Au-vg flexible strain sensor, preparation method and application thereof

Abstract

The application discloses an Au-VG flexible strain sensor, a preparation method and application thereof. The flexible strain sensor is based on gold modified vertical graphene (Au-VG). Under the condition that the original length is 3 cm and the strain range is 3.3% to 13.3% (Delta L=1 to 4 mm), the sensor can realize that the average GF is more than 240, the linearity R² is greater than 0.98, the surface resistance is less than or equal to 500 ohms / sq, the response time is less than 100 ms, and the resistance drift is less than 3% after 10,000 cycles, and excellent comprehensive performance is exhibited. The Au-VG sensor is integrated in a smart glove, and finally realizes high-precision classification of 26 English letter ASL gestures, up to 99.6%. Compared with existing recognition systems based on crack metal or carbon-based composite materials, the application has significant advantages in micro-strain recognition accuracy, power consumption control, device life and signal stability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of sensors, and particularly relates to an Au-VG flexible strain sensor, a preparation method and application thereof. BACKGROUND

[0002] With the continuous growth of global demand for barrier-free information technology, natural interaction systems and intelligent wearable devices, the demand for "high sensitivity, low power consumption, structural integration and mass production" of flexible wearable devices is becoming more and more urgent. In particular, the governments and international organizations actively promote barrier-free information technology, and relevant regulations and standards continue to be introduced; the global wearable device and unmanned aerial vehicle market is rapidly expanding, driving the demand for new high-performance sensing materials; and the fusion of some new conductive materials (such as vertical graphene) and micro-nano processing technology improves the performance and manufacturability of sensors; the important needs of special groups for high-precision wearable gesture recognition devices to assist communication even extend to consumer electronics, industrial control, medical rehabilitation and other wider fields.

[0003] At present, sign language recognition and natural gesture interaction technology is still in the research and development or initial commercialization stage, and there are problems such as low recognition accuracy, high delay and bulky equipment. Although some methods based on cameras and visual recognition have been put into the market, they rely on specific environments, are sensitive to occlusion, and have large computational complexity, making it difficult to popularize in practical applications. In contrast, solutions based on flexible strain sensors can work in non-visual environments and have stronger practicality and reliability. As a representative of emerging electronic materials, the commercialization of flexible sensors is still limited by unstable material performance and difficulty in mass production.

[0004] In recent years, flexible wearable strain sensors have attracted widespread attention as key sensing units for gesture recognition, human-computer interaction and health monitoring systems. Current research mainly focuses on material configuration design and conductive mechanism optimization, and typical technical routes include metal crack structure, laser-induced graphene (LIG), carbon nanotube / graphene composite conductive network, and three-dimensional vertical graphene (VG).

[0005] In the metal crack structure class, this type of scheme usually deposits a thin film of metal such as silver (Ag) or gold (Au) on the surface of an elastomer, and induces crack propagation during strain loading to achieve a strain resistance response. Its advantage is high initial sensitivity (GF can reach thousands), but the crack structure is not reversible, which leads to significant response attenuation after more than 1000 cycles, making it difficult to meet long-term use requirements and lacking stability.

[0006] Laser-induced graphene (LIG) can be obtained by laser direct etching of polyimide (PI), which is convenient to prepare and has good flexibility. The bionic hydrophobic LIG strain sensor reported in the existing research has a GF of 565 and has certain environmental adaptability. However, in the 0-4.5% micro-strain range, the change of ΔR / R0 is slow, the sensitivity is insufficient, and the surface resistance is high (>10 kΩ / sq), which limits its application in low-power precise sensing scenarios.

[0007] In addition, related research has adopted a flexible pressure sensor based on ion-composite nanofiber membranes to realize intelligent sign language recognition combined with a BPNN (back propagation neural network) model. Among the 650 collected data samples, the overall recognition accuracy of 24 letter gestures is 96.8%. However, this method performs poorly in distinguishing between the letters U and V, which is a shortcoming of the system. In practical applications, the recognition accuracy of sign language recognition systems is particularly critical, and the improvement of recognition accuracy is directly related to its practical value in communication assistance for deaf and hearing-impaired populations.

[0008] US 12025437 B2 Stretchable Strain Sensor Based on Vertical Graphene and Its Application discloses a VG-PDMS flexible strain sensor: this patent grows a vertical graphene (VG) network on the surface of PDMS through the PECVD process, constructs a three-dimensional conductive framework, realizes linear response and good cycle stability (>10,000 times) under high strain (up to 50%). The device shows good mechanical compliance and overall conductivity, suitable for wearable scene applications.

[0009] However, this scheme has the following shortcomings: the initial surface resistance of the material is still high (1-5 kΩ / sq), which is not conducive to low-power electronic systems; the GF in the micro-strain range (0-10%) is low, making it difficult to meet the requirements of high-resolution gesture or physiological signal recognition; no crack structure or sensitization mechanism is introduced, and the response sensitivity has limited room for improvement.

[0010] A VG / CNT / PDMS sandwich structure is reported in the prior art, which constructs a multilayer conductive structure cross-linked by wrinkle-type vertical graphene (VG) and carbon nanotubes (CNT). The device exhibits an extremely high strain coefficient (GF>1300) in the 60-100% strain range, while the linearity is good (R² ≈ 0.98), and has strong tensile strength and structural continuity. However: GF ≈ 45 only in the 0-10% range, with limited micro-strain detection capability; the proportion of CNT composition has a great influence on the interface stability, and local delamination is prone to occur in high strain cycles; the device preparation involves multiple operations such as CNT coating, transfer and compounding, which is not conducive to batch production and consistency control.

[0011] A research has a sign language recognition system based on a flexible pressure sensor of ion composite nanofiber membrane, which combines a BPNN learning model to recognize 24 English letters of American sign language with an accuracy of 96.8%. However, U / V letters are very easy to confuse, and it is difficult to accurately recognize all 26 letters. The fuzzy recognition of U / V letters is still a key bottleneck in this field. SUMMARY

[0012] The purpose of the present application is to provide an Au-VG-based flexible strain sensor, a preparation method thereof, and an application thereof in gesture recognition.

[0013] To achieve the above purpose, the present application adopts the following technical solutions:

[0014] A preparation method of an Au-VG-based flexible strain sensor, the process being as follows:

[0015] (1) The rigid substrate is cleaned and dried;

[0016] (2) Vertical graphene is prepared on the rigid substrate of step (1) by chemical vapor deposition; the nanosheet of the obtained vertical graphene has a sheet height of 4-6 μm;

[0017] (3) A gold thin film is physically vapor deposited on the vertical graphene of step (2), and the thickness of the gold thin film is 30-100 nm, to obtain a VG-Au conductive structure;

[0018] (4) The VG-Au conductive structure is transferred to the cured lower PDMS substrate by peeling, and PDMS liquid is covered on the VG-Au conductive structure and cured to form an upper PDMS layer;

[0019] (5) The flexible sensing layer is cut into a suitable size, silver paste electrodes are formed at both ends of the device, and a medical breathable film is used for secondary packaging.

[0020] Further, the rigid substrate in step (1) is carbon paper, silicon wafer, alumina ceramic substrate, quartz wafer, etc. The rigid substrate needs to be cleaned and dried before use. The thickness of the rigid substrate is 100-500 μm.

[0021] Preferably, the rigid substrate in step (1) is sequentially cleaned with acetone, isopropyl alcohol and deionized water by ultrasonic cleaning for 5-15 minutes each time. After drying, it is baked on a hot plate at 100-150°C for standby use.

[0022] Further, in step (2), the process conditions of chemical vapor deposition are as follows: raw gas: methane (CH4), gas flow rate is 5-10 sccm; cavity pressure is 6-12 Pa; radio frequency power is 1000-2000 W; cavity temperature is 600-900°C; growth time is 5-40 minutes.

[0023] Preferably, the process conditions of chemical vapor deposition are as follows: raw gas: methane (CH4), gas flow rate is 5-8 sccm; cavity pressure is 8-10 Pa; radio frequency power is 1000-1500 W; cavity temperature is 600-900°C; growth time is 5-40 minutes.

[0024] Further, in step (3), the process conditions of physical vapor deposition are as follows: working gas: argon (Ar), gas flow rate is 10-20 sccm; cavity pressure is 0.5-2 Pa; radio frequency power is 100-200 W; cavity temperature is 100±10°C; deposition time is 10-50 seconds.

[0025] Preferably, the process conditions of physical vapor deposition are as follows: working gas: argon (Ar), gas flow rate is 10-20 sccm; cavity pressure is 0.5-2 Pa; radio frequency power is 120 W; cavity temperature is 100±10°C; deposition time is 20-30 seconds.

[0026] Further, in step (4), the PDMS pre-curing conditions are as follows: heating at 90-110°C for 5-15 minutes, and the thickness of the upper and lower PDMS is 100-500 μm.

[0027] Further, in step (5), silver paste leads are printed on both ends of the device and heat cured to form silver paste electrodes. The curing temperature of the silver paste electrodes is 60-80°C, and the curing time is 10-30 minutes.

[0028] The Au-VG flexible strain sensor prepared by the above preparation method.

[0029] The application of the Au-VG flexible strain sensor in the field of wearable devices and / or human-computer interaction.

[0030] Application of the above Au-VG flexible strain sensor in gesture recognition. Six Au-VG flexible strain sensors were fixed on the back of the fingers of the five fingers of the glove and the wrist, respectively, and were connected to the integrated circuit board with wires for signal collection. A wearable signal acquisition system was formed.

[0031] Collection and classification of gesture data: 26 letter gestures were demonstrated by five volunteers aged 22 to 30 years old while wearing smart gloves, each gesture was repeated 10 times, and finally 26x10x5=1300 sets of multi-channel electrical signal samples were obtained. According to the ratio of 6:2:2, it is divided into training set (780), verification set (260) and test set (260).

[0032] Data preprocessing: Extract the original signal within 3 seconds (300 sampling points x 6 channels in total) from each sample, and calculate the mean value of the first 100 sampling points of each channel as the baseline value. Then, subtract the corresponding baseline from the 300x6 signal matrix point by point in each channel to complete the static offset correction. The corrected signal is transposed to 6x300 channel-time format and merged into the sample set. Finally, traverse all samples to obtain the global maximum absolute value, and normalize each data point to map the signal amplitude uniformly to the interval [-1, 1].

[0033] Gesture recognition model training and deployment: The model used for training is an Inception-BiLSTM network that combines multi-scale convolution and bidirectional time series network. The Inception module extracts features from the input signal through multi-scale parallel convolution, captures local dynamic features under different time receptive fields, and concatenates high-dimensional time series feature maps in the channel dimension. The bidirectional LSTM module transmits in the forward and reverse directions to model the complete time series dependency. The preprocessed data is input into the model for training, and then the model achieves a classification accuracy of 99.6% for 26 letters in the test set.

[0034] The application provides a flexible strain sensor product, the core functional layer of which is a vertical graphene-metal composite conductive network, which is finally transferred and solidified on a polydimethylsiloxane (PDMS) flexible substrate. In the device preparation process, first, a vertical graphene (VG) is in-situ grown on a rigid substrate such as carbon paper by using a plasma enhanced chemical vapor deposition (PECVD) technology, the sheet height of which is controlled to be 4-6 μm, to construct a three-dimensional conductive framework; then, a gold thin film (Au) with a nominal thickness of 50±10 nm is deposited on the surface of the VG by using a radio frequency physical vapor deposition (RF-PVD) technology, to construct a crack-tunneling synergistic conductive structure and enhance the micro-strain response. After the conductive sensitive layer is constructed, the VG / Au layer is integrated with the PDMS by using a pretreated PDMS substrate to achieve high adhesion. The edges of the device are connected to a signal collector by using a printed silver paste or metal evaporation to form a lead zone and a metal wire. The flexible sensing unit is fixed on the back of five fingers and the wrist, respectively, to form a multi-channel distributed sensing network that is attached to the skin, for collecting local strain signals caused by gestures.

[0035] The flexible strain sensor of the application can be applied to physiological signal monitoring, and a wearable gesture recognition system based on the sensor can recognize gesture signals in real time; it is also suitable for human-computer interaction systems to complete the operation of unmanned aerial vehicles. It is suitable for the following uses:

[0036] 1. Gesture recognition and sign language recognition: the product can be attached to the back of the fingers and the wrist of a glove, to monitor the bending state of the fingers and joints in real time, to recognize complex dynamic gestures such as American Sign Language (ASL) through strain signals, and to be used for communication assistance for the hearing-impaired population or virtual reality control.

[0037] 2. Wearable human-computer interaction system: by integrating with a wireless communication module, the sensor can realize action interaction control with mobile devices, robots or unmanned aerial vehicles, has the natural interaction ability of low power consumption and high accuracy, and is suitable for remote operation and augmented reality scenarios. In addition, the sensor can also be integrated into a motion analysis and human-computer interaction system. The sensor glove is worn on the hand, and an additional sensor is attached at the elbow joint, which can be used for shot motion recognition and analysis. The continuous change of the resistance value of the sensor in the whole shooting process, from the ball, lifting, wrist pressing to the hand, can be used to identify the shooting posture and assist in optimizing the training strategy to improve the hit rate. In table tennis, wearing a sensor glove can identify the mechanical response difference of the forehand and backhand actions in the case of light block and heavy block, for action judgment and skill analysis. The glove is also suitable for grip analysis. Different objects have different grip forces due to differences in stiffness, weight or shape, which can be effectively identified through resistance value changes.

[0038] 3. Physiological signal monitoring: can be adhered to the pharynx, eyelid, tendon, joint and other parts, can be used for detecting small deformation such as speech, swallowing, blinking, muscle vibration, and can also detect the signal of larger amplitude deformation in the region such as finger joint and knee joint. It is helpful for fatigue detection, rehabilitation evaluation, silent input, human-computer interaction and other application scenarios. In the aspect of human physiological signal monitoring, the sensor attached to the chest area can realize the strain response monitoring under normal breathing and deep breathing state, showing significant resistance change difference, which can effectively distinguish the breathing frequency and rhythm under standing, sitting, speaking and different gait (such as slow walking, fast walking) state. The sensor attached to the cheek or upper eyelid position can sensitively respond to facial expressions and eye movements, such as monitoring the resistance value change caused by smiling, frowning, blinking, yawning and other subtle actions, which is suitable for emotion recognition and neuromuscular disorder evaluation.

[0039] In the aspect of joint motion monitoring, the sensor pasted on the knee can accurately identify the strain response signals corresponding to different bending amplitudes from the knee joint straightening (180°) to slight bending (120°), moderate bending (60°) and complete deep squatting state (0°); pasted on the ankle, it can be used to monitor the resistance value change caused by different force (light stepping, medium stepping, heavy stepping) in the process of stepping on the accelerator and brake in the simulated driving environment, which is helpful for driving behavior analysis and fatigue driving warning.

[0040] The present application proposes to construct a crack-tunneling synergistic structure by depositing a metal thin film in situ on the surface of vertical graphene (VG), thereby achieving the following key objectives:

[0041] Improve micro-strain resolution: linear response R 2 >0.98, and the average value of strain coefficient GF exceeds 240, which is significantly better than the existing VG scheme, and can distinguish subtle differences when collecting different physiological signals, which helps to improve gesture recognition signal recognition;

[0042] Reduce surface resistance to match the low-power system requirement: the initial surface resistance is controlled to be ≤500Ω / sq, which supports Bluetooth and other wireless communication modules for direct reading;

[0043] Improve cycle stability and response rate: the drift of 10,000 cycles ΔR / R0 is <3%, and the response time is <100ms, which meets the dynamic recognition;

[0044] Enhance system adaptability: integrate the sensor into a glove to collect six-channel signals on the back of the finger and the wrist, and combine with the deep learning model Inception-BiLSTM model to greatly improve the ASL 26 letter recognition accuracy, which can reach 99.6%, verifying its practicability.

[0045] In summary, the present application proposes a new solution path of material structure, conduction mechanism and system integrated collaborative optimization for the key problems of the existing flexible strain sensing technology, such as "insufficient micro-strain sensitivity, high surface resistance, unstable response and complex manufacturing, recognition defects integrated in sign language recognition system", etc., which has good academic innovation and engineering practical prospect.

[0046] The core idea of the present application is to construct a crack-tunneling synergistically enhanced three-dimensional conductive network structure to significantly improve the sensitivity, linearity and stability of the flexible strain sensor in the micro-strain interval, and the sensor can be integrated into a wearable device to realize accurate gesture recognition combined with a deep learning model. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 Figure 1 is a structural schematic diagram of the sensor of the present application;

[0048] Figure 2 Figure 2 is an SEM image of vertical graphene (VG) in Example 1;

[0049] Figure 3 Figure 3 is the resistance change rate ΔR / R0 of the sensor in Example 1 under different strains;

[0050] Figure 4 Figure 4 is the response time and recovery time of the sensor in Example 1 during loading and unloading;

[0051] Figure 5 Figure 5 is the significant resistance response change of the sensor in Example 1 under different tensile displacements (1-9 mm);

[0052] Figure 6 Figure 6 is the cyclic sensing performance of the sensor in Example 1 during loading and unloading, and no significant performance decline is monitored after 10,000 cycles;

[0053] Figure 7 Figure 7 is a confusion matrix of the classification of 26 English letter gestures by the wearable glove containing the sensor in Example 1. DETAILED DESCRIPTION

[0054] The following examples merely afford the skilled person a complete disclosure and description of how to make and evaluate the compounds, compositions, articles, devices, and / or methods claimed in the present application and are intended to be purely exemplary and are not intended to be limiting to the scope of the inventors' claimed application. Efforts have been made to ensure accuracy with respect to numbers (e.g., amounts, temperature, etc. ) but some errors and deviations should be accounted for. Unless indicated otherwise, parts are parts by weight, molecular weight is weight average molecular weight, temperature is in degrees Celsius, and pressure is at or near atmospheric.

[0055] Plasma enhanced chemical vapor deposition apparatus (PECVD)

[0056] For in-situ growth of vertical graphene (VG) on rigid carbon paper substrate. The device should have the ability to independently control the radio frequency power, gas flow, temperature and cavity pressure, and the appropriate reaction gas is methane (CH4). The radio frequency power regulation range is 1000~2000 W, the working temperature is 600~900℃, the gas precision control range is 1~20 sccm, and the reaction cavity vacuum control ability is less than 5 Pa.

[0057] Radio frequency physical vapor deposition device (RF-PVD)

[0058] For depositing a metal thin film with a thickness of 50 nm on the surface of VG. The device should be equipped with a high-purity gold target, with a radio frequency power regulation function of 100~200 W, a cavity temperature controllable to about 100℃, and a high vacuum system (background pressure <1×10 -3 Pa). Argon gas (10~20 sccm) can be stably introduced for working atmosphere.

[0059] Six-channel signal acquisition control board and Bluetooth module

[0060] In the present application, a customized circuit board is used to connect sensor electrodes and collect strain electrical signals. The control board includes six trans-impedance amplifiers (TIAs), an analog multiplexer, a signal bias amplifier, a 32-bit MCU (with built-in ADC), and a Bluetooth module supporting BLE 5.0 and above protocols, with a communication power consumption of ≤5 mW. It is used to send multiple channel signals in a serial manner to a mobile phone or other terminal. Thus forming a six-channel synchronous acquisition low-power gesture recognition system, the core uses an 8-bit high-performance microcontroller, which can simultaneously acquire analog signals from multiple finger bending sensors, and after being processed by a dual-channel high-precision operational amplifier, it is digitized by the 12-bit ADC of the main control chip; The present application only uses six channels, and the processing results are transmitted through low-power Bluetooth; it is powered by a 3.7V lithium battery, supports USB charging and 3.3V voltage output, and is equipped with a hardware power switch, realizing a one-stop solution for multi-channel signal synchronous acquisition, high-precision processing and wireless transmission.

[0061] Example 1

[0062] A preparation method of a flexible strain sensor based on Au-VG is as follows:

[0063] (1) Preparation of the lower substrate PDMS: Sylgard 184 PDMS prepolymer and curing agent were mixed in a mass ratio of 10:1 and stirred at 300 rpm for 5 minutes. The mixture was degassed in a vacuum environment for not less than 10 minutes to remove bubbles. The degassed PDMS was poured into a clean mold, the film thickness was controlled to be about 200 μm, and the film was heat-cured at 80°C for 2 hours to obtain a flexible lower substrate PDMS film.

[0064] (2) Growth of vertical graphene (VG) framework: A conductive carbon paper with a thickness of 150-200 μm was selected as a temporary rigid carrier, and was ultrasonically cleaned in acetone, isopropyl alcohol and deionized water for 10 minutes each time, and was dried on a hot plate at 120°C for 10 minutes. The carbon paper was placed in a plasma enhanced chemical vapor deposition (PECVD) chamber, and the process parameters were set as follows: methane (CH4), methane flow rate 5 sccm, chamber pressure 10 Pa, radio frequency power 1000 W, and temperature 800°C. The carbon paper was grown for about 20 minutes to form a three-dimensional vertical graphene network with a sheet height of about 5 μm. The SEM image is shown in FIG. 2(a), which is an SEM image of the vertical graphene (VG) prepared in the application under a magnification of 7k. It can be seen that the VG grows uniformly on the surface of the substrate, forming a dense and high vertical sheet structure. The sheet layers are staggered to form multi-scale pore channels, and the whole has a uniform three-dimensional network morphology. The SEM image of the VG under a magnification of 40k is shown in FIG. 2(b). The sheet edges are sharp and wavy, the thickness is nanoscale, the surface is smooth and has a three-dimensional relief, and the inter-sheet gaps can be identified as sub-micron pores under high magnification. Figure 2

[0065] (3) Deposition of 50 nm gold film: After growth, the carbon paper was directly transferred to a physical vapor deposition (PVD) chamber for gold film deposition without air exposure. High-purity gold target material was used, and the sputtering was performed under the following conditions: argon flow rate 15 sccm, chamber pressure 1 Pa, radio frequency power 120 W, and temperature 100°C. The sputtering time was 30 seconds, and a gold film with a thickness of about 50 nm was obtained. This layer cooperates with the VG network to construct crack-induced and conductive compensation channels, thereby improving the sensitivity and conductive stability of the device.

[0066] (4) Transfer and lamination of the whole structure: The Au-VG layer was transferred to the cured lower PDMS substrate by mechanical peeling. During the transfer process, the lamination was kept flat to avoid curling or crack propagation. The laminated structure constitutes the lower three layers of the sensing core of the device.

[0067] ​(5) Upper PDMS encapsulation: Repeat the PDMS mixing and degassing process in step one to uniformly cover the Au layer surface with PDMS liquid, controlling the thickness to 200μm. After pre-curing at room temperature for 15 minutes, heat-cur at 80℃ for 2 hours to form a flexible upper encapsulation structure, as shown in the figure. Figure 1 As shown, it includes a lower substrate PDMS, a vertical graphene layer, a gold film layer, and an upper substrate PDMS.

[0068] (6) Device cutting and lead encapsulation: The entire sensor is cut into 35 mm × 5 mm sections using laser cutting technology, with the middle 30 mm area serving as the effective strain sensing zone. Silver paste leads are printed at both ends of the device and thermally cured, followed by secondary encapsulation using a medical breathable membrane. This encapsulation effectively prevents oxidation of the Au layer and corrosion from sweat, improving device stability.

[0069] The sensor prepared in Example 1 was subjected to progressive deformation within a strain range of 0-17% using a tensile platform. The corresponding resistance changes were recorded, and GF was calculated. The results are as follows: Figure 3 As shown, the results indicate that the strain remains highly linear (R0) within the 3.3%–13.3% strain range. 2 >0.98), GF reached 232.5, and the average GF exceeded 240.

[0070] The sensor of Example 1 was subjected to periodic stretching-releasing at a 1Hz cycle, and the resistance change curve over time was recorded. See the results for details. Figure 4 ,Depend on Figure 4 It can be seen that both the response time and the recovery time are less than 100ms.

[0071] Figure 5 The relative resistance change rate (ΔR / R0) of the sensor in Example 1 under different strain amplitudes (1 mm, 3 mm, 5 mm, 7 mm, and 9 mm) is shown. As the displacement increases stepwise, the peak signal shows a significant increasing trend. At the same time, after multiple consecutive loading at each displacement level, the waveform peaks and valleys overlap, and the baseline after stretching and releasing always returns to near zero, without drifting or hysteresis.

[0072] Figure 6 The sensor of Example 1 shown was subjected to 10,000 tensile-release cycles at 3% strain, and the changes in resistance after the initial cycle were monitored. The results showed that the drift was less than 3% after 10,000 tensile-release cycles.

[0073] Six sensors were fixed on the back of the five fingers and the wrist of the glove (such as commercially available disposable PVC gloves) respectively, and were connected to the integrated circuit board for signal collection. The fixed time step of each gesture was 300 (3s, 100Hz sampling frequency). The data set was collected by 5 people wearing gloves respectively, and each person repeated 26 gestures for 10 times, a total of 10*5*26=1300 groups of data. Among them, the collected data set is divided into training set, validation set and test set according to 6:2:2. The collected data is preprocessed, first, the original data of the first 3 seconds (300 sampling points) of each channel is extracted from the data, and then the average value of the first 100 sampling points in the signal is taken as the baseline in each channel; then the whole 300*6 original signal matrix is subtracted from the baseline point by point to realize the static offset correction; then the corrected signal is transposed to 6*300 channel-time format and collected into the sample list; after processing all samples, the global maximum absolute value is found, and each value is normalized to map the signal distribution to the [-1, 1] interval, so as to obtain a standardized tensor that removes the baseline offset and unifies the amplitude range. After the preprocessed data is input into the designed Inception-BiLSTM model for training, high-precision recognition of the test set data is realized, and the results are shown in Figure 7 As shown in Figure 7 It can be seen that the wearable sensing system of the application combines inception-BiLSTM to process, recognize and classify the gesture signals of 26 English letters, and finally the accuracy of the test set reaches 99.6%,

[0074] The performance parameters of the sensor prepared in example 1 and the existing sensor are shown in table 1.

[0075] Table 1: Comparison of Au-VG sensor of example 1 and existing VG sensor in performance index

[0076]

[0077] Example 2

[0078] The difference from example 1 is that the vertical graphene growth process is optimized: the CH4 gas flow range is recommended to be accurately controlled at 5~8 sccm; the cavity pressure is recommended to be controlled at 8~10 Pa, the VG sheet height is concentrated at 4~6 μm, the structure is more uniform, and the conductive channel is more stable; the radio frequency power in the range of 1000~1500W can improve the verticality of the sheet structure.

[0079] Although the preferred embodiments of the application have been illustrated and described, it will be appreciated by those skilled in the art that changes can be made to the described embodiments without deviating from the inventive concept, and all modifications and changes come within the scope of the present application as defined by the following claims.

Claims

1. A method for fabricating a flexible strain sensor based on Au-VG, characterized in that, The process is as follows: (1) Clean and dry the rigid substrate; (2) Vertical graphene was prepared by chemical vapor deposition on a rigid substrate in step (1); the height of the obtained vertical graphene nanosheets was 4~6 μm. (3) Physical vapor deposition of gold film on the vertical graphene in step (2) with a thickness of 30~100nm to obtain VG-Au conductive structure; (4) The above-mentioned VG-Au conductive structure is transferred to the cured lower PDMS substrate by peeling, and PDMS liquid is covered on the VG-Au conductive structure and cured to form the upper PDMS; (5) Cut the flexible sensing layer to a suitable size, form metal electrodes at both ends of the device, and then use a medical breathable membrane for secondary encapsulation.

2. The fabrication method of the Au-VG flexible strain sensor according to claim 1, characterized in that, In step (2), the process conditions for chemical vapor deposition are as follows: raw material gas: methane (CH4), gas flow rate is 5~10 sccm; chamber pressure is 6~12 Pa; radio frequency power is 1000~2000 W; chamber temperature is 600~900℃; growth time is 5~40 minutes.

3. The fabrication method of the Au-VG flexible strain sensor according to claim 1, characterized in that, In step (3), the physical vapor deposition process conditions are as follows: working gas: argon (Ar), gas flow rate is 10~20 sccm; chamber pressure is 0.5~2 Pa; radio frequency power is 100~200 W; chamber temperature is 100±10℃; deposition time is 10~50 seconds.

4. The method for fabricating an Au-VG flexible strain sensor according to claim 1, characterized in that, In step (4), the PDMS pre-curing conditions are: heating at 90~110℃ for 5~15 minutes, and the thickness of the upper and lower PDMS layers is 100~500μm.

5. The method for fabricating an Au-VG flexible strain sensor according to claim 1, characterized in that, In step (5), silver paste is printed on both ends of the device and thermally cured to form silver paste electrodes. The curing temperature of the silver paste electrodes is 60~80℃ and the curing time is 10~30 minutes.

6. The method for fabricating an Au-VG flexible strain sensor according to claim 2, characterized in that, The process conditions for chemical vapor deposition are as follows: raw material gas: methane (CH4), gas flow rate: 5~8 sccm; chamber pressure: 8~10 Pa; radio frequency power: 1000~1500 W; chamber temperature: 600~900℃; growth time: 5~40 minutes.

7. The method for fabricating an Au-VG flexible strain sensor according to claim 3, characterized in that, The physical vapor deposition process conditions are as follows: working gas: argon (Ar), gas flow rate: 10~20 sccm; chamber pressure: 0.5~2 Pa; radio frequency power: 120 W; chamber temperature: 100±10℃; deposition time: 20~30 seconds.

8. A flexible strain sensor based on Au-VG, characterized in that, It is prepared by the preparation method according to any one of claims 1 to 7.

9. An application of Au-VG flexible strain sensor in wearable devices and / or human-computer interaction, characterized in that, The Au-VG-based flexible strain sensor described in claim 8 is used.

10. The application according to claim 9, characterized in that, The application of Au-VG flexible strain sensor in gesture recognition and sign language.

Images