A data glove based on the skin surface capacitance of the back of a human hand

Abstract

The application discloses a data glove based on human back skin surface capacitance, which comprises a glove body and a watch type collecting unit arranged at a wrist, a flexible printed circuit board electrode array is arranged at the back position of the glove body, the electrode array forms a detection loop with the human body through a watch ground electrode, gesture is detected by using the capacitance change difference signal caused by the change of the electrode and skin contact area when the finger moves, and the motion capture of ten degrees of freedom is realized; the wrist posture quaternion of the application is obtained by a nine-axis inertial measurement unit as a reference zero point, the capacitance signal and the posture data are input into a recurrent neural network for coupling signal decoupling, sampling calibration is carried out by using a joint limiting mechanism, and finally the high-precision reconstruction of the hand posture is realized, and a long-term wearable, comfortable and convenient real scene hand motion data collecting scheme is provided for the embodied intelligent research.

Description

Technical Field

[0001] This invention relates to the field of sensor equipment technology, and in particular to a data glove based on the capacitance of the skin surface on the back of the human hand. Background Technology

[0002] Against the backdrop of the rapid development of embodied intelligence technology, human hand motion capture technology has become a key breakthrough for robots to learn human behavior. Currently, traditional data gloves mainly use bending sensors or magnetic sensors to detect finger posture. Both types of sensors need to be installed in key joints of the glove to reconstruct hand movements by sensing joint bending or changes in magnetic fields.

[0003] The aforementioned sensors are large in size and usually require complex mechanical structures or external magnetic field generating devices, resulting in bulky data gloves that significantly restrict natural hand movements when worn, making them difficult to use for extended periods in daily life. In addition, the high cost of the sensors and the complexity of system construction limit their application in continuous, large-scale real-world data acquisition and fail to meet the needs of embodied intelligence for massive hand movement samples.

[0004] Therefore, in response to the problems mentioned above, this invention proposes a data glove based on the capacitance of the skin surface on the back of the human hand. Summary of the Invention

[0005] To overcome the problems of existing data glove sensors being bulky, inconvenient to wear, and difficult to use for extended periods, this invention proposes a data glove based on the capacitance of the skin surface on the back of the human hand. By using flexible FPC electrodes to contact the skin on the back of the hand, gestures are detected by changes in capacitance. This achieves lightweight and comfortable wear, facilitating continuous collection of hand movement data in daily work and life, and providing rich real-world motion data for embodied intelligence learning.

[0006] The technical solution of the present invention is as follows: a data glove based on the capacitance of the skin surface of the back of the human hand, comprising a glove body and a watch-style acquisition unit disposed at the wrist. The glove body is made of flexible fabric material, and a flexible printed circuit board electrode array is disposed on its inner side corresponding to the position of the back of the human hand. The electrode array includes ten metal electrodes for direct contact with the skin surface of the back of the hand. The watch-style acquisition unit is disposed at the wrist and includes a capacitance acquisition circuit and a nine-axis inertial measurement unit. The watch-style acquisition unit is electrically connected to the electrode array through a flexible printed circuit board adapter.

[0007] The capacitance acquisition circuit forms a detection loop with the human body through the watch's ground electrode. It is used to collect the capacitance difference signal caused by the change in the contact area between the electrode and the skin on the back of the hand during finger movement. When the finger bends or moves horizontally, the contact area between the electrode and the skin at the corresponding position changes, resulting in a corresponding change in the capacitance value between the electrode and the human body. The capacitance acquisition circuit collects the capacitance change data of each electrode in real time.

[0008] The electrode array is specifically arranged as follows: the first to fifth electrodes are respectively located on the back of the hand at the metacarpophalangeal joints of the thumb to the little finger, for detecting the independent bending degree of freedom of the five fingers; the sixth electrode is located on the back of the hand between the first metacarpal bone and the base epiphysis; the seventh electrode is located on the back of the hand between the first metacarpal bone and the second metacarpal bone, for detecting the translational degree of freedom of the thumb; the eighth electrode is located on the side of the second proximal phalanx near the thumb; the ninth electrode is located on the back of the hand between the second metacarpal bone and the second middle phalanx, for detecting the translational degree of freedom of the index finger; and the tenth electrode is located at the center of the back of the hand, for detecting the forward and backward flexion and extension degree of freedom of the palm. The above electrode arrangement realizes the detection of ten degrees of freedom of hand movements, including the independent bending of the five fingers, the lateral and vertical swing of the index finger and thumb, and the forward and backward flexion and extension of the palm.

[0009] The nine-axis inertial measurement unit is used to obtain the wrist's posture quaternion with the wrist as the reference zero point, providing a global reference coordinate system for hand movements.

[0010] Since there is signal coupling between the translation electrode and the curvature electrode, decoupling processing is required. This invention uses a recurrent neural network to fuse the capacitance change difference signal with the attitude data of the nine-axis inertial measurement unit. By using the "gating" characteristic of the deep learning model, the coupling signal between the electrodes is decoupled, and a mapping relationship between the capacitance signal and the hand posture is established, finally outputting the complete posture and movement of the hand.

[0011] In addition, the present invention also includes a sampling calibration module, which uses a joint limiting mechanism to perform initial calibration or process calibration. The joint limiting includes a maximum finger bending angle of 120°, an index finger abduction and extension range of -15° to +15°, a thumb translation range of ±40°, an index finger rotation range of 0° to 30°, and a thumb rotation range of 0° to 60°. Through this joint limiting mechanism, rapid and convenient personalized calibration can be achieved.

[0012] The training data of the deep learning model is established by synchronously acquiring the output signal of the data glove and the hand image captured by the RealSense camera, thereby constructing a precise mapping relationship between the capacitance signal and the hand posture, thus achieving high-precision reconstruction of the hand movement process.

[0013] The beneficial effects of this invention are: 1. This invention replaces traditional bending and magnetic sensors with a detection principle based on the capacitance of the skin surface on the back of the human hand. Gesture detection is achieved by the capacitance change of the contact area between the flexible FPC electrode and the skin, avoiding the problems of large sensor size and complex structure. This results in a lightweight and highly comfortable data glove that can meet the long-term wearing needs in daily life and provides continuous data acquisition capabilities in real-world scenarios for embodied intelligence research.

[0014] 2. This invention utilizes the unique design of flexible FPC electrodes, enabling the electrodes to closely conform to the curved surface of the back of the hand. This not only ensures the stability of signal acquisition but also eliminates the foreign body sensation and skin pressure problems caused by rigid electrodes, significantly improving the wearing experience. This allows the data gloves to be worn for extended periods like ordinary gloves without affecting normal hand activities.

[0015] 3. This invention achieves the detection of ten degrees of freedom, including the bending of the five fingers, the lateral and vertical movement of the index finger and thumb, and the forward and backward flexion and extension of the palm, through the precise layout of ten electrodes on the back of the hand. While ensuring comprehensive detection, it avoids redundancy in the number of sensors and covers the main movement modes of daily hand activities with the simplest electrode configuration.

[0016] 4. This invention establishes a correspondence between capacitance signals and action angles by using a sampling calibration method based on joint limiting mechanisms and utilizing the inherent range of motion of human hand joints. This enables rapid, simple, and personalized calibration without the need for external equipment, effectively solving the problem of individual differences in the skin capacitance method and significantly improving the system's adaptability and detection accuracy.

[0017] 5. This invention integrates a nine-axis inertial measurement unit with a capacitance acquisition circuit, establishes a global coordinate system with the wrist as the reference zero point, and decouples the coupled signals using a recurrent neural network. This not only solves the signal aliasing problem between the translation electrode and the curvature electrode, but also achieves high-precision reconstruction of hand posture, providing high-quality input data for subsequent deep learning mapping. Attached Figure Description

[0018] Figure 1 The diagram shown illustrates the distribution of the ten electrodes of this invention. Figure 2 The diagram shown is a schematic representation of the electrode signal acquisition data of this invention. Figure 3 The diagram shown is a schematic representation of the hand reconstruction posture restoration of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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 some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figures 1-3 The present invention provides an embodiment: The data glove of this invention comprises two parts: a glove body and a watch-style data acquisition unit located at the wrist. The glove body is made of a flexible fabric material, which has good breathability and elasticity, can adapt to different hand shapes, and ensures comfort during long-term wear. A flexible printed circuit board electrode array is arranged on the inner side of the glove body corresponding to the back of the hand. This electrode array includes ten metal electrodes for direct contact with the skin surface of the back of the hand. These electrodes are made using flexible printed circuit board technology, which not only has good conductivity but also conforms to the curved structure of the back of the hand, avoiding the foreign body sensation and signal instability problems caused by rigid electrodes.

[0021] The watch-style data acquisition unit is located on the wrist, resembling a regular smartwatch for easy everyday wear. Internally, it integrates a capacitive data acquisition circuit and a nine-axis inertial measurement unit, electrically connected to an electrode array on the glove body via a flexible printed circuit board adapter. This separate design concentrates the acquisition circuitry on the wrist, reducing weight burden on the back of the hand and further improving wearing comfort. The watch-style unit also includes a built-in power module and data storage module, enabling it to operate independently for extended periods without external devices for data acquisition and storage.

[0022] To achieve accurate detection of multiple degrees of freedom of the hand, this invention designs the positions of ten electrodes. The first to fifth electrodes are respectively positioned on the back of the hand at the metacarpophalangeal joints of the thumb to the little finger, used to detect the independent bending degrees of freedom of the five fingers. When a finger bends, the skin at the metacarpophalangeal joint undergoes significant stretching and displacement, causing a change in the contact area between the electrode and the skin, resulting in a corresponding change in capacitance. This change has a certain mapping relationship with the finger bending angle. By collecting and analyzing this capacitance change, the bending state of each finger can be reconstructed.

[0023] The sixth electrode is positioned on the back of the hand between the first metacarpal bone and the basal epiphysis, and the seventh electrode is positioned on the back of the hand between the first and second metacarpal bones. These two electrodes are used together to detect the translational degree of freedom of the thumb. The thumb is the human finger with the largest range of motion and the most complex movements. In addition to flexion and extension, it also has obvious adduction, abduction, and opposition movements. By combining the sixth and seventh electrodes, the lateral and vertical movements of the thumb at the metacarpophalangeal joints can be captured, achieving accurate perception of the thumb's multi-directional movements.

[0024] The eighth electrode is positioned on the side of the second proximal phalanx near the thumb, and the ninth electrode is positioned on the back of the hand between the second metacarpal and the second middle phalanx. These two electrodes are used to detect the translational degree of freedom of the index finger. As the most active finger in daily operations, the lateral movement of the index finger is crucial for fine motor skills such as grasping or pinching. The placement of the eighth and ninth electrodes fully considers the movement characteristics of the index finger, effectively capturing the adduction and abduction movements of the index finger at the metacarpophalangeal joint.

[0025] The tenth electrode is positioned at the center of the back of the hand to detect the flexion and extension degrees of freedom of the palm. The flexion and extension movements of the palm involve the activity of the wrist and metacarpophalangeal joints and are the basis of all hand movements. The tenth electrode is located in the central area of ​​the back of the hand. When the palm flexes and extends, the skin in this area undergoes a wide range of stretching and contraction, generating a significant change in capacitance signal.

[0026] Through the arrangement of the ten electrodes described above, this invention achieves hand motion detection with ten degrees of freedom, specifically including: one independent bending degree of freedom for each of the five fingers (a total of five), two degrees of freedom for the index finger's lateral and vertical movements, two degrees of freedom for the thumb's lateral and vertical movements, and one degree of freedom for the palm's forward and backward flexion and extension, for a total of ten degrees of freedom. This degree of freedom configuration basically covers the main movement patterns of the hand in daily activities and can meet the needs of most motion capture applications.

[0027] The capacitance acquisition circuit forms a detection loop with the human body through the watch's ground terminal. This design makes full use of the conductivity of the human body itself. When the hand moves, the contact area between each electrode and the skin on the back of the hand changes, causing the capacitance value between the electrode and the human body to change accordingly. At this time, the capacitance acquisition circuit collects the capacitance value of each electrode in real time and converts it into a digital signal for subsequent processing.

[0028] There is an inherent signal coupling between the translation electrode and the bending electrode. For example, between the first electrode (thumb bending electrode) and the sixth and seventh electrodes (thumb translation electrode), since they are adjacent in anatomical position and share part of the skin area, when the thumb makes a compound movement, the capacitance signals of the three electrodes will change simultaneously, and there is a complex nonlinear relationship between the changes. This coupling phenomenon is an inherent characteristic of the skin surface capacitance detection method. If decoupling is not performed, it will be impossible to accurately separate the independent motion information of each degree of freedom.

[0029] This invention discovers that this coupling phenomenon has a "gating" characteristic. The gating characteristic means that the motion of certain degrees of freedom will modulate the capacitance detection of other degrees of freedom. For example, when the thumb is in different bending angles, the capacitance change amplitude corresponding to the same translational motion will be different. This modulation relationship is non-linear and difficult to handle by traditional linear decoupling methods.

[0030] To address the aforementioned coupling problem, this invention employs a recurrent neural network (RNN) to decouple the capacitive signal. The reason for choosing an RNN is that hand movement is a continuous time-series process; the current hand posture depends not only on the current sensor signal but also on the posture from the previous moment. The unique recurrent connection structure of an RNN makes it naturally suitable for processing this type of time-series data, enabling it to capture the dynamic characteristics of the movement process.

[0031] In implementation, the constructed RNN model comprises an input layer, several hidden layers, and an output layer. The input layer receives ten capacitive signals and attitude data from a nine-axis inertial measurement unit (IMU). The ten capacitive signals are normalized data obtained from preprocessing of the raw sampled values. The attitude data from the nine-axis IMU is represented as quaternions to provide a global attitude reference for the wrist. The hidden layers employ Long Short-Term Memory (LSTM) units or Gated Recurrent Units (GRUs). These unit structures effectively address the vanishing gradient problem in traditional RNNs and are suitable for processing long sequences of data. The output layer outputs the ten degrees of freedom angle values ​​of the hand at the current moment.

[0032] The training process of the model requires a large amount of synchronous data. This invention constructs a training dataset by synchronously acquiring the output signal of the data glove and the hand images captured by the RealSense camera. Specifically, the subject wears the data glove of this invention and performs various hand movements, while the RealSense camera records the actual hand posture from multiple angles. The precise hand posture parameters at each moment are extracted from the video using image processing algorithms and used as training labels. The synchronously acquired capacitance signals and IMU data are used as input features, and the posture parameters obtained from image processing are used as the expected output to supervise the training of the RNN model. After training, the model establishes a precise mapping relationship between capacitance signals and hand posture, and can reconstruct hand posture in real time from sensor signals.

[0033] In the signal decoupling process, the RNN model achieves automatic decoupling by learning the nonlinear mapping relationship between the coupled signal and the motion of each degree of freedom. The recurrent connections within the model enable it to use temporal information to distinguish the influence of different degrees of freedom motions on the same electrode signal. For example, when the thumb simultaneously bends and translates, the model can identify the components corresponding to bending and translation respectively based on the signal's change over time. Compared with traditional mathematical decoupling methods, this learning-based decoupling method has stronger adaptability and higher accuracy.

[0034] This invention designs a sampling calibration method based on a joint limiting mechanism to address the issues that different users have different hand shapes and skin characteristics, and the contact state between the electrode and the skin may also be different each time it is worn. This calibration method can be used for initial calibration when wearing it for the first time, as well as for periodic calibration during use, thereby ensuring the consistency of detection accuracy.

[0035] The joint limiting mechanism utilizes the inherent range of motion of the human hand joints. Specifically, the maximum bending angle of the fingers is generally considered to be 120°, at which point the fingers are clenched into a fist as much as possible; the abduction and extension range of the index finger is usually -15° to +15°, that is, the index finger can be retracted 15° towards the thumb and extended 15° towards the little finger; the translational range of the thumb is relatively large, at ±40°; the rotational degree of freedom of the index finger is 0° to 30°; and the rotational degree of freedom of the thumb is 0° to 60°.

[0036] The calibration process is as follows: The user performs several standard movements sequentially as prompted, including clenching the fist to its maximum extent, extending the palm to its maximum extent, abducting the index finger to its maximum extent, adducting the index finger to its maximum extent, and opposing the thumb to its maximum extent. During each movement, the system records the capacitance values ​​of each electrode and correlates them with known joint limit positions. In this way, the system establishes a correspondence between the dynamic range of the capacitance signal and the range of motion of each user, thereby achieving personalized calibration.

[0037] The advantage of this calibration method is that it is simple and quick, requiring no external measuring equipment. Users only need to complete a few natural movements according to the prompts to complete the calibration. At the same time, because it utilizes the physical limits of the joints, the calibration results have clear physiological significance, ensuring the accuracy and reliability of the calibration.

[0038] This invention provides an embodiment: This example designed a systematic experimental scheme and invited 10 healthy subjects (aged 25-35, half male and half female) to participate in the test. Each subject wore the data gloves of this invention and performed hand movements according to a preset set of movements. At the same time, an optical motion capture system was used as a reference standard to record the actual movement posture of the hand.

[0039] The experimental action set includes two categories: single-degree-of-freedom actions and compound actions. Single-degree-of-freedom actions include independent flexion and extension of each finger, independent lateral movement of the index finger and thumb, and independent flexion and extension of the palm, used to evaluate the detection accuracy of each degree of freedom. Compound actions include various common hand gestures, such as clenching a fist, extending a palm, pinching, grasping, and digital gestures, used to evaluate the overall performance of the system in real-world application scenarios.

[0040] (1) For each single-degree-of-freedom movement, the subject moves from one end of the joint limit to the other end, repeating 10 times. The root mean square error (RMSE) and Pearson correlation coefficient (r) between the angle detected by the present invention and the angle measured by the optical capture system are calculated. The results are shown in Table 1: Table 1. Test results of single-degree-of-freedom detection accuracy As shown in Table 1, the detection errors of each degree of freedom are all within 5°, and the correlation coefficients are all above 0.94, indicating that the present invention has high detection accuracy. The detection accuracy of the finger bending degree of freedom is better than that of the translation degree of freedom, which is related to the stronger and more stable capacitance change signal generated by the bending motion. The error of the translation degree of freedom is slightly larger, but it can still meet the needs of most application scenarios.

[0041] (2) For each compound action, this example designed 11 common gestures in daily life, including: clenched fist, extended palm, spread fingers, OK gesture, thumbs-up gesture, victory gesture, and number 1-5 gesture. Each subject repeated each gesture 10 times. The trained RNN model was used to recognize the gestures, and the recognition accuracy was calculated. The results are shown in Table 2. Table 2. Accuracy Test Results of Compound Action Recognition As shown in Table 2, the average recognition accuracy of the 11 gestures reached 95.3%. The recognition rates of single-finger gestures and large-amplitude gestures were relatively high, while the recognition rate of complex gestures was slightly lower but still remained above 92%. This indicates that the present invention can effectively recognize complex hand gestures and has good practical value.

[0042] (3) This experiment verifies the decoupling effect of the RNN model. A specific comparative experiment was designed in this invention. Subjects performed a compound movement of their thumb, simultaneously bending and translating. The original, undecoupled signal and the decoupled signal were recorded and compared with the actual angle measured by the optical capture system. The correlation coefficients between the original and decoupled signals and the true angle were calculated, and the results are shown in Table 3. Table 3 Comparison of Decoupling Effects As shown in Table 3, the original signal has a high correlation with multiple degrees of freedom, which shows obvious coupling phenomenon. After decoupling by the RNN model, the correlation between the bending signal and the translational true value is significantly reduced, and the correlation between the translational signal and the bending true value is also significantly reduced. At the same time, the correlation between each signal and the corresponding degree of freedom remains at a high level. This shows that the RNN model has successfully separated the coupled signals and effectively solved the signal aliasing problem when multiple degrees of freedom move at the same time.

[0043] (4) This experiment compared the detection accuracy of the system before and after calibration. After the subject first wore the gloves, the standard action set was executed directly without calibration, and the detection error was recorded. Then, the calibration was performed according to the calibration procedure, and the same action set was executed again, and the detection error was recorded. The error changes before and after calibration were compared. The results are shown in Table 4. Table 4 Comparison of errors before and after calibration As shown in Table 4, the detection errors of each degree of freedom were significantly reduced after calibration, with an improvement rate generally above 40%. This indicates that the calibration method based on joint limits can effectively eliminate the influence of individual differences and wearing differences, and greatly improve the detection accuracy and stability of the system.

[0044] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A data glove based on the capacitance of the skin surface on the back of the human hand, characterized in that, include: The glove body has a flexible printed circuit board electrode array on the inner side corresponding to the back of the human hand. The electrode array includes ten metal electrodes for contact with the skin surface of the back of the hand. A watch-style data acquisition unit, located on the wrist, includes a capacitive data acquisition circuit and a nine-axis inertial measurement unit. The watch-style data acquisition unit is electrically connected to the electrode array via a flexible printed circuit board adapter. Among them, the capacitance acquisition circuit forms a detection loop with the human body through the ground of the watch, which is used to collect the difference signal of capacitance change caused by the change of the contact area between the electrode and the skin on the back of the hand during finger movement. The electrode array layout includes bending detection electrodes located at the metacarpophalangeal joints of the five fingers on the back of the hand, lateral and vertical swing detection electrodes located at the base of the index finger and thumb, and palm flexion and extension detection electrodes located at the center of the palm, for detecting hand movements of ten degrees of freedom. The nine-axis inertial measurement unit is used to obtain the wrist's attitude quaternion with the wrist as the reference zero point; The capacitance change difference signal and the attitude data of the nine-axis inertial measurement unit are input into a recurrent neural network to decouple the coupling signals between the electrodes and map the hand posture and movement through a deep learning model.

2. The data glove based on the capacitance of the skin surface on the back of the human hand according to claim 1, characterized in that, The specific layout of the ten electrodes is as follows: The first to fifth electrodes are respectively placed on the back of the hand at the metacarpophalangeal joints of the thumb to the little finger, and are used to detect the bending freedom of each finger. The sixth electrode is located on the back of the hand between the first metacarpal bone and the base epiphysis, and the seventh electrode is located on the back of the hand between the first metacarpal bone and the second metacarpal bone. It is used to detect the translational degree of freedom of the thumb. The eighth electrode is located on the side of the second proximal phalanx near the thumb, and the ninth electrode is located on the back of the hand between the second metacarpal and the second middle phalanx, used to detect the translational degree of freedom of the index finger. The tenth electrode is positioned at the center of the back of the hand and is used to detect the forward and backward flexion and extension degrees of freedom of the palm.

3. A data glove based on the capacitance of the skin surface on the back of the human hand according to claim 2, characterized in that: The first to fifth electrodes are finger bending detection electrodes, the sixth to ninth electrodes are finger root translation detection electrodes, and the tenth electrode is palm forward and backward bending detection electrode.

4. A data glove based on the capacitance of the skin surface on the back of the human hand according to claim 1, characterized in that: The capacitance acquisition circuit forms a circuit with the human body through the watch's ground electrode. It utilizes the capacitance change caused by the change in the contact area between the electrode and the skin to achieve non-contact detection of finger bending, translation, and palm flexion and extension movements.

5. A data glove based on the capacitance of the skin surface on the back of the human hand according to claim 1, characterized in that: The electrodes are made of flexible printed circuit board material to adapt to the curved surface of the back of the hand.

6. A data glove based on the capacitance of the skin surface on the back of the human hand according to claim 1, characterized in that: The data glove also includes a sampling calibration module, which uses a joint limiting mechanism to perform initial or process calibration. The joint limiting mechanism includes a maximum finger bending angle of 120°, an index finger abduction and extension range of -15° to +15°, a thumb translation range of ±40°, an index finger rotation range of 0° to 30°, and a thumb rotation range of 0° to 60°.

7. A data glove based on the capacitance of the skin surface on the back of the human hand according to claim 1, characterized in that: The recurrent neural network is used to decouple the coupled signals between the curvature electrode and the translation electrode. The coupling exists between the first electrode and the sixth and seventh electrodes. Signal separation is achieved through the gating characteristics of the deep learning model.

8. A data glove based on the capacitance of the skin surface on the back of the human hand according to claim 1, characterized in that: The nine-axis inertial measurement unit is used to provide a reference zero point for wrist posture, and the capacitance change difference signal is used to provide relative motion information between the fingers and the palm. The two are fused and input into a deep learning model to reconstruct the complete hand posture.

9. A data glove based on the capacitance of the skin surface on the back of the human hand according to claim 1, characterized in that: The training data of the deep learning model is established by synchronously acquiring the output signal of the data glove and the hand image captured by the RealSense camera, thereby constructing a mapping relationship between the capacitance signal and the hand posture.

10. A data glove based on the capacitance of the skin surface on the back of the human hand according to claim 1, characterized in that, The detection of the ten degrees of freedom includes: one independent bending degree of freedom for each of the five fingers, totaling five; two degrees of freedom for the lateral and vertical movement of the index finger; two degrees of freedom for the lateral and vertical movement of the thumb; and one degree of freedom for the forward and backward flexion and extension of the palm.

Images