A coordinate system alignment method for inertial sensor network motion capture glove

Abstract

The application discloses a kind of coordinate system alignment methods for inertial sensing network motion capture glove, collect the acceleration data of multiple inertial measurement unit sensor nodes arranged on glove body during first alignment action, determine the unit vector of Z axis of human body coordinate system;Collect the angular velocity data of sensor node during second alignment action, determine the unit vector of X axis of human body coordinate system by principal component analysis;Z axis unit vector and X axis unit vector are subjected to Schmidt orthogonalization processing, and the modified X axis unit vector and Y axis unit vector are obtained;The projection component of the modified X axis unit vector, Y axis unit vector and Z axis unit vector in each axis direction of sensor coordinate system is arranged in column, and the rotation transformation matrix from sensor coordinate system to human body coordinate system is constructed;Original inertial data is transformed to human reference frame using the rotation transformation matrix.The method of the application has the advantages of high alignment stability, high alignment accuracy and short alignment time.

Description

Technical Field

[0001] This invention relates to the field of inertial measurement technology, and in particular to a coordinate system alignment method for motion capture gloves for inertial sensor networks. Background Technology

[0002] In the field of inertial measurement technology, motion capture gloves, as an important human-computer interaction device, are widely used in virtual reality, robot teleoperation, motion analysis, and other scenarios. Existing motion capture gloves mainly employ the following technical solutions to achieve attitude measurement:

[0003] The inertial alignment scheme based on Earth's rotation, disclosed in application number CN202410848931.0, utilizes a gyroscope and accelerometer in an inertial measurement unit (IMU) in conjunction with the Earth's rotation angular velocity for initial alignment. Specifically, this scheme requires the user to maintain a specific hand posture and remain still for at least 5 seconds. During the first 2 seconds, the system uses the angular velocity information output by the gyroscope and the specific force information output by the accelerometer, along with the Earth's rotation angular velocity and local gravitational acceleration, to analytically calculate the initial attitude rotation matrix. During the following 3 seconds, the system employs Kalman filtering for fine alignment, estimating gyroscope drift and attitude error, ultimately obtaining the initial attitude information.

[0004] The calibration scheme based on multi-sensor fusion, disclosed in application number CN202410848931.0, employs a combination of fiber optic flexible sensors and inertial sensors for calibration. Fiber optic sensors are positioned at the finger joints to measure joint bending angles; inertial sensors are positioned on the back of the hand and fingertips to measure posture. The system uses high-precision joint angle information obtained from the fiber optic sensors, combined with inertial sensor data, to obtain hand posture through a data fusion algorithm. The initial alignment of this scheme still relies on the aforementioned inertial alignment method based on Earth's rotation.

[0005] The alignment scheme based on mechanical calibration, proposed in application number CN201610231809.4, employs a fiber optic bending sensor or strain sensor. The operator performs specific calibration actions, recording the maximum and minimum values ​​of the sensor output to establish a mapping relationship between the sensor output and the joint angle. This scheme does not involve inertial coordinate system alignment, is only applicable to bending sensors, and cannot obtain the absolute posture of the hand in three-dimensional space.

[0006] However, the aforementioned prior art has the following drawbacks:

[0007] Inertial alignment schemes based on Earth's rotation suffer from poor stability. The core of this scheme lies in using the Earth's rotational angular velocity for initial alignment. However, the Earth's rotational angular velocity is extremely small, and the measurement accuracy of existing low-cost MEMS gyroscopes is insufficient to accurately capture such a weak angular velocity signal. Furthermore, this scheme requires the hand to remain absolutely still for an extended period, while the human hand inevitably experiences physiological tremors, leading to instability and significant alignment errors.

[0008] The initial alignment posture requirements are stringent. Existing solutions require users to maintain a specific hand posture, which does not conform to the natural state of the human hand. Users find it difficult to maintain this posture accurately and are prone to muscle fatigue during the process, further deteriorating alignment accuracy. The complete alignment process requires a static time of more than 5 seconds, forcing users to consciously remain still while waiting for the system to complete alignment. This passive waiting method results in a poor user experience, and in practical applications, the excessive preparation time reduces operational efficiency.

[0009] The sensor coordinate system and the human coordinate system are not precisely aligned. Due to the randomness of glove wearing, there is often an unknown deviation angle between the sensor coordinate system and the human coordinate system. Existing solutions fail to effectively establish a precise transformation relationship from the sensor coordinate system to the human coordinate system, resulting in systematic errors in subsequent attitude measurements. Summary of the Invention

[0010] To address the aforementioned problems, this invention provides a coordinate system alignment method for motion capture gloves designed for inertial sensor networks.

[0011] The above-mentioned technical objective of the present invention is achieved through the following technical solution: a coordinate system alignment method for motion capture gloves for inertial sensor networks, comprising the following steps:

[0012] Acceleration data of multiple inertial measurement unit sensor nodes arranged on the glove body are collected during the first alignment action, which is the action of the user placing his hand flat on a horizontal surface with his palm facing down, his fingers naturally extended and remaining still.

[0013] The Z-axis unit vector of the human body coordinate system is determined based on the acceleration data;

[0014] During the second alignment action, the angular velocity data of the sensor node is collected. The second alignment action is the action in which the user keeps his fist clenched and rotates his forearm as the axis.

[0015] The X-axis unit vector of the human coordinate system is determined based on the angular velocity data;

[0016] The Z-axis unit vector and the X-axis unit vector are subjected to Schmitt orthogonalization to obtain the corrected X-axis unit vector and Y-axis unit vector;

[0017] Based on the Z-axis unit vector, the corrected X-axis unit vector, and the Y-axis unit vector, construct a rotation transformation matrix from the sensor coordinate system to the human body coordinate system;

[0018] The original inertial data is transformed to a human body reference frame using the rotation transformation matrix.

[0019] Furthermore, the glove body is equipped with seven inertial measurement unit sensor nodes, including one back-of-hand node, one proximal joint node for each of the index, middle, ring, and little fingers, and one mid-joint node and one distal joint node for the thumb. All nodes are connected to the data processing unit via flexible cables.

[0020] Furthermore, determining the Z-axis unit vector of the human coordinate system based on the acceleration data includes: calculating the average acceleration vector of the acceleration data collected during the first alignment action within the sampling window; and determining the Z-axis unit vector using the opposite direction of the average acceleration vector.

[0021] Furthermore, the formula for calculating the average acceleration vector is as follows:

[0022] ;

[0023] in, For the number of samples, For the first The triaxial acceleration vector of the subsampled sample It is the average acceleration vector;

[0024] The formula for calculating the Z-axis unit vector is:

[0025] ;

[0026] in, The Z-axis unit vector, Let be the magnitude of the average acceleration vector.

[0027] Furthermore, determining the X-axis unit vector of the human coordinate system based on the angular velocity data includes: constructing a covariance matrix based on the angular velocity data collected during the second alignment action; performing eigenvalue decomposition on the covariance matrix to extract the eigenvector corresponding to the largest eigenvalue; and using the eigenvector as the X-axis unit vector.

[0028] Furthermore, the formula for calculating the covariance matrix is ​​as follows:

[0029] ;

[0030] in, Let covariance matrix be the variance matrix. For the first The triaxial angular velocity vector of the subsampled sample The number of data acquisitions during the second alignment action;

[0031] The eigenvalue decomposition formula is:

[0032] ;

[0033] in, It is an eigenvalue diagonal matrix. The eigenvector matrix;

[0034] The formula for calculating the X-axis unit vector is:

[0035] ;

[0036] The unit vector of the X-axis in the human coordinate system. This is the eigenvector corresponding to the largest eigenvalue.

[0037] Furthermore, the Z-axis unit vector and the X-axis unit vector are subjected to Schmitt orthogonalization to obtain the corrected X-axis unit vector and Y-axis unit vector, including: projecting the X-axis unit vector onto a plane perpendicular to the Z-axis unit vector and normalizing it to obtain the corrected X-axis unit vector; calculating the cross product of the Z-axis unit vector and the corrected X-axis unit vector to obtain the Y-axis unit vector.

[0038] Furthermore, the formula for calculating the corrected X-axis unit vector is as follows:

[0039]

[0040] in, The corrected X-axis unit vector. The X-axis unit vector, The Z-axis unit vector;

[0041] The formula for calculating the Y-axis unit vector is:

[0042] ;

[0043] in, It is the unit vector along the Y-axis.

[0044] Furthermore, the projection components of the corrected X-axis unit vector, Y-axis unit vector, and Z-axis unit vector along each axis of the sensor coordinate system are arranged in columns to construct a rotation transformation matrix from the sensor coordinate system to the human body coordinate system, including:

[0045] The rotation transformation matrix is ​​constructed by using the projection components of the corrected X-axis unit vector onto the three axes of the sensor coordinate system as the first column, the projection components of the Y-axis unit vector onto the three axes of the sensor coordinate system as the second column, and the projection components of the Z-axis unit vector onto the three axes of the sensor coordinate system as the third column. :

[0046] ;

[0047] in, , , The corrected X-axis unit vector is projected onto the three axes of the sensor coordinate system as its projection components. , , These are the projection components of the Y-axis unit vector onto the three axes of the sensor coordinate system. , , The Z-axis unit vector is projected onto the three axes of the sensor coordinate system as its projection components.

[0048] The process of transforming the original inertial data to a human body reference frame using a rotation transformation matrix includes:

[0049] Multiply the rotation transformation matrix by the original inertial data in the sensor coordinate system to obtain the inertial data in the human body reference frame.

[0050] In summary, the present invention has the following beneficial effects:

[0051] This invention completely eliminates the dependence on the Earth's rotation and uses the significant angular velocity changes generated by active rotational motion for principal component analysis (PCA), which greatly improves signal strength, significantly enhances alignment stability, and controls alignment error within 1-2 degrees.

[0052] The first and second alignment actions of this invention are both performed in a natural and comfortable posture, which greatly reduces the requirements for alignment posture and significantly improves the user experience.

[0053] This invention utilizes the strong inertial effect generated by active rotation to accurately extract the main motion direction through principal component analysis, improving the alignment accuracy by more than 50% compared to existing technologies.

[0054] This invention can complete alignment using only accelerometer and gyroscope data, without relying on geomagnetic sensors, thus having stronger resistance to electromagnetic interference and a wider range of applications.

[0055] This invention controls the entire alignment process to within 5 seconds, significantly improving operational efficiency. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the composition of an inertial sensor network motion capture glove system according to an embodiment of this application. Detailed Implementation

[0057] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0058] like Figure 1 As shown in the illustration, this application discloses an inertial sensor network motion capture glove. The glove includes a glove body and multiple inertial measurement unit (IMU) sensor nodes arranged on the glove body. Each sensor node includes a three-axis gyroscope and a three-axis accelerometer. Specifically, one back-of-hand node is arranged on the back of the hand to measure the overall posture of the palm; one sensor node is arranged on each of the index, middle, ring, and little fingers, located at the proximal joints of each finger; and two sensor nodes are arranged on the thumb, located at the mid-thumb joint and distal-thumb joint, respectively. A total of seven sensor nodes are provided, and all nodes are connected to a data processing unit via flexible cables.

[0059] This application also discloses a coordinate system alignment method for motion capture gloves based on inertial sensor networks, comprising the following steps:

[0060] S1. Collect acceleration data from multiple inertial measurement unit sensor nodes arranged on the glove body during the first alignment action.

[0061] The first alignment action involves the user placing their hand flat on a horizontal surface, palm down, fingers naturally extended, and remaining still. In one embodiment, the user maintains this still state for 3 seconds. In this state, the specific force sensed by the accelerometer is primarily due to gravitational acceleration. Composition. Due to the geometric relationship of the palm facing down, the direction of gravity physically indicates the negative direction of the Z-axis of the human body coordinate system.

[0062] Let the acceleration data sequence acquired during the first alignment action be... ,in The number of samples (e.g., 300 times). For the first The three-axis acceleration vectors are sampled once. The average acceleration within the sampling window can then be calculated. :

[0063] ;

[0064] Then, the unit vector of the Z-axis of the human coordinate system is determined using the opposite direction of the gravity vector. :

[0065] ;

[0066] in, Let be the magnitude of the average acceleration vector.

[0067] S2. Collect the angular velocity data of the sensor nodes during the second alignment action.

[0068] The second alignment action involves the user maintaining a clenched fist and rotating their forearm around an axis. In one embodiment, the rotation angle is 90 degrees, taking 2 seconds. During this time, the hand sensors rotate along the forearm's aligning axis. According to rigid body kinematics, the direction in which the angular velocity vector signal collected by the gyroscope has the strongest characteristic indicates the direction of the forearm's rotation axis, which is defined as the X-axis (heading axis) of the human coordinate system.

[0069] Let the angular velocity sequence acquired during the second action be... Then the covariance matrix can be constructed: Calculate angular velocity covariance matrix To characterize energy distribution:

[0070] ;

[0071] in, For the first The triaxial angular velocity vector of the subsampled sample This refers to the number of data acquisitions during the second alignment action.

[0072] Principal component extraction using PCA was used to analyze the covariance matrix. Perform eigenvalue decomposition:

[0073] ;

[0074] in, It is an eigenvalue diagonal matrix. Given an eigenvector matrix, extract the largest eigenvalue. corresponding feature vector Using the main rotation direction, the feature vector is taken as the unit vector of the X-axis in the human coordinate system. :

[0075] .

[0076] S3. Perform Schmitt orthogonalization on the Z-axis unit vector and the X-axis unit vector to obtain the corrected X-axis unit vector and Y-axis unit vector.

[0077] Due to measurement errors, the initially determined and It may not be perfectly orthogonal. To construct a rigorous spatial rectangular coordinate system, using the highest precision... The X-axis unit vector is corrected using the Z-axis as a reference. The X-axis unit vector is projected onto a plane perpendicular to the Z-axis unit vector and normalized to obtain the corrected X-axis unit vector. :

[0078] ;

[0079] in, The corrected X-axis unit vector. The X-axis unit vector, The Z-axis unit vector.

[0080] Both the human coordinate system and the sensor coordinate system follow the right-hand coordinate rule. Therefore, the Y-axis unit vector is obtained by calculating the cross product of the Z-axis unit vector and the corrected X-axis unit vector. :

[0081] ;

[0082] in, The unit vector along the Y-axis, thus forming a structure composed of... The standard orthogonal basis is formed.

[0083] S4. Arrange the projection components of the corrected X-axis unit vector, Y-axis unit vector, and Z-axis unit vector along each axis of the sensor coordinate system in columns to construct a rotation transformation matrix from the sensor coordinate system to the human body coordinate system, including:

[0084] Construct the rotation transformation matrix from the sensor coordinate system to the human body coordinate system. Its matrix elements consist of the projection components of each axis of the human body coordinate system onto the sensor axis. The projection components of the corrected X-axis unit vector onto the three axes of the sensor coordinate system are used as the first column of the rotation transformation matrix; the projection components of the Y-axis unit vector onto the three axes of the sensor coordinate system are used as the second column; and the projection components of the Z-axis unit vector onto the three axes of the sensor coordinate system are used as the third column.

[0085] ;

[0086] in, , , The corrected X-axis unit vector is projected onto the three axes of the sensor coordinate system as its projection components. , , These are the projection components of the Y-axis unit vector onto the three axes of the sensor coordinate system. , , The Z-axis unit vector is the projection component of the three axes of the sensor coordinate system.

[0087] S5. Transform the original inertial data to the human body reference frame using the aforementioned rotation transformation matrix:

[0088] ;

[0089] Inertial data in the human body frame of reference. This provides the raw inertial data in the sensor coordinate system. This step eliminates systematic measurement bias caused by the randomness of the glove wearing position.

[0090] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A coordinate system alignment method for motion capture gloves based on inertial sensor networks, characterized in that, Includes the following steps: Acceleration data of multiple inertial measurement unit sensor nodes arranged on the glove body are collected during the first alignment action, which is the action of the user placing his hand flat on a horizontal surface with his palm facing down, his fingers naturally extended and remaining still. The Z-axis unit vector of the human body coordinate system is determined based on the acceleration data; During the second alignment action, the angular velocity data of the sensor node is collected. The second alignment action is the action in which the user keeps his fist clenched and rotates his forearm as the axis. The X-axis unit vector of the human coordinate system is determined based on the angular velocity data; The Z-axis unit vector and the X-axis unit vector are subjected to Schmitt orthogonalization to obtain the corrected X-axis unit vector and Y-axis unit vector; Based on the Z-axis unit vector, the corrected X-axis unit vector, and the Y-axis unit vector, construct a rotation transformation matrix from the sensor coordinate system to the human body coordinate system; The original inertial data is transformed to a human body reference frame using the rotation transformation matrix.

2. The coordinate system alignment method for a motion capture glove for inertial sensor networks according to claim 1, characterized in that: The glove body is equipped with seven inertial measurement unit sensor nodes, including one back-of-hand node, one proximal joint node for each of the index, middle, ring, and little fingers, and one mid-joint node and one distal joint node for the thumb. All nodes are connected to the data processing unit via flexible cables.

3. The coordinate system alignment method for a motion capture glove for inertial sensor networks according to claim 1, characterized in that, Determining the Z-axis unit vector of the human body coordinate system based on the acceleration data includes: Calculate the average acceleration vector of the acceleration data acquired during the first alignment action within the sampling window; The Z-axis unit vector is determined by using the opposite direction of the average acceleration vector.

4. The coordinate system alignment method for a motion capture glove for inertial sensor networks according to claim 3, characterized in that, The formula for calculating the average acceleration vector is: ； in, For the number of samples, For the first The triaxial acceleration vector of the subsampled sample It is the average acceleration vector; The formula for calculating the Z-axis unit vector is: ； in, The Z-axis unit vector, Let be the magnitude of the average acceleration vector.

5. The coordinate system alignment method for a motion capture glove for inertial sensor networks according to claim 1, characterized in that, Determining the X-axis unit vector of the human coordinate system based on the angular velocity data includes: A covariance matrix is ​​constructed based on the angular velocity data collected during the second alignment action; The covariance matrix is ​​subjected to eigenvalue decomposition to extract the eigenvector corresponding to the largest eigenvalue; The feature vector is used as the X-axis unit vector.

6. The coordinate system alignment method for a motion capture glove for inertial sensor networks according to claim 5, characterized in that, The formula for calculating the covariance matrix is ​​as follows: ； in, Let covariance matrix be the variance matrix. For the first The triaxial angular velocity vector of the subsampled sample The number of data acquisitions during the second alignment action; The eigenvalue decomposition formula is: ； in, It is an eigenvalue diagonal matrix. The eigenvector matrix; The formula for calculating the X-axis unit vector is: ； The unit vector of the X-axis in the human coordinate system. This is the eigenvector corresponding to the largest eigenvalue.

7. The coordinate system alignment method for a motion capture glove for inertial sensor networks according to claim 1, characterized in that, The Z-axis unit vector and the X-axis unit vector are subjected to Schmitt orthogonalization to obtain the corrected X-axis unit vector and Y-axis unit vector, including: The X-axis unit vector is projected onto a plane perpendicular to the Z-axis unit vector and normalized to obtain the corrected X-axis unit vector; The cross product of the Z-axis unit vector and the corrected X-axis unit vector is calculated to obtain the Y-axis unit vector.

8. A coordinate system alignment method for a motion capture glove for inertial sensor networks according to claim 7, characterized in that, The formula for calculating the corrected X-axis unit vector is as follows: ； in, The corrected X-axis unit vector. The X-axis unit vector, The Z-axis unit vector; The formula for calculating the Y-axis unit vector is: ； in, It is the unit vector along the Y-axis.

9. A coordinate system alignment method for a motion capture glove for inertial sensor networks according to claim 7, characterized in that, The projection components of the corrected X-axis unit vector, Y-axis unit vector, and Z-axis unit vector along each axis of the sensor coordinate system are arranged in columns to construct a rotation transformation matrix from the sensor coordinate system to the human body coordinate system, including: The rotation transformation matrix is ​​constructed by using the projection components of the corrected X-axis unit vector onto the three axes of the sensor coordinate system as the first column, the projection components of the Y-axis unit vector onto the three axes of the sensor coordinate system as the second column, and the projection components of the Z-axis unit vector onto the three axes of the sensor coordinate system as the third column. : ； in, , , The corrected X-axis unit vector is projected onto the three axes of the sensor coordinate system as its projection components. , , These are the projection components of the Y-axis unit vector onto the three axes of the sensor coordinate system. , , The Z-axis unit vector is projected onto the three axes of the sensor coordinate system as its projection components. The process of transforming the original inertial data to a human body reference frame using a rotation transformation matrix includes: Multiply the rotation transformation matrix by the original inertial data in the sensor coordinate system to obtain the inertial data in the human body reference frame.

Images