute corresponding volume increase, volume decrease, power on and power off operations.

[0027] Reference Figure 2 The specific implementation steps of the attitude-compensated air gesture control method in this embodiment are as follows: Step S1: Synchronous Data Acquisition and Preprocessing The main control processing unit 2 uses a hardware synchronization trigger module to synchronously drive the first IMU module 11 and the second IMU module 12 at a sampling frequency of 200Hz. It synchronously reads the arm reference posture data (denoted as Acc1, Gyr1, Mag1) collected by the first IMU module 11 and the fingertip composite posture data (denoted as Acc2, Gyr2, Mag2) collected by the second IMU module 12. The sampling time difference between the two data is ≤10μs, which eliminates the compensation cumulative error caused by the sampling time difference.

[0028] The two data streams are preprocessed as follows: First, a 5th-order moving average filter is used to eliminate random high-frequency noise during sensor acquisition; then, a power-on static zero-bias calibration is performed to eliminate the zero-bias error of the accelerometer and gyroscope; finally, the built-in temperature compensation model is used to eliminate the error caused by sensor temperature drift, resulting in preprocessed reference data and composite data.

[0029] Step S2: Coordinate System Alignment and Attitude Calculation First, three right-handed rectangular coordinate systems are predefined: Geodetic coordinate system n: The origin is the ground point where the user initially stands, the x-axis points due east, the y-axis points due north, and the z-axis is perpendicular to the ground and pointing upwards; Arm rigid body coordinate system b1: The origin is the installation center of the first IMU module 11, the x-axis points from the forearm to the wrist, the y-axis points horizontally to the right along the forearm, and the z-axis is perpendicular to the forearm and upward, and is rigidly connected to the forearm rigid body. The fingertip coordinate system b2 has its origin at the installation center of the second IMU module 12, the x-axis points from the fingertip to the fingertips, the y-axis points horizontally to the right along the fingertip, and the z-axis points vertically upwards from the fingertip and is fixed to the fingertip.

[0030] Upon initial power-on, an initial zero-point calibration is performed: the user keeps their arm and fingertips still for 3 seconds, and the main control processing unit 2 collects two channels of static data during this time period and calculates the initial relative attitude quaternion, which serves as the reference zero point for subsequent attitude compensation.

[0031] Based on the preprocessed two-channel synchronous data, the quaternion-gradient descent method is used to calculate the first attitude quaternion q_n_b1 of the first IMU module 11 relative to the geodetic coordinate system n, and the second attitude quaternion q_n_b2 of the second IMU module 12 relative to the geodetic coordinate system n. Using the coordinate system transformation formula, the second attitude quaternion q_n_b2 is transformed to the arm rigid body coordinate system b1, yielding the initial relative attitude quaternion q_b1_b2 of the fingertip relative to the arm. The transformation formula is: q_b1_b2 = q_n_b1 -1 q_n_b2 where q_n_b1 -1 The conjugate quaternion of the first-position quaternion. This is a quaternion multiplication operator. Using the initial relative attitude quaternion, it calculates the initial relative attitude data of the fingertip relative to the arm, including relative Euler angles (roll, pitch, yaw), relative acceleration, and relative angular velocity.

[0032] Step S3 Adaptive Attitude Decoupling and Disturbance Compensation The core of this step is to eliminate rigid body motion interference caused by arm swinging, extract pure waving motion data of the fingertips, and refer to... Figure 5The specific implementation steps are as follows: S301 Constructing the state equation: Using the fingertip relative attitude quaternion and relative angular velocity as state variables, construct the discrete state equation: X(k) = A·X(k-1) + B·U(k) + W(k), where X(k) is the state vector at time k, A is the state transition matrix, B is the control input matrix, U(k) is the control input vector, and W(k) is the system process noise; S302 Constructing the observation equation: Using the arm reference attitude data collected by the first IMU module 11 as the observation value, construct the observation equation: Z(k) = H·X(k) + V(k), where Z(k) is the observation vector at time k, H is the observation matrix, and V(k) is the observation noise; S30 3. Frequency Identification: The arm reference posture data is analyzed in the frequency domain using Fast Fourier Transform (FFT) to identify the peak frequency of the current arm swing in real time and update the cutoff frequency threshold of the filter. In this embodiment, the motion frequency range of the arm swing is 0.5Hz~2Hz, and the frequency range of the high-frequency motion component of the fingertip active waving is 2Hz~8Hz. S304 Prediction and Correction: Based on the updated filter threshold, the Kalman prediction and update step is performed to adaptively filter out the low-frequency interference component corresponding to the arm swing, retain the high-frequency motion component of the fingertip active waving, and output the corrected fingertip relative posture increment and angular velocity increment data, i.e., pure waving motion data, completely eliminating the interference of the arm swing.

[0033] Step S4: Dynamic Threshold Effective Gesture Recognition and Command Mapping Reference Figure 4 The basic action threshold and instruction mapping rules for fingertip waving gestures are preset. The specific rules in this embodiment are as follows: Volume adjustment gesture: The fingertip is waved left and right along the horizontal direction (yaw angle direction). The basic angle increment threshold is 15° and the basic angular velocity threshold is 50° / s. When the absolute value of the relative yaw angle increment is greater than the current dynamic threshold and the peak value of the relative angular velocity is greater than the current dynamic threshold, it is determined to be a valid horizontal waving gesture. Among them, waving the fingertip to the left is mapped to the robot volume decrease command, and waving to the right is mapped to the robot volume increase command. A single waving gesture corresponds to a single level of volume adjustment, and continuous waving corresponds to continuous volume adjustment.

[0034] Switch control gesture: The fingertip is waved back and forth along the vertical direction (pitch angle direction). The basic angle increment threshold is 20° and the basic angular velocity threshold is 60° / s. When the absolute value of the relative pitch angle increment is greater than the current dynamic threshold and the peak value of the relative angular velocity is greater than the current dynamic threshold, it is determined to be a valid back and forth waving gesture. Among them, waving the fingertip forward is mapped to the robot power-on / start command, and waving backward is mapped to the robot power-off / stop command.

[0035] Meanwhile, a dynamic threshold calibration mechanism is set up: based on the average amplitude and speed of the user's first 10 valid actions, the action threshold is automatically adjusted to adapt to the operating habits of different users; valid gesture judgment conditions are set: only when the duration of a pure waving action is between 50ms and 500ms is it judged as a valid gesture, filtering out jitter noise with too short a duration and unconscious actions with too long a duration.

[0036] The main control processing unit 2 performs threshold judgment and feature matching on the extracted pure waving motion data, identifies valid fingertip waving gestures, and maps them into corresponding robot control commands according to the above rules.

[0037] Step S5: Instruction Transmission and Execution The main control processing unit 2 sends the mapped control commands to the wireless communication unit 3 via the UART serial port. The wireless communication unit 3 then transmits the commands wirelessly to the robot execution unit 4 via Bluetooth. After receiving the control commands, the robot execution unit 4 adjusts the playback volume through the built-in volume control module or performs power-on / power-off operations through the power switch control module, thus completing the entire air gesture control process.

[0038] Example 2: Low-cost consumer-grade example The core difference between this embodiment and Embodiment 1 lies in the hardware selection and lightweight algorithm adaptation, specifically as follows: Both the first IMU module 11 and the second IMU module 12 adopt 6-axis MEMSMU sensors (with only a built-in 3-axis accelerometer and 3-axis gyroscope), and the sampling frequency is set to 100Hz. The main control processing unit 2 adopts an STM32F103 microcontroller, and the wireless communication unit 3 adopts a Bluetooth 4.2 module, which significantly reduces hardware costs. In this embodiment, the frequency range of the adaptive Kalman filter is fixed at 0.5Hz~2Hz (low-frequency interference range) and 2Hz~8Hz (high-frequency action range), simplifying the algorithm calculation and adapting to the computing power of the low-cost microcontroller. The gesture recognition accuracy of this embodiment is ≥98%, the false touch rate is ≤0.5%, and the single-frame data processing latency is ≤15ms, which can meet the control requirements of consumer-grade service robots and has a very high cost advantage.

[0039] Those skilled in the art should understand that the above embodiments are merely preferred embodiments of the present invention, and any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An aerial gesture control method based on attitude compensation, characterized in that, Includes the following steps: S1 Synchronous Data Acquisition and Preprocessing: Through a hardware synchronous triggering mechanism, the first IMU module (11) fixed to the user's forearm and the second IMU module (12) fixed to the user's index fingertip are synchronously driven to acquire the arm reference posture data and fingertip composite posture data in real time at the same sampling frequency; the two posture data are sequentially preprocessed by sliding mean filtering, power-on static zero bias calibration, and temperature drift compensation to eliminate random noise, sensor zero bias error and temperature drift error; S2 coordinate system alignment and attitude calculation: predefine the geodetic coordinate system, the arm rigid body coordinate system and the fingertip coordinate system, and complete the initial zero point calibration upon power-on; Based on the preprocessed two-channel synchronous data, the first attitude quaternion of the first IMU module (11) relative to the geodetic coordinate system and the second attitude quaternion of the second IMU module (12) relative to the geodetic coordinate system are solved by the quaternion-gradient descent method respectively; by coordinate system transformation, the second attitude quaternion is transformed to the rigid body coordinate system of the arm to obtain the initial relative attitude quaternion of the fingertip relative to the arm and the initial relative attitude data. S3 Adaptive Attitude Decoupling and Disturbance Compensation: Based on the arm reference attitude data, the motion frequency range of arm swing is identified in real time, and an adaptive Kalman filter model is constructed. Using the arm reference posture data as the observation value, the initial relative posture data is predicted and corrected. Rigid body motion interference components in the corresponding frequency range are adaptively filtered out, while the high-frequency motion components of the fingertips are retained, and the pure waving motion data of the fingertips relative to the arm is extracted. S4 Dynamic Threshold Effective Gesture Recognition and Command Mapping: Predefined basic action thresholds and command mapping rules for fingertip waving gestures, dynamically adjusting action thresholds based on user's historical action data; Threshold judgment and feature matching are performed on pure waving motion data to identify valid gestures and map them to corresponding robot control commands; the command mapping rules include: horizontal left and right waving of the fingertip corresponds to robot volume adjustment command, and vertical forward and backward waving of the fingertip corresponds to robot on / off control command. S5 command transmission and execution: The mapped control commands are wirelessly transmitted to the robot, which then executes the corresponding control operations.

2. The aerial gesture control method based on attitude compensation according to claim 1, characterized in that, In step S1, both the first IMU module (11) and the second IMU module (12) use 6-axis or 9-axis MEMSMU sensors, and the sampling frequency is set to 100Hz~200Hz. The hardware synchronization triggering mechanism is as follows: the main control processing unit (2) synchronously outputs the trigger signal to the external interrupt pin of the two IMU modules through the same GPIO pin to realize the microsecond-level synchronous acquisition of the two data.

3. The aerial gesture control method based on attitude compensation according to claim 1, characterized in that, In step S2, the specific steps for power-on initial zero-point calibration are as follows: the user keeps his arm and fingertips still for 3s~5s, and the main control processing unit (2) collects two static data during this time period and calculates the initial relative attitude quaternion as the reference zero point for subsequent attitude compensation.

4. The aerial gesture control method based on attitude compensation according to claim 1, characterized in that, In step S2, the conversion formula for the initial relative attitude quaternion is: q_b1_b2=q_n_b1 -1 q_n_b2 where q_n_b1 is the first pose quaternion, q_n_b1 -1 Let q_n_b2 be the conjugate quaternion of the first pose quaternion, and q_n_b2 be the quaternion of the second pose quaternion. q_b1_b2 is the quaternion multiplication operator, and q_b1_b2 is the initial relative attitude quaternion.

5. The aerial gesture control method based on attitude compensation according to claim 1, characterized in that, In step S3, the arm reference posture data is analyzed in the frequency domain by fast Fourier transform to identify the peak frequency of the current arm swing in real time and adaptively update the filter cutoff frequency threshold; the motion frequency range of the arm swing is 0.5Hz~2Hz, and the frequency range of the high-frequency motion component of the fingertip active waving is 2Hz~8Hz.

6. The aerial gesture control method based on attitude compensation according to claim 1, characterized in that, In step S3, the extracted pure waving motion data includes the relative posture increment of the fingertip relative to the arm, the angular velocity increment, and the acceleration increment.

7. The aerial gesture control method based on attitude compensation according to claim 1, characterized in that, In step S4, the basic action thresholds include the angle increment threshold and the angular velocity threshold; the instruction mapping rules are as follows: a horizontal waving of the fingertip to the left corresponds to a robot volume decrease instruction, a horizontal waving of the fingertip to the right corresponds to a robot volume increase instruction, a vertical waving of the fingertip forward corresponds to a robot power on / start instruction, and a vertical waving of the fingertip backward corresponds to a robot power off / stop instruction.

8. An air gesture control system based on attitude compensation, used to implement the air gesture control method based on attitude compensation as described in any one of claims 1-7, characterized in that, The system includes a wearable sensing unit (1), a main control processing unit (2), a wireless communication unit (3), and a robot execution unit (4). The wearable sensing unit (1) includes a first IMU module (11) and a second IMU module (12). The first IMU module (11) is fixed to the user's forearm near the wrist by a strap and is used to collect arm reference posture data. The second IMU module (12) is fixed to the user's index fingertip by a flexible finger sleeve and is used to collect fingertip composite posture data. The main control processing unit (2) communicates with the wearable sensing unit (4) via an I2C bus. 1) Electrical connection, with built-in hardware synchronous trigger module, data preprocessing module, posture calculation module, adaptive posture compensation module, gesture recognition and command mapping module, used to execute the corresponding operations of steps S1 to S4 as described in claim 1; the wireless communication unit (3) is electrically connected to the main control processing unit (2) through UART serial port, used to wirelessly transmit control commands to the robot execution unit (4); the robot execution unit (4) has built-in wireless receiving module, volume control module and switch control module, used to receive control commands and execute corresponding volume adjustment or switch control operations.

9. The air gesture control system based on attitude compensation according to claim 8, characterized in that, Both the first IMU module (11) and the second IMU module (12) have a built-in 3-axis accelerometer and a 3-axis gyroscope, or a built-in 3-axis accelerometer, a 3-axis gyroscope and a 3-axis magnetometer; the main control processing unit (2) adopts an STM32F4 series microcontroller with a built-in ARM Cortex-M4 core.

10. The attitude-compensated aerial gesture control system according to claim 8, characterized in that, The hardware synchronization trigger module of the main control processing unit (2) synchronously outputs a trigger signal to the external interrupt pins of the first IMU module (11) and the second IMU module (12) through the same GPIO pin, so as to realize the microsecond-level synchronous acquisition of the two data channels.

Images