Glove gesture detection method based on palmar pressure and dorsal strain collaborative compensation and intelligent glove

Abstract

The application relates to a glove gesture detection method based on palmar pressure and dorsal strain collaborative compensation and an intelligent glove. The detection method comprises the following steps: acquiring initial angles of finger joints; determining pressure background values under corresponding hand gestures according to the initial angles, and correcting pressure signals according to the pressure background values to calculate actual contact pressures; correcting the angles of the finger joints to obtain corrected joint angles; comparing the corrected joint angles with current joint angles to obtain a change amount, and comparing the change amount with a first preset threshold; when the change amount does not meet a preset stop condition, recalculating the actual contact pressures and again correcting the joint angles; and outputting the actual contact pressures, the angles of the finger joints and the action state of the hand when the preset stop condition is met. The intelligent glove comprises a glove body, a flexible resistance pressure array layer, a liquid metal strain sensing layer, a synchronous acquisition module and a processor. The application can reduce detection errors.

Description

Technical Field

[0001] This invention relates to the field of smart glove technology, and in particular to a glove posture detection method and a smart glove based on the coordinated compensation of palmar pressure and dorsal strain. Background Technology

[0002] Smart gloves typically have a flexible resistive pressure array on the palm side to collect pressure signals from each array unit, in order to obtain the position and force distribution when the palm comes into contact with an external object; and a liquid metal strain sensing layer on the back of the hand to collect deformation signals to estimate the angle of each finger joint.

[0003] In existing technologies, flexible pressure sensors and flexible strain sensors are combined in the same glove, outputting pressure and bending information respectively. However, they typically operate independently, lacking an effective collaborative compensation mechanism between the palm-side pressure signal and the dorsal-side deformation signal. For flexible resistive pressure arrays that conform to the palm and fingers, even without external object contact, palm bending, finger flexion and extension, glove wrinkling, and wearing pre-tightening can alter the initial resistance of the array units. When using a fixed zero point or zeroing only in a flat palm posture, the array output generated by bending is easily misinterpreted as contact pressure; the same external pressure may also correspond to different initial resistance changes under different hand postures. Existing technologies do not obtain the corresponding pressure background value based on the current hand posture, resulting in spurious pressure caused by posture changes being mixed into the palm-side pressure signal. The dorsal liquid metal strain sensing layer is also affected by wearing position, glove slippage, material hysteresis, and reaction loads when grasping objects. Converting the deformation signal into joint angles based solely on a single mapping under non-contact conditions may produce different errors between empty-hand bending and grasping objects. Existing technologies do not utilize the distribution and magnitude of actual contact pressure to correct for joint angles, resulting in the inability to suppress interference from contact loads and wear slippage on the dorsal strain mapping.

[0004] Therefore, it is necessary to provide a new glove posture detection method and smart glove based on the coordinated compensation of palmar pressure and dorsal strain to solve the above-mentioned problems in the prior art. Summary of the Invention

[0005] The technical problem to be solved by this application is how to provide a glove posture detection method and a smart glove based on the coordinated compensation of palmar pressure and dorsal strain with small detection error.

[0006] To address the aforementioned technical problems, according to embodiments of this application, a glove posture detection method based on coordinated compensation of palmar pressure and dorsal strain is provided, comprising the following steps: synchronously acquiring pressure signals from each array unit on the palm side of the glove and deformation signals from the dorsal side of the glove; obtaining the initial angle of each finger joint based on the deformation signals; using the initial angle as the current joint angle; obtaining the pressure background value of the pressure signal under the corresponding hand posture based on the current joint angle; correcting the pressure signal based on the pressure background value; and calculating the actual contact pressure; wherein different hand postures have different pressure background values; obtaining contact features from the actual contact pressure, and calculating the contact features based on the deformation signals and the... The angles of each finger joint are corrected based on contact characteristics to obtain corrected joint angles. The corrected joint angles are compared with the current joint angles to obtain the change in joint state between two adjacent iterations. It is determined whether the change satisfies a preset stopping condition. If the preset stopping condition is not met, the corrected joint angle is used as the current joint angle. The pressure background value under the corresponding hand posture is updated according to the corrected joint angle, the actual contact pressure is recalculated, and the joint angles are corrected again until the preset stopping condition is met or the preset number of iterations is reached. The pressure background value, the actual contact pressure, the angles of each finger joint, and the hand movement state when the preset stopping condition is met are output.

[0007] According to an embodiment of this application, obtaining the pressure background value of the pressure signal under the corresponding hand posture based on the current joint angle includes: acquiring joint state reference values ​​and palm side array signals for different hand postures under the condition that the palm side of the glove does not bear external load; averaging the joint state reference values ​​under each hand posture to obtain a reference joint angle; averaging the palm side array signals under each hand posture to obtain a reference pressure value; calculating the angle difference between the current joint angle and the reference joint angle; obtaining the weight of the corresponding hand posture based on the angle difference; and weighting and summing the reference pressure values ​​according to the weights to obtain the pressure background value.

[0008] According to an embodiment of this application, the step of correcting the pressure signal based on the pressure background value and calculating the actual contact pressure includes: subtracting the pressure background value from the pressure signal to obtain the intermediate pressure value of each array unit; defining array units whose intermediate pressure values ​​are less than a second preset threshold as non-contact units; obtaining the common drift amount of the non-contact units; subtracting the common drift amount from the intermediate pressure value of each array unit, and setting the result that is less than zero to zero, to obtain the actual contact pressure.

[0009] According to an embodiment of this application, obtaining contact features from the actual contact pressure includes: dividing the palm side of the glove into multiple groups of determination areas; the determination areas include the fingertip area, finger pad area, palm area, thenar and hypothenar eminence areas, and palm heel area; grouping the actual contact pressure into different determination areas; taking the determination area in contact with an external object as the contact area; calculating the resultant pressure force, pressure center position, and effective contact area of ​​the contact area, and taking the resultant pressure force, pressure center position, and effective contact area as contact features.

[0010] According to an embodiment of this application, the step of correcting the angle of each finger joint based on the deformation signal and the contact feature to obtain the corrected joint angle includes: calculating the base angle of each finger joint based on the deformation signal; determining the contact constraint correction amount of each finger joint based on the contact area and the pressure magnitude of the contact area; calculating the load coupling correction amount of each finger joint based on the coupling relationship between the deformation signal and the contact feature; summing the base angle, the contact constraint correction amount, and the load coupling correction amount to obtain the corrected joint angle; if the summed angle exceeds the preset physiological activity range of the corresponding joint, then the boundary value of the range is taken as the corrected joint angle.

[0011] According to an embodiment of this application, the preset stopping condition includes: stopping the iteration when the change in joint state obtained from two adjacent iterations is less than a first preset threshold, and the current correction and the previous correction are in the same contact area; and outputting the last correction result and reducing the confidence level when the number of iterations reaches a preset upper limit and the preset stopping condition is still not met.

[0012] According to an embodiment of this application, obtaining the weight of the corresponding hand posture based on the angle difference includes: normalizing the angle difference to obtain a distance value; and determining the weight of each hand posture based on the distance value.

[0013] A smart glove for the aforementioned detection method, the smart glove comprising: a glove body; a flexible resistive pressure array layer disposed on the palm side of the glove body to collect pressure signals from each array unit; a liquid metal strain sensing layer disposed on the back side of the glove body to collect deformation signals; a synchronous acquisition module disposed on the wrist of the glove body and electrically connected to the flexible resistive pressure array layer and the liquid metal strain sensing layer respectively, to synchronously acquire the pressure signals and the deformation signals in corresponding sampling frames, and assign frame numbers and timestamps to the pressure signals and the deformation signals to align the pressure signals and the deformation signals in time; and a processor disposed on the wrist of the glove body, electrically connected to the synchronous acquisition module, and configured to: determine the initial angle of each finger joint based on the deformation signals; determine the pressure background value of the corresponding hand posture based on the initial angles; correct the pressure signals using the pressure background value to obtain the actual contact pressure; obtain contact characteristics and correct the joint state based on the actual contact pressure; feed back the corrected joint state to update the pressure background value, and iterate until a preset stopping condition is met.

[0014] According to an embodiment of this application, the flexible resistive pressure array layer includes multiple row electrodes and multiple column electrodes, each row electrode intersecting with each column electrode to form an array unit, and the resistance of each array unit changes when pressure is applied; a row readout circuit, located at the end of the row electrode, includes a shared row drive operational amplifier and a row selection switch group, the row selection switch group having a pair of drive gating switches and feedback gating switches for each row, the two switches being synchronously controlled by the same row selection signal, the sampling point of the feedback gating switch being located between the drive gating switch and the row electrode, the row drive operational amplifier being adjusted in a closed loop according to the sampling point to compensate for the on-state voltage drop of the drive gating switch; and a column readout circuit, located at the end of the column electrode, to read the pressure signal of each array unit.

[0015] According to an embodiment of this application, the liquid metal strain sensing layer includes multiple acquisition units, which cover the four proximal interphalangeal joints of the four fingers, at least one and at most four metacarpophalangeal joints of the four fingers, at least one and at most four distal interphalangeal joints of the four fingers, the metacarpophalangeal joint of the thumb, and the interphalangeal joint of the thumb; the four fingers include the index finger, middle finger, ring finger, and little finger; the multiple acquisition units cover the four finger gaps respectively located between the thumb and index finger, the index finger and middle finger, the middle finger and ring finger, and the ring finger and little finger.

[0016] By adopting the above technical solution, a pressure background value that varies with hand posture is established in the dorsal deformation signal, effectively eliminating the false pressure detected on the palm side due to changes in hand posture. At the same time, the angle of each finger joint is corrected by using contact features extracted from the actual contact pressure, suppressing the influence of contact load and wearing slippage on the mapping of the dorsal deformation signal. By feeding the corrected joint angle back to the step of determining the pressure background value, a two-way closed-loop iteration is formed to correct palm pressure with dorsal posture and correct dorsal posture with palm contact, so that the actual contact pressure and joint angle are consistent with each other. In addition, the liquid metal strain sensing layer simultaneously covers all four proximal interphalangeal joints, selected metacarpophalangeal joints, selected distal interphalangeal joints, thumb metacarpophalangeal joints and interphalangeal joints and four finger gaps, which can simultaneously describe finger flexion and extension and finger gap opening. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of a smart glove according to an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram showing the position of the flexible resistive pressure array layer on the palm side of a smart glove according to an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram showing the position of the liquid metal strain sensing layer on the back side of a smart glove according to an embodiment of the present invention.

[0020] Figure 4 This is a schematic diagram of a row readout circuit and a column readout circuit according to an embodiment of the present invention.

[0021] Figure 5 This is a schematic diagram of the connection method of a signal conditioning circuit according to an embodiment of the present invention.

[0022] Figure 6 This is a step diagram of a method for detecting the motion state of a smart glove according to an embodiment of the present invention.

[0023] Figure 7 This diagram illustrates the steps of determining whether a preset stopping condition is met, i.e., the subsequent processing method, according to an embodiment of the present invention.

[0024] Figure label: 100. Glove body; 200. Flexible resistive pressure array layer; 210. Row electrode; 220. Column electrode; 230. Array unit; 240. Row readout circuit; 241. Row drive operational amplifier; 242. Row selection switch group; 250. Column readout circuit; 300. Liquid metal strain sensing layer; 310. Acquisition unit; 320. Proximal interphalangeal joint; 330. Metacarpophalangeal joint; 340. Distal interphalangeal joint; 350. Thumb metacarpophalangeal joint; 360. Thumb interphalangeal joint; 370. Finger gap; 400. Synchronous acquisition module; 500. Processor; 600. Analog-to-digital converter; 700. Parameter storage module; 800. Communication module. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, 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. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, but does not exclude other elements or objects.

[0026] The following is in conjunction with the appendix Figures 1-5 The specific embodiments of the present invention will be further described in detail below.

[0027] The present invention provides a glove posture detection method and a smart glove based on palmar pressure and dorsal strain collaborative compensation. The glove posture detection method is used to detect the action state of the smart glove, such as actual contact pressure, angle of each finger joint and hand action state. The smart glove includes a glove body 100, a flexible resistive pressure array, a liquid metal strain sensing layer 300, a synchronous acquisition module 400 and a processor 500.

[0028] In some embodiments, the glove body 100 includes a palm portion, a finger portion, and a wrist portion. The finger portion and wrist portion are respectively located at opposite ends of the palm portion. The user inserts their hand from the wrist portion, placing their fingers inside the finger portion, their palm inside the palm portion, and their wrist inside the wrist portion, thereby completing the wearing of the glove body 100. The finger portion includes the thumb, index finger, middle finger, ring finger, and little finger. The glove body 100 also has a palm side and a back side. The palm side of the glove is the side facing the palm when worn, and it is used with a flexible resistive pressure array for contact with external objects and sensing pressure. The back side of the glove is the side facing the back of the hand when worn, and it is used with a liquid metal strain sensor for detecting deformation of the finger joints.

[0029] In some embodiments, a flexible resistive pressure array layer 200 is disposed on the palm side of the glove body 100 to collect pressure signals from each array unit 230. Specifically, the flexible resistive pressure array layer 200 is disposed on the palm surface of the glove body 100 and the fingertip area where contact needs to be sensed. The flexible resistive pressure array layer 200 is a row and column array with multiple rows and columns of array units 230. The number of rows and columns of the array is determined according to the glove size and the requirements of partition wiring. It contains at least two rows and at least two columns. In this embodiment, it is preferably thirty-two rows and thirty-two columns, forming a total of one thousand and twenty-four array units 230. The array units 230 are used to generate resistance changes when the palm or fingertip contacts an external object, thereby facilitating the collection of pressure signals.

[0030] In some specific embodiments, the flexible resistive pressure array layer 200 includes multiple row electrodes 210 and multiple column electrodes 220. Each row electrode 210 intersects with each column electrode 220 to form an array unit 230. The resistance of each array unit 230 changes when pressure is applied. A row readout circuit 240, located at the end of the row electrode 210, includes a shared row drive operational amplifier 241 and a row selection switch group 242. The row selection switch group 242 sets a pair of drive gating switches and feedback gating switches for each row. The two switches are synchronously controlled by the same row selection signal. The sampling point of the feedback gating switch is located between the drive gating switch and the row electrode 210. The row drive operational amplifier 241 is adjusted in a closed loop according to the sampling point to compensate for the on-state voltage drop of the drive gating switch. A column readout circuit 250, located at the end of the column electrode 220, cooperates with the row readout circuit 240 to read the pressure signal of each array unit 230. Specifically, the row drive operational amplifier 241 has an inverting input terminal; the row drive reference is denoted as... The column adjustment reference is denoted as The row-driven benchmark is denoted as The column adjustment reference is denoted as the reference set potential provided to the row drive operational amplifier 241. The column readout circuit 250 is used to adjust the column reference potential. The output of the row drive operational amplifier 241 is connected to the selected row electrode 210 through a drive gating switch. Due to the on-resistance of the drive gating switch, a voltage drop is generated when current flows through it, causing the actual potential of the selected row electrode 210 to be lower than the output potential of the row drive operational amplifier 241. Therefore, the feedback gating switch returns the actual potential of the selected row electrode 210 to the inverting input of the row drive operational amplifier 241. The row drive operational amplifier 241 is denoted according to the row drive reference. The difference between the output potential and the inverting input potential adjusts the output to compensate for the on-state voltage drop of the drive gating switch, thereby stabilizing the actual potential of the selected row electrode 210 at the row drive reference. The selected row electrode 210 is stabilized by the closed loop at the row drive reference, denoted as... The column readout node is stable at the column adjustment reference, denoted as... Furthermore, under the condition that the weak current of the non-selected row leakage is negligible, the current of the selected array cell 230 approximately satisfies Its current amplitude is ,in The resistor of the selected array cell 230; the column readout circuit 250 will measure the current amplitude. After being converted to voltage, the signal is sent to an analog-to-digital converter 600 for digitization. The analog-to-digital converter 600 converts the analog voltage output from the column readout circuit 250 into a digital signal; the column adjustment reference is denoted as... The driving reference is denoted as All are set within the linear output range of the operational amplifier; each row is selected and each column is read in parallel. After all rows are selected in sequence, a complete frame of pressure signal is obtained.

[0031] In some embodiments, a liquid metal strain sensing layer 300 is disposed on the back side of the glove body 100 to collect deformation signals. Specifically, the liquid metal strain sensing layer 300 is made of a flexible material, or more specifically, the liquid metal strain sensing layer 300 is an integrated flexible layer or a structure with a common flexible tail end. The liquid metal is sealed in an elastomer microchannel or a conductive channel covered by an elastic encapsulation. Specifically, the elastomer microchannel is a cavity structure processed on a flexible substrate, and the liquid metal fills the cavity. When the finger is bent or the finger gap 370 is opened, the cavity deforms, causing the resistance of the liquid metal to change. The elastic encapsulation is an elastomer film covering the surface of the conductive channel to prevent liquid metal leakage and ensure the flexibility and extensibility of the conductive channel. Multiple acquisition units 310 are formed in the liquid metal strain sensing layer 300 to generate resistance changes when the finger is bent or the finger gap 370 is opened, so as to collect deformation signals.

[0032] In some specific embodiments, the liquid metal strain sensing layer 300 includes a plurality of acquisition units 310, which cover the four proximal interphalangeal joints 320 of the four fingers, at least one and at most four metacarpophalangeal joints 330 of the four fingers, at least one and at most four distal interphalangeal joints 340 of the four fingers, the metacarpophalangeal joint of the thumb 350 and the interphalangeal joint of the thumb 360; the four fingers include the index finger, middle finger, ring finger and little finger; the plurality of acquisition units 310 cover the four finger gaps 370 respectively spanning between the thumb and index finger, the index finger and middle finger, the middle finger and ring finger, and the ring finger and little finger. Specifically, each acquisition unit 310 of the liquid metal strain sensing layer 300 extends to the common flexible tail end of the wrist via an independent lead. Each acquisition unit 310 and its connecting lead constitute an independent detection channel. Each detection channel is connected to the analog-to-digital converter 600. The number of channels K of the liquid metal strain sensing layer 300 satisfies K=4+nM+nD+2+4=10+nM+nD, where nM is the number of selected metacarpophalangeal joints 330 (the joints connecting the fingers to the palm), nD is the number of selected distal interphalangeal joints 340 (the joints of the fingers furthest from the palm), and 1≤nM≤4, 1≤nD≤4, that is, 12≤K≤ 18; The four proximal interphalangeal joints 320 (the joints between the distal interphalangeal joint 340 and the metacarpophalangeal joint 330) of the non-thumb are four mandatory channels, the metacarpophalangeal joint 330 and the interphalangeal joint of the thumb are two mandatory channels, and the acquisition units 310 of the four finger gaps 370 are four mandatory channels; Preferably, the liquid metal strain sensing layer 300 has seventeen independent detection channels; Each detection channel of the liquid metal strain sensing layer 300 is connected to a signal conditioning circuit and an analog-to-digital converter 600 (ADC), the signal conditioning circuit includes an operational amplifier and an adjustable reference provided by a digital-to-analog converter (DAC); the signal conditioning circuit includes pull-up resistors. With liquid metal resistance channel Formation of pressure divider nodes Under the connection method of grounding the liquid metal resistor channel, it satisfies When the non-inverting input of the operational amplifier is connected The inverting input is connected to the input resistor. Connect adjustable reference And through feedback resistor When connecting the output terminal, the output satisfies Processor 500 adjusts This ensures that the outputs of channels with different initial resistances all enter a preset linear window; when the smart glove is in a specified no-load position, the processor 500 searches the digital-to-analog converter code channel by channel and saves the zero-adjustment code that makes the output enter the linear window. And save the zero-point analog-to-digital conversion value corresponding to the code. During online measurement, the processor 500... , Calculate the liquid metal channel resistance based on the current analog-to-digital conversion value and channel circuit parameters. Then according to Forming the first The normalized form of the channel, where, The resistance is the current resistance, which is the actual resistance value of the k-th detection channel at time t. When the finger is bent or the gap between the fingers is opened, the liquid metal channel is stretched / compressed, and the resistance changes accordingly. The zero-point resistance is the initial resistance value of the k-th detection channel in an unloaded posture (i.e., a specified initial posture where no external force is applied to the hand and the hand is not bent). More specifically, the smart glove also includes a parameter storage module 700, which is electrically connected to the processor 500 and is used to store the reference joint angles, reference pressure values, and zero-adjustment codes of each detection channel for each calibration posture obtained during the calibration process. and the corresponding zero-point analog-to-digital conversion value (Analog-to-digital conversion value when input voltage is 0), zero-adjustment code and zero-point analog-to-digital conversion value Write to parameter storage module 700.

[0033] In some embodiments, the synchronous acquisition module 400 is located at the wrist of the glove body 100 and is electrically connected to the flexible resistive pressure array layer 200 and the liquid metal strain sensing layer 300, respectively, to synchronously acquire pressure signals and deformation signals in the corresponding sampling frames, and assign frame numbers and timestamps to the pressure signals and deformation signals to align the pressure signals with the deformation signals in time. Specifically, the synchronous acquisition module 400 uses a frame number shared by the palm side and the back side and a time reference shared by the palm side and the back side to assign a frame number and timestamp to each complete palm side scanning frame, and reads all detection channels in the scanning frame. When the palm side scanning time is not negligible, the detection channels are read once at the beginning and once at the end of the frame, and linear interpolation is performed according to the sampling time of each row to align the pressure signals with the deformation signals in time.

[0034] In some embodiments, the processor 500 is disposed at the wrist of the glove body 100 and electrically connected to the synchronous acquisition module 400 to output the pressure background value, actual contact pressure, angles of each finger joint, and hand movement state. Specifically, the processor 500 is configured to synchronously acquire the pressure signals of each array unit 230 of the flexible resistive pressure array layer 200 and the deformation signals of each detection channel of the liquid metal strain sensing layer 300 through the synchronous acquisition module 400; obtain the initial angle of each finger joint based on the deformation signals, use the initial angle as the current joint angle, and retrieve the reference joint angle of each calibrated posture from the parameter storage module 700. The pressure background value under the current posture is obtained by normalizing the distance and weighting the pressure signal based on the reference pressure value. The pressure background value and common drift are subtracted from the pressure signal to obtain the actual contact pressure. The angles of each finger joint are corrected according to the distribution and magnitude of the actual contact pressure to obtain the corrected joint angles. The corrected joint angles are used as the current joint angles to redetermine the pressure background value and recalculate the actual contact pressure. After bidirectional iterative collaborative compensation of back posture correction of palmar pressure and palmar contact correction of back posture until the preset stopping condition is met, the pressure background value, actual contact pressure, angles of each finger joint, and hand movement state are output. In some specific embodiments, the smart glove also includes a communication module 800, which is electrically connected to the processor 500 and is used to transmit the actual contact pressure, angles of each finger joint, and hand movement state to an external device. The communication module 800 is well known to those skilled in the art and will not be described in detail here.

[0035] The embodiments of this application disclose a method for detecting the motion state of a smart glove, which includes the following steps: S1. Synchronously acquire the pressure signals of each array unit 230 on the palm side of the glove and the deformation signals on the back side of the glove, and obtain the initial angles of each finger joint based on the deformation signals. Specifically, by synchronously acquiring the pressure signals of each array unit 230 on the palm side and the deformation signals on the back side of the glove, the data on both sides are brought to a unified time reference. The palm side pressure signal reflects the contact state between the palm and the external object, while the back side deformation signal reflects the bending state of the finger joints. Synchronizing the two can avoid time misalignment caused by hand movements. If the palm side data comes from before the fist is clenched and the back side data comes from after the fist is clenched, the two will cause errors due to mutual compensation. Therefore, the synchronization relationship is a prerequisite for ensuring the accuracy of subsequent iteration convergence. The initial angles are obtained based on the deformation signals, which will be described in detail later.

[0036] S2. Using the initial angle as the current joint angle, obtain the pressure background value of the pressure signal under the corresponding hand posture based on the current joint angle, correct the pressure signal based on the pressure background value, and calculate the actual contact pressure; wherein, different hand postures have different pressure background values; specifically, by using the initial angle as the current joint angle, obtaining the pressure background value under the corresponding hand posture based on the current joint angle, and subtracting the pressure background value from the pressure signal (details below), the false pressure detected on the palm side due to changes in hand posture can be eliminated, so that the same external pressure corresponds to a consistent pressure response under different hand postures, thereby obtaining the actual contact pressure generated only by external contact and improving the accuracy of contact pressure measurement.

[0037] S3. Obtain contact features from the actual contact pressure, and correct the angles of each finger joint according to the deformation signal and the contact features to obtain the corrected joint angles. Specifically, correcting the angles of each finger joint according to the distribution and magnitude of the actual contact pressure is to use the real contact information on the palm side to perform reverse constraint correction on the joint state inferred from the deformation signal on the back side. This can reduce the influence of contact load and wearing slippage on the strain mapping on the back side, making the output joint angles closer to the real hand posture, thereby improving the distinguishability between bare-handed actions, contact actions and grasping actions.

[0038] S4. Compare the corrected joint angle with the current joint angle to obtain the change in joint state between two adjacent iterations, and determine whether the change meets the preset stopping condition. Specifically, by comparing the change in joint state between two adjacent iterations with the preset stopping condition, it can be determined whether the joint state tends to stabilize between two iterations, ensuring the reliability of the output results.

[0039] S5. When the preset stopping condition is not met, the corrected joint angle is used as the current joint angle. The pressure background value under the corresponding hand posture is updated according to the corrected joint angle, the actual contact pressure is recalculated, and the joint angle is corrected again until the preset stopping condition is met or the preset number of iterations is reached. Specifically, when the change in joint state obtained from two adjacent iterations does not meet the preset stopping condition, that is, when it is greater than or equal to the first preset threshold, the corrected joint angle is used as the current joint angle to recalculate the actual contact pressure and the joint angle is corrected again, forming a two-way closed-loop iteration of back posture correcting palm pressure and palm contact correcting back posture, so that the pressure distribution and joint state gradually converge to a consistent result. In the prior art, palm pressure and back posture are output independently, which makes the false pressure generated by palm bending easily misjudged as real contact, while the back side is subjected to reaction load during gripping, causing the joint angle to be distorted. The two results are contradictory. This step eliminates the above contradiction by iteratively updating the pressure distribution and joint state to gradually converge to a consistent result.

[0040] S6. Output the pressure background value, actual contact pressure, angles of each finger joint, and hand movement state when the preset stopping condition is met. Specifically, when the preset stopping condition is met during iteration, the outputs consistent actual contact pressure, angles of each finger joint, and hand movement state. Due to bidirectional iterative compensation, the output pressure distribution has eliminated posture background interference, and the joint state has suppressed the influence of contact load, which can be directly used for accurate identification of hand movements.

[0041] In some embodiments, an offline calibration step is included before step S1 to establish contactless mapping and reference data for multiple hand postures. Under the condition that the palm side of the glove does not bear external load, the wearer maintains a specified initial posture, and the processor 500 completes the zeroing code and zero-point analog-to-digital conversion value calibration of each back channel. At the same time, the pressure signal of each array unit 230 of the flexible resistive pressure array layer 200 under the initial posture without external load is collected as the initial reading for electrical zeroing. This zeroing process is used to make the analog front end work in the linear range and does not replace the subsequent posture-conditional pressure background value calibration. Subsequently, under the condition that the palm side of the glove does not bear external load, the wearer is required to maintain multiple hand postures. These hand postures include palms together, palms open, small-angle (15-30°) flexion of each finger, medium-angle (30-75°) flexion, thumb abduction, and four-finger layered flexion without contacting the palm. For each hand posture, multiple sampling frames are maintained, and dorsal deformation signals, joint state reference values, and palmar array signals are acquired simultaneously. The joint state reference values ​​are the actual angle values ​​of each finger joint measured by mechanical limiting, and the palmar array signals are the raw electrical signals acquired by the flexible resistive pressure array layer 200 under this posture. The multiple sampling frames include transition frames where the hand is in an intermediate state of motion during posture switching. The data of the transition frames does not belong to the current calibration posture or the previous calibration posture, and these transition frames need to be discarded. After discarding the transition frames, the dorsal deformation center of the corresponding posture is calculated. Joint Status Center Palm-side unloaded mean matrix dorsal deformation center The average value of the deformation signals of each channel on the back side under this posture is used to calibrate the contactless mapping. Contactless mapping To establish the mapping relationship from deformation signals to joint states, the dorsal deformation center of each calibrated posture is used. With the corresponding joint state center As input-output sample pairs, they are fitted using piecewise linear lookup tables or polynomial regression. Specifically, piecewise linear lookup tables use the deformation center and joint state center of the calibrated posture as nodes, and the initial angle is obtained through interpolation during online measurement; polynomial regression fits the polynomial relationship between deformation and joint state using the least squares method; the joint state center... This represents the average value of the joint state reference values ​​under this posture, i.e., the reference joint angle; palmar unloaded mean matrix. The average pressure signal of each array unit 230 on the palm side under this posture is the reference pressure value. Through the above offline calibration, multiple reference joint angles and reference pressure values ​​under non-external contact postures are pre-established, providing a data basis for posture matching and background subtraction during online measurement. Using multiple hand postures instead of a single flat palm zero point can cover a variety of hand postures in actual use, thereby avoiding the problem of false pressure generated by a fixed zero point under different postures.

[0042] In some embodiments, the initial angle of each finger joint is obtained based on the deformation signal, including: based on non-contact mapping. Calculate the initial angle Where e(t) is the back-side deformation vector at time t, composed of the normalized deformation of each detection channel of the liquid metal strain sensing layer 300. This is the initial joint angle vector; in this embodiment, there is no contact mapping. Using a quadratic mapping, i.e. ,in , , This represents element-wise multiplication, where 'a' is the fundamental bias vector when the sensor remains unchanged. Let a represent a J-dimensional real vector space, where a is a column vector containing J real elements, each element corresponding to a basic bias of a joint state; This is the matrix showing the main linear relationship between deformation and joint states. Describe the space of real matrices with J rows and K columns, i.e. It is a matrix containing J×K real elements, used to describe the main linear mapping relationship between the deformation of each channel on the dorsal side and the joint state; The quadratic coefficient matrix is ​​used to compensate for the nonlinearity of the liquid metal channel; e is the back-side deformation vector; J is the number of joint states to be output; K is the number of back-side detection channels. The relevant parameters are determined by the calibration samples. That is to say, a is the basic bias when the sensor does not change, i.e., the zero-point offset of each joint. Describe the main linear relationship between deformation and joint state; This is used to compensate for the nonlinearity of the liquid metal channel. Through the above mapping, the initial angle of the hand joint can be quickly estimated from the deformation signal; at the same time, the quadratic mapping can compensate for the nonlinear characteristics of the liquid metal strain channel, improving the accuracy of joint angle estimation.

[0043] In some embodiments, obtaining the pressure background value of the pressure signal under the corresponding hand posture based on the current joint angle includes: collecting joint state reference values ​​and palm side array signals for different hand postures under the condition that the palm side of the glove does not bear external load; averaging the joint state reference values ​​under each hand posture to obtain a reference joint angle; averaging the palm side array signals under each hand posture to obtain a reference pressure value; calculating the angle difference between the current joint angle and the reference joint angle; obtaining the weight of the corresponding hand posture based on the angle difference; and weighting and summing the reference pressure values ​​according to the weights to obtain the pressure background value. The process of obtaining the weight of the corresponding hand posture based on the angle difference includes: normalizing the angle difference to obtain a distance value; determining the weight of each hand posture based on the distance value, where a larger distance value results in a smaller weight, and the sum of all weights is one. Specifically, because different joints have different ranges of motion—for example, the metacarpophalangeal joint 330° (MCP) may have a larger range, while the distal interphalangeal joint 340° (DIP) may have a smaller range—if the angle differences are directly added together, the joint with the larger range will unreasonably dominate the distance results. Therefore, it is necessary to normalize the angle differences to obtain distance values, so as to convert the differences between different joints to a comparable scale. The normalization process uses a joint state normalization matrix. Divide the angle difference of each joint by the range or standard deviation of the corresponding joint to obtain the joint state normalization matrix. This is a diagonal matrix, where the diagonal elements represent the range or standard deviation of each joint. The normalized differences between each joint are combined into a single distance value, calculated as follows: ,in for The inverse matrix, The L2 norm is represented by the square root of the sum of the squares of the normalized differences between the joints; the distance value is used to measure the closeness between the current hand posture and the postures of other hands.

[0044] More specifically, since calibration poses closer to the current pose should receive greater weights and those farther away should receive smaller weights, and the sum of all weights must be one, the following Gaussian weight calculation method is used as an implementation for determining weights based on distance values:

[0045] in, For the first The pose weights of each hand gesture; Current joint angle With the Reference joint angles for each hand posture Normalized distance between them; The scale is determined by the pose neighborhood, which dictates the range of poses to be referenced. Smaller sizes are more sensitive to recent posture changes and are primarily based on recent posture. A larger value allows for the reference of more postures simultaneously, resulting in smoother changes; This represents the total number of hand gestures. Indicates all Summing of individual hand poses; pose weights satisfy... That is, the sum of all pose weights is one.

[0046] The background pressure value corresponding to the current posture is:

[0047] in, For the first The baseline pressure value for each hand posture; Current joint angle The corresponding pressure background value. By combining the above distance values ​​and posture weights, the reference pressure values ​​of multiple hand postures are mixed according to their weights to obtain the continuously changing pressure background value under the current posture, avoiding the problem of inaccurate background subtraction caused by using only a single flat palm zero point.

[0048] In some embodiments, the pressure signal is corrected based on the pressure background value, and the actual contact pressure is calculated. This includes subtracting the pressure background value from the pressure signal to obtain the intermediate pressure value of each array unit 230; defining array units 230 with intermediate pressure values ​​less than a second preset threshold as non-contact units; obtaining the common drift amount of the non-contact units; subtracting the common drift amount from the intermediate pressure value of each array unit 230, and setting any result less than zero to zero to obtain the actual contact pressure. Specifically, the second preset threshold is determined based on the noise statistical characteristics of the flexible resistive pressure array under conditions without external contact. For example, it is the sum of the mean of the intermediate pressure values ​​under non-contact conditions and several times the standard deviation, to ensure that the true contact pressure is not missed and the common drift is not misjudged as contact. The second preset threshold is a contact candidate threshold used to determine whether the array unit 230 may not be in contact with an external object. The second preset threshold cannot be too large, otherwise the true small pressure will be regarded as non-contact, nor can it be too small, otherwise the drift will be regarded as true contact. In obtaining the common drift of uncontacted units, the common drift can be obtained from the residuals of multiple uncontacted units. The common drift can be obtained using the median, truncated mean, or other robust statistics. The common drift is used to compensate for the overall offset caused by temperature rise, slow material recovery, and common power source drift. Determining the common drift by the median of the intermediate pressure values ​​of uncontacted units avoids the drift result being skewed by a small number of misjudged contact points, thereby improving the calculation accuracy of the actual contact pressure. More specifically, the calculation process of the actual contact pressure is as follows: first, uncontacted candidate array units 230 are selected based on the comparison between the intermediate pressure value and the second preset threshold; then, the common drift is determined based on the median of the residuals of the uncontacted candidate array units 230; finally, the common drift is subtracted from the intermediate pressure value of each array unit 230, and negative values ​​are set to zero; uncontacted units are set to zero after subtracting the common drift, while contact units retain the pressure generated by actual contact after subtracting the common drift.

[0049] In some embodiments, obtaining contact characteristics from the actual contact pressure includes: dividing the palm side of the glove into multiple groups of determination areas; the determination areas include the fingertip area, fingertip area, palm area, thenar and hypothenar eminence areas, and palm heel area; grouping the actual contact pressure into different determination areas; taking the determination area in contact with an external object as the contact area; calculating the resultant pressure force, pressure center position, and effective contact area of ​​the contact area, and taking the resultant pressure force, pressure center position, and effective contact area as contact characteristics. Specifically, a hysteresis determination method with entry and exit thresholds is used to obtain the contact area. Hysteresis determination means that when the actual contact pressure rises from below the exit threshold to above the entry threshold, the contact area of ​​the determination area is set to a contact state; when the actual contact pressure drops from above the entry threshold to below the exit threshold, the contact area of ​​the determination area is set to a non-contact state. During the pressure change between the entry and exit thresholds, the contact area remains unchanged. The entry threshold is greater than the exit threshold, thereby avoiding repeated jumps in the contact area when the pressure fluctuates around a single threshold. The resultant pressure is the sum of all effective pressures in a certain contact area, reflecting the total force on that area. The pressure center position is the location where the contact pressure is concentrated in the area, reflecting the point of application of the force. The effective contact area is the number of array units 230 exceeding the entry threshold or the corresponding area, reflecting the contact range.

[0050] In some embodiments, the angles of each finger joint are corrected based on the deformation signal and the contact feature to obtain the corrected joint angle. This includes: calculating the base angle of each finger joint based on the deformation signal; determining the contact constraint correction amount for each finger joint based on the contact area and the pressure magnitude of the contact area; calculating the load coupling correction amount for each finger joint based on the coupling relationship between the deformation signal and the contact feature; summing the base angle, the contact constraint correction amount, and the load coupling correction amount to obtain the corrected joint angle; if the summed angle exceeds the preset physiological activity range of the corresponding joint, the boundary value of that range is taken as the corrected joint angle. Specifically, the contact constraint correction amount reflects the direct influence of the contact area and the pressure magnitude on the angles of each joint; the load coupling correction amount reflects the coupling compensation with different meanings under different contact loads for the same dorsal deformation, i.e., the deviation caused by the reaction load affecting the dorsal strain when grasping an object; the preset physiological activity range is the reasonable activity range of the corresponding joint, used to limit the correction result within an anatomically reasonable range. The joint angle correction can be divided into three parts. The first part is the base angle obtained from the dorsal deformation signal, which only considers the joint state obtained from the deformation of the back of the hand. The second part is a direct correction based on the palmar contact area and pressure magnitude; for example, stable fingertip contact increases the system correction by 3 degrees. The third part is a coupling correction that addresses the different meanings of the same back-of-the-hand deformation under different contact loads; for example, the load coupling term reduces the system correction by 1 degree. Finally, the three parts are summed. If the calculated proximal interphalangeal joint angle is 320° = 130°, but the physiological upper limit is set to 110°, then 110° is output. More specifically, the joint state correction formula is:

[0051] in, For the first The corrected joint angle obtained in the next iteration; The base value of the joint state obtained from the dorsal deformation signal is the base angle. For the contact constraint correction matrix, , express OK The space of real matrices of columns, i.e. For inclusion A matrix of real elements is used to describe the correction amount of a certain anatomical contact area to the relevant joint state. The contact feature vector includes the contact area, resultant pressure force, pressure center location, and effective contact area. The dimension of the contact feature; This is the load coupling correction matrix. , express OK The space of real matrices of columns, i.e. For inclusion A matrix of real elements is used to describe the change in the sensitivity of the liquid metal strain channel when the contact state changes. The coupled feature vector is obtained by combining the dorsal deformation and palmar contact information into the same region dimension one by one. For coupling feature dimensions; and The preset lower and upper limits of the physiological range of motion for the corresponding joints; This indicates that the results will be limited to a preset range of physiological activities.

[0052] Coupled feature vectors From the region mapping matrix and Construct, satisfy ,in, , This is the mapping matrix from the backside deformation to the region dimension. Describe the space of real matrices with rows D and columns K, i.e. It is a matrix containing D×K real elements, used to transform backside deformation to the region dimension; , This is a mapping matrix from contact features to region dimensions. Describe the space of real matrices with rows D and columns M, i.e. It is a matrix containing D×M real elements, used to transform contact features to the region dimension; and This is used to convert dorsal deformation and palmar contact information to the same regional dimension before combining them item by item. That is, the coupling feature first converts the dorsal deformation and palmar contact information to the same regional dimension, and then combines them item by item; for example, the index finger PIP deformation channel and the index finger pad contact force form a coupling feature, while the unrelated little finger channel and thumb contact do not need to be forcibly combined. and This can be understood as the correspondence between the sensor and the anatomical region. During firmware implementation, it's not necessarily necessary to save it as a large matrix; a fixed index and coefficient table can also be used. The parameters of the contactless mapping (i.e., the basic bias vector a, the principal linear relationship matrix...) and quadratic coefficient matrix The contact constraint correction matrix H, the load coupling correction matrix G, and the region mapping matrix are also included. and The parameters are obtained from calibration samples with known joint and contact states using regularized least squares method to avoid overfitting and improve generalization ability. By combining and limiting the base angle, contact constraint correction, and load coupling correction within the physiological range, the complementarity of dorsal deformation and palmar contact information can be fully utilized to suppress the error of a single sensor, so that the corrected joint angle simultaneously meets the requirements of anatomical rationality and measurement accuracy.

[0053] In some embodiments, a preset stopping condition is included: when the change in joint state obtained from two adjacent iterations is less than a first preset threshold, and the current correction and the previous correction are in the same contact area, the iteration is stopped; when the number of iterations reaches a preset upper limit and the preset stopping condition is still not met, the final correction result is output and the confidence level is reduced. Specifically, the first preset threshold is determined comprehensively based on the physiological range of motion of the corresponding joint and the measurement noise level; for example, for the proximal interphalangeal joint 320 with a range of 110°, the first preset threshold can be set to 1% to 5% of the range after normalization; the change in joint state obtained from two adjacent iterations is the normalized difference between the corrected joint angle and the current joint angle; the first preset threshold is the joint angle change threshold; the stopping condition requires that the change in joint state obtained from two adjacent iterations is less than the first preset threshold and the contact area obtained from the two corrections is simultaneously satisfied, in order to prevent pseudo-convergence where the angle converges but the contact area is still changing. In some specific embodiments, when a preset stopping condition is met, the final result is output; when the preset stopping condition is not met, the steps of correcting the pressure signal and revising the joint angle are repeated; when the preset stopping condition is not met even after the number of iterations reaches a preset upper limit, the final correction result is output and the confidence level is reduced; the specific logic is as follows: Figure 7 As shown.

[0054] More specifically, the normalized calculation method for the change in joint state obtained between two adjacent iterations is as follows: ,in Let Dq be the inverse of the joint state normalization matrix. The infinity norm represents the maximum change in all joints after normalization; the change in joint states obtained from two consecutive iterations is no greater than a first preset threshold. If the contact area obtained from the two consecutive corrections remains unchanged, the iteration stops. Specifically, the system compares the current joint state with the previous one; if the maximum change in all joints after normalization does not exceed a threshold, the joint state is considered basically stable. Furthermore, the contact area must remain unchanged between adjacent iterations; both conditions must be met before stopping. If the iteration count reaches the upper limit and stability is still not achieved, the system outputs the final result but lowers the confidence level of that frame. The iteration count is limited to two to five times to limit processing latency; when the preset iteration count is reached and the stopping condition is still not met, the final correction result is output with a lower confidence level, alerting the system that the data quality of that frame is poor and preventing the use of unstable results for subsequent action recognition.

[0055] In some embodiments, the output of the hand's movement state includes: determining the hand's movement state based on the final angles of each finger joint, the opening amount of the four finger gaps 370 (obtained from data collected by the acquisition units at the finger gaps 370, calculated in the same way as at the joints, and will not be elaborated here), the resultant pressure force of each contact area, the position of the pressure center, and the contact area. Specifically, the hand's movement states include opening the hand, clenching the fist, pinching, enveloping grasping, lateral pinching, and empty-hand flexion; the movement state is output after being input into a rule-determiner, an action template matcher, or a pre-trained classification model based on the feature vector formed by the final joint angles, the opening amount of the finger gaps 370, the resultant pressure force, the position of the pressure center, and the contact area. Taking empty-hand flexion as an example, the bending of the palm and fingers may cause the original pressure signal to show a non-zero response in the crease area; the dorsal deformation first determines the flexion posture, the pressure background value eliminates the corresponding crease response, and if there is no stable contact area in the actual contact pressure, it is determined to be empty-hand flexion rather than grasping. Taking the grasping of an object as an example, the actual contact pressure forms a stable contact area in the fingertips and palm. These contact features correct the back strain mapping through contact constraint correction and load coupling correction, so that the joint state deviation caused by glove slippage or object reaction load is suppressed. The corrected joint state is then used to update the pressure background value, thus forming a closed loop of back posture correcting palm pressure and palm contact correcting back posture, and finally outputting consistent actual contact pressure, joint angle and grasping action.

[0056] While embodiments of the present invention have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations fall within the scope and spirit of the invention as set forth in the claims. Furthermore, the invention described herein may have other embodiments and can be implemented or carried out in various ways.

Claims

1. A glove posture detection method based on coordinated compensation of palmar pressure and dorsal strain, characterized in that, Includes the following steps: The pressure signals of each array unit on the palm side of the glove and the deformation signals on the back side of the glove are collected simultaneously, and the initial angle of each finger joint is obtained based on the deformation signals. The initial angle is used as the current joint angle. The pressure background value of the pressure signal under the corresponding hand posture is obtained based on the current joint angle. The pressure signal is corrected based on the pressure background value, and the actual contact pressure is calculated. Different hand postures have different pressure background values. The contact characteristics are obtained from the actual contact pressure, and the angles of each finger joint are corrected according to the deformation signal and the contact characteristics to obtain the corrected joint angles; The corrected joint angle is compared with the current joint angle to obtain the change in the joint state obtained in two adjacent iterations, and it is determined whether the change satisfies the preset stopping condition. When the preset stopping condition is not met, the corrected joint angle is used as the current joint angle, the pressure background value under the corresponding hand posture is updated according to the corrected joint angle, the actual contact pressure is recalculated, and the joint angle is corrected again until the preset stopping condition is met or the preset number of iterations is reached. Output the pressure background value, the actual contact pressure, the angle of each finger joint, and the hand movement state when the preset stopping condition is met.

2. The glove posture detection method according to claim 1, characterized in that, The pressure background value of the pressure signal under the corresponding hand posture is obtained based on the current joint angle. include, Under the condition that the palm side of the glove does not bear external load, joint state reference values ​​and palm side array signals of different hand postures are collected; The reference values ​​of joint states under each hand posture are averaged to obtain the reference joint angle; the reference pressure value is obtained by averaging the palmar array signals under each hand posture. Calculate the angle difference between the current joint angle and the reference joint angle; The weight of the corresponding hand posture is obtained based on the angle difference; The pressure background value is obtained by summing the reference pressure values ​​according to the weights.

3. The glove posture detection method according to claim 1, characterized in that, The step of correcting the pressure signal based on the pressure background value and calculating the actual contact pressure includes, The intermediate pressure value of each array unit is obtained by subtracting the pressure background value from the pressure signal. The array unit whose intermediate pressure value is less than the second preset threshold is defined as an uncontacted unit; Obtain the common drift amount of the uncontacted units; The actual contact pressure is obtained by subtracting the common drift from the intermediate pressure value of each array unit and setting any result less than zero to zero.

4. The glove posture detection method according to claim 1, characterized in that, The step of obtaining contact characteristics from the actual contact pressure includes: The palm side of the glove is divided into multiple judgment areas; the judgment areas include the fingertip area, finger pad area, palm area, thenar and hypothenar eminence areas, and palm root area; The actual contact pressure is grouped according to different determination areas; The area in contact with an external object is defined as the contact area; the resultant pressure force, the position of the pressure center, and the effective contact area of ​​the contact area are calculated, and the resultant pressure force, the position of the pressure center, and the effective contact area are defined as contact features.

5. The glove posture detection method according to claim 4, characterized in that, The step of correcting the angles of each finger joint based on the deformation signal and the contact characteristics to obtain the corrected joint angles includes, Calculate the basic angle of each finger joint based on the deformation signal; Based on the contact area and the pressure magnitude in the contact area, determine the contact constraint correction amount for each finger joint; Based on the coupling relationship between the deformation signal and the contact feature, the load coupling correction amount for each finger joint is calculated; The corrected joint angle is obtained by summing the base angle, the contact constraint correction, and the load coupling correction. If the summed angle exceeds the preset physiological range of the corresponding joint, then the boundary value of that range is taken as the corrected joint angle.

6. The glove posture detection method according to claim 4, characterized in that, The preset stop conditions include: The iteration stops when the change in joint state between two consecutive iterations is less than the first preset threshold, and the current correction and the previous correction are in the same contact area. If the preset stopping condition is not met even after the number of iterations reaches the preset upper limit, the final correction result is output and the confidence level is reduced.

7. The glove posture detection method according to claim 2, characterized in that, The step of obtaining the weight of the corresponding hand posture based on the angle difference includes, The angle difference is normalized to obtain the distance value; The weights of each hand gesture are determined based on the distance values.

8. A smart glove, characterized in that, For implementing the detection method according to any one of claims 1-7, the smart glove comprises: The glove itself; A flexible resistive pressure array layer is disposed on the palm side of the glove body to collect pressure signals from each array unit; A liquid metal strain sensing layer is disposed on the back side of the glove body to collect the deformation signal; A synchronous acquisition module is located at the wrist of the glove body and is electrically connected to the flexible resistive pressure array layer and the liquid metal strain sensing layer, respectively, so as to synchronously acquire the pressure signal and the deformation signal in the corresponding sampling frame, and assign frame number and timestamp to the pressure signal and the deformation signal so that the pressure signal and the deformation signal are time-aligned. The processor, located at the wrist of the glove body, is electrically connected to the synchronous acquisition module and is configured to: determine the initial angle of each finger joint based on the deformation signal; determine the pressure background value of the corresponding hand posture based on the initial angle; correct the pressure signal using the pressure background value to obtain the actual contact pressure; obtain contact characteristics and correct the joint state based on the actual contact pressure; feed back the corrected joint state to update the pressure background value, and iterate until a preset stopping condition is met.

9. The smart glove according to claim 8, characterized in that, The flexible resistive pressure array layer includes, Multiple row electrodes and multiple column electrodes, each row electrode intersecting with each column electrode to form an array unit, and the resistance of each array unit changes when subjected to pressure; The row readout circuit, located at the end of the row electrode, includes a common row drive operational amplifier and a row selection switch group. The row selection switch group is configured with a pair of drive gating switches and feedback gating switches for each row. The two switches are synchronously controlled by the same row selection signal. The sampling point of the feedback gating switch is located between the drive gating switch and the row electrode. The row drive operational amplifier is adjusted in a closed loop according to the sampling point to compensate for the on-state voltage drop of the drive gating switch. A column readout circuit is located at the end of the column electrode to read out the pressure signal of each array unit.

10. The smart glove according to claim 8, characterized in that, The liquid metal strain sensing layer includes multiple acquisition units, which cover four proximal interphalangeal joints of four fingers, at least one and at most four metacarpophalangeal joints of the four fingers, at least one and at most four distal interphalangeal joints of the four fingers, the metacarpophalangeal joint of the thumb, and the interphalangeal joint of the thumb; the four fingers include the index finger, middle finger, ring finger, and little finger; The multiple acquisition units cover the four finger gaps respectively located between the thumb and index finger, the index finger and middle finger, the middle finger and ring finger, and the ring finger and little finger.

Images