A strain sensing film for multi-language handwritten text recognition, a flexible strain sensor, and its preparation method and application

Abstract

The present invention belongs to the technical field of flexible strain sensors, and specifically relates to a strain sensing film for multilingual handwritten text recognition, a flexible strain sensor, and a preparation method and application thereof. PPy is attached to the surface of HEC cellulose to form a uniform PPy / HEC conductive film, and the conductive film is distributed between the MXene layers. The ultra-high conductivity of the MXene increases the sensitivity of the strain sensing film, and the PPy / HEC conductive film is connected to the MXene under stretching to ensure the connectivity of the conductive network on the strain sensing film. The flexible strain sensor has excellent sensing performance such as a wide linear strain detection range, high sensitivity, reliable repeatability, and ultra-fast response-recovery time. The MXene / PPy / HEC sensor can detect different levels of human physiological activities, and exhibits excellent performance in measuring external strain changes and real-time motion detection. The MXene / PPy / HEC sensor obtains information through machine training and learning based on the wavelet time scattering network-long short-term memory neural network (WTSN-LSTM) algorithm, realizing high-precision recognition of handwritten text in multiple languages ​​such as English words, Arabic numerals, and Chinese character signals.

Description

Technical Field

[0001] The present invention belongs to the technical field of flexible strain sensors, and in particular relates to a strain sensing film for multi-language handwritten text recognition, a flexible strain sensor, and a preparation method and application thereof. Background Art

[0002] In recent years, many research results have been published in the field of flexible strain sensors. Researchers have shifted their research focus from the basic performance of sensors to the practical application of sensors, and have achieved major breakthroughs in many fields such as soft robots, electronic skin, human-computer interaction and medical monitoring. Strain sensors have achieved real-time sensing applications in human health and motion detection due to their excellent elasticity and toughness and perfect fit with human skin. Strain sensors that fully detect human motion activities need to meet the characteristics of high sensitivity, fast response, stable dynamic performance and high linearity. In addition, the size, volume, wearing comfort, wear resistance and life of the sensor are also issues that designers need to focus on. Although flexible strain sensors have made novel developments in the fields of human-computer interaction, medical monitoring and handwriting recognition, they also face many challenges such as accurate capture of weak signals, comprehensive acquisition of signal information and accurate recognition.

[0003] The development of artificial neural networks has made sensors develop in the direction of intelligence, and has also made human-computer interaction closely related to human activities, such as speech recognition, handwriting recognition, gesture recognition and gait recognition, become a research hotspot. Among them, handwriting recognition is a very important part of human activities and an indispensable communication connection between people and computers. The biggest challenge facing handwriting recognition is the interference of similarities between different characters. At present, some flexible strain sensors with high performance and good stability have been developed, but they still face huge challenges in practical applications, specifically: flexible sensors need to accurately capture weak signals, have excellent biocompatibility, be comfortable and soft to wear, and interact with smart devices. In addition to the dilemma of accurately collecting handwriting signals, the accuracy, operation speed and complexity of intelligent algorithms for accurate recognition have limited the development of handwriting recognition. WTSN (wavelet time scattering network) has excellent data dimensionality reduction function and is widely used in feature extraction of time series. LSTM (long short-term memory neural network) can solve the long-term dependency problem of general neural networks and is suitable for processing and predicting important events with relatively long intervals and delays in time series. The excellent performance of WTSN and LSTM makes them have great potential in the field of handwritten text recognition.

[0004] Flexible sensors have broad application prospects in the fields of health monitoring, human motion detection, human-computer interaction and text recognition. They can fit closely to the human body and sense small deformations. Through wireless transmission technology, they can detect the human body state in real time on smart terminals, and are expected to be applied in medical monitoring, motion detection and electronic skin. In addition, the use of flexible strain sensors as text signal acquisition boards combined with artificial intelligence algorithms has made great progress in the fields of multilingual text recognition, voice control and recognition. Therefore, there is an urgent need for a strain sensing film, flexible strain sensor and its preparation method and application for multilingual handwritten text recognition that is highly stretchable and highly sensitive, has reliable repeatability and ultra-fast response-recovery time. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a strain sensing film, a flexible strain sensor and its preparation method and application for multi-language handwritten text recognition, and realizes online monitoring of human motion signals and recognition of handwritten text in multiple languages, playing a role in the fields of human motion detection, medical health monitoring and handwritten letter recognition.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution: a strain sensing film for multi-language handwriting recognition, PPy is attached to the surface of HEC cellulose to form a uniform PPy / HEC conductive film, and the conductive film is distributed between MXene layers. The ultra-high conductivity of the MXene increases the sensitivity of the strain sensing film, and the PPy / HEC conductive film connects each MXene under stretching to ensure the connectivity of the conductive network on the strain sensing film.

[0007] Further, the preparation steps include:

[0008] (1) Preparation of MXene

[0009] Slowly add LiF to the hydrochloric acid solution, stir evenly, then slowly add Ti3AlC2 in several times, stir the solution in a 35°C water bath for 24 hours, wash the solution, centrifuge it with deionized water several times until there is no supernatant, ultrasonically treat the solution in an ice bath for 1 hour, then centrifuge it at 3500 rpm, and save the supernatant for later use;

[0010] (2) Preparation of PPy / HEC

[0011] The HEC cellulose solution was added to a dimethylformamide (DMF) solution and stirred vigorously at 25°C to completely disperse the HEC cellulose. Pyrrole was added to the HEC suspension and stirred vigorously for 0.5 h to allow the pyrrole to completely self-assemble on the surface of the HEC cellulose. A mixed solution of FeCl3·6H2O and HCl was then added and reacted at 4°C for 8 h. After the reaction, the precipitate was washed with ethanol, deionized water, and 1 M HCl to remove the monomer and the oxidant. The washed product was redissolved in 1 M HCl to form a PPy / HEC black suspension.

[0012] (3) Preparation of MXene / PPy / HEC strain sensing film

[0013] The MXene and PPy / HEC solutions were mixed and stirred to obtain a MXene / PPy / HEC solution, and the MXene / PPy / HEC composite sensing material was sprayed on the surface of the substrate, and then dried to obtain a strain sensing film.

[0014] Furthermore, the mass ratio of LiF and Ti3AlC2 in (1) is 1:1, and the concentration of the hydrochloric acid solution is 9M.

[0015] Furthermore, the mass fraction of the HEC solution in (2) is 0.2 wt %, and the volume ratio of the HEC solution to dimethylformamide is 2:1. The mass ratio of HEC to pyrrole is 1:10; The mass ratio of pyrrole to FeCl3·6H2O is 1:1.89; the concentration of the hydrochloric acid solution is 1M, The molar ratio of hydrochloric acid to pyrrole is 3:1 .

[0016] Furthermore, in (3) The volume ratio of MXene solution and PPy / HEC solution was 2:1.

[0017] A flexible strain sensor for multi-language handwritten text recognition is fabricated by depositing MXene / PPy / HEC on the PDMS surface by spraying.

[0018] Furthermore, the sensing principle of the sensor is:

[0019] The sensing material is adsorbed on the surface of PDMS. When the sensor is stretched, cracks appear in the sensing layer, causing an increase in resistance. When the stretching is restored, the cracks disappear and the resistance of the strain sensor returns to its original state.

[0020] PPy attached to the HEC surface forms a uniform film, which plays a good connecting role. Under small strain conditions, a small separation occurs between the MXenes, resulting in a significant change in the resistance of the sensing layer, causing the strain sensor based on MXene / PPy / HEC to maintain a high sensitivity under small strains. Under larger strains, the MXenes will be completely separated from each other, and PPy / HEC acts as a bridge to connect the MXene to ensure the access to the sensing layer, greatly extending the sensing range of the strain sensor.

[0021] A method for preparing a flexible strain sensor for multi-language handwritten text recognition.

[0022] (1) PDMS prepolymer and curing agent were uniformly mixed in a ratio of 10:1 and placed in a vacuum drying oven for 20 min to remove bubbles;

[0023] (2) The mixed reagent is evenly coated on a glass slide and placed in a vacuum drying oven at 80° C. for 2 h to form a PDMS film;

[0024] (3) Spraying the MXene / PPy / HEC composite sensing material onto the surface of the PDMS film by spraying, and spraying and drying for multiple times to form a strain sensing film;

[0025] (4) Cut the PDMS film into long strips, install conductive tape on both sides, and lead out electrodes;

[0026] (5) A layer of PDMS is further covered on the PDMS surface with the strain sensing film and placed in an oven at 60°C for curing, thereby preparing a flexible strain sensor with a sandwich structure.

[0027] An application of flexible strain sensor for multi-language handwritten text recognition.

[0028] The strain sensor is used for monitoring human health and motion status, and detecting swallowing, voice recognition, breathing and joint bending signals.

[0029] Furthermore, the sensor measured and obtained the handwritten English words, Arabic numerals and Chinese character signals of four volunteers, and recognized different English letters and Arabic numerals through the WTSN-LSTM algorithm, with the recognition accuracy higher than 96%.

[0030] The present invention provides a strain sensing film, a flexible strain sensor for multilingual handwriting recognition, and its preparation method and application. The present invention designs a flexible strain sensor with a sandwich structure, which is composed of a MXene / PPy / HEC (titanium carbide / polypyrrole / hydroxyethyl cellulose) sensitive material film and a PDMS (polydimethylsiloxane) flexible substrate. The sensor has excellent sensing performance such as a wide linear strain detection range (0-94%), high sensitivity (gauge factor up to 357.5), reliable repeatability (>1300 cycles) and ultra-fast response-recovery time (300ms / 300ms). The MXene / PPy / HEC sensor is capable of detecting different levels of human physiological activities and shows excellent performance in measuring external strain changes and real-time motion detection. For the first time, the present invention combines the handwritten English words, Arabic numerals and Chinese characters of four volunteers measured by sensors with the wavelet time scattering network-long short-term memory neural network (WTSN-LSTM) algorithm to successfully identify different English letters and Arabic numerals with a recognition accuracy of more than 96%, indicating that the device can be used as an intelligent handwriting board for handwritten text recognition. The flexible strain sensor is expected to play a role in the fields of human motion detection, medical health monitoring and handwritten letter recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a preparation process diagram of the MXene / PPy / HEC composite sensing material according to an embodiment of the present invention;

[0032] Figure 2 X-ray diffraction (XRD) patterns of MXene, PPy / HEC, HEC and MXene / PPy / HEC (Rigaku D / Max2500PC) according to the embodiments of the present invention;

[0033] Figure 3 Fourier spectra of MXene, PPy, HEC and MXene / PPy / HEC materials according to the embodiments of the present invention;

[0034] Figure 4 a is a SEM image of the PPy / HEC material of the embodiment of the present invention;

[0035] Figure 4 b is a SEM image of the MXene material of an embodiment of the present invention;

[0036] Figure 4 c is a SEM image of the PPy material of the embodiment of the present invention;

[0037] Figure 4 d is the SEM image of the MXene / PPy / HEC material of the embodiment of the present invention;

[0038] Figure 5 a is a schematic diagram of an experimental test device according to an embodiment of the present invention;

[0039] Figure 5 b is the preparation process of the flexible strain sensor according to an embodiment of the present invention;

[0040] Figure 6 a is a schematic diagram of the structure of a sensor and a sensitive material installed on a human body according to an embodiment of the present invention;

[0041] Figure 6 b is a schematic diagram of the change of microcracks during the stretching process of an embodiment of the present invention;

[0042] Figure 7 a is a typical response strain curve of MXene, PPy / HEC and MXene / PPy / HEC sensors according to an embodiment of the present invention;

[0043] Figure 7 b is the typical response curve of MXene, PPy / HEC and MXene / PPy / HEC sensors under the same strain in the embodiment of the present invention;

[0044] Figure 7 c is a typical response strain curve of the MXene / PPy / HEC sensor of an embodiment of the present invention;

[0045] Figure 7 d is the IV characteristic curve of the MXene / PPy / HEC strain sensor under different strains according to an embodiment of the present invention;

[0046] Figure 8 a. Figure 8 b and Figure 8 c is the relative resistance change of the MXene / PPy / HEC strain sensor under different strains according to an embodiment of the present invention;

[0047] Figure 8 d is the response time and recovery time of the MXene / PPy / HEC strain sensor according to an embodiment of the present invention;

[0048] Figure 8 e is the response change of the strain sensor at different frequencies according to the embodiment of the present invention;

[0049] Figure 8 f is the response change of the strain sensor under different strains and the stretching speed under different strains according to the embodiment of the present invention;

[0050] Figure 8 g is the response change of the strain sensor in the embodiment of the present invention during the step stretching process and recovery;

[0051] Figure 8h is the hysteresis curve of the embodiment of the present invention at different strains of 20-80%;

[0052] Figure 8 i is the repeatability of the MXene / PPy / HEC strain sensor of the present invention. The inset shows the stable response signal output during the test;

[0053] Fig. 9 A schematic diagram of the strain sensor preparation of the embodiment of the present invention and sensor response diagrams and mechanism descriptions in different application scenarios (pulse, exercise, real-time monitoring of joint movement, and handwritten text recognition);

[0054] Fig.10 a. Fig.10 b. Fig.10 c and Fig.10 d is the response of the strain sensor according to the embodiment of the present invention when detecting the bending of human joints (elbows, fingers, wrists and knees);

[0055] Fig.10 e and Fig.10 f is the response of the strain sensor of the embodiment of the present invention when it is connected to a finger and written on a table;

[0056] Fig.10 g is the sensor response of different mouse clicks in the embodiments of the present invention;

[0057] Fig.10 h is the response of the strain sensor during walking in the embodiment of the present invention.

[0058] Fig.11 a is a wrist pulse diagram of an embodiment of the present invention;

[0059] Fig.11 b is an enlarged view of a single pulse wave according to an embodiment of the present invention;

[0060] Fig.11 c is the pronunciation test result of words with different syllables according to the embodiment of the present invention;

[0061] Fig.11 d and Fig.11 e is a diagram showing the response of human swallowing and swelling according to an embodiment of the present invention;

[0062] Fig.11 f is a breathing state test according to an embodiment of the present invention;

[0063] Fig.12 The responses of the MXene / PPy / HEC strain sensor of the embodiment of the present invention to the handwritten letters "A", "B", "C", "D", "P" and "U" of four volunteers;

[0064] Fig.13Response of the MXene / PPy / HEC strain sensor when volunteers in the embodiments of the present invention handwrite the letters "E-Z" (except "P" and "U");

[0065] Fig.14 Time-frequency spectrum of the response of the strain sensor in the embodiments of the present invention to the handwritten letters "A", "B", "C", "D", "P" and "U";

[0066] Fig.15 Time-frequency spectrum of the response of the strain sensor in the embodiments of the present invention to the handwritten digits "1-6";

[0067] Fig.16 Time-frequency spectrum of the response of the strain sensor in the embodiments of the present invention to the handwritten digits "7-9, 0";

[0068] Fig.17 a is the flowchart of handwritten recognition in the embodiments of the present invention;

[0069] Fig.17 b is the color map letter waterfall diagram of the response data in the embodiments of the present invention ("A", "B", "C", "D", "P" and "U");

[0070] Fig.17 c is the recognition confusion matrix of the handwritten English letters by Volunteer 1 in the embodiments of the present invention;

[0071] Fig.17 d is the recognition confusion matrix of four volunteers in the embodiments of the present invention;

[0072] Fig.17 e is the response curve of the sensor to the Arabic numerals "0-9" in the embodiments of the present invention;

[0073] Fig.17 f is the color map Arabic numeral "0-9" waterfall diagram of the response data in the embodiments of the present invention;

[0074] Fig.17 g is the recognition confusion matrix of the handwritten Arabic numerals by Volunteer 1 in the embodiments of the present invention;

[0075] Fig.18 a is the intelligent application description of the strain sensor in the embodiments of the present invention;

[0076] Fig.18 b is the intelligent device detection diagram of the index finger bending signal in the embodiments of the present invention;

[0077] Fig.18 c and Fig.18 d are the response curves of the strain sensor to the Chinese characters "中" and "国" in the embodiments of the present invention;

[0078] Fig.18e is the time spectrum of Chinese characters '中' and '国' in the embodiments of the present invention. Detailed implementation manners

[0079] The following further elaborates on its detailed implementation manners in conjunction with the accompanying drawings.

[0080] The reagents and instruments in the following embodiments are all conventional experimental reagents and instruments.

[0081] Embodiment 1:

[0082] A strain-sensing film for multi-language handwritten text recognition, where PPy is attached to the surface of HEC cellulose to form a uniform PPy / HEC conductive film, and the conductive film is distributed between the layers of MXene. The ultra-high conductivity of MXene increases the sensitivity of the strain-sensing film, and the PPy / HEC conductive film connects each MXene under stretching to ensure the connectivity of the conductive network on the strain-sensing film.

[0083] Layered MXene was prepared by an etching method, and at the same time, MXene / PPy / HEC composite nanomaterials were prepared by mixing with in-situ polymerized PPy / HEC. MXene has a controllable layered structure and excellent conductivity, and can exhibit excellent responses during stretching. The introduction of the polymer improves the connection effect between materials and enhances the detection range of the sensor.

[0084] A preparation method of a strain-sensing film for multi-language handwritten text recognition

[0085] Appropriate deionized water was added to concentrated hydrochloric acid to prepare 9M HCl, and then 2g of weighed LiF was slowly added to the hydrochloric acid and waited for it to be stirred evenly. 2g of Ti3AlC2 was slowly added to the prepared solution in multiple portions to prevent splashing and make the etching more uniform. The solution was stirred in a water bath at 35°C for 24h. The above solution was washed, centrifuged with deionized water multiple times until there was no supernatant, and the solution was ultrasonically treated. The centrifugation speed was increased and centrifuged again until there was no supernatant. The obtained solution was ultrasonically treated in an ice bath for 1 hour, and then centrifuged at 3500 r / min, and the supernatant was taken and stored for later use.

[0086] PPy / HEC conductive nanocomposites were prepared by in situ oxidative polymerization of pyrrole self-assembly on HEC cellulose. The synthesis was as follows: 50 ml (0.2 wt%) HEC solution was added to 25 ml dimethylformamide (DMF) solution and stirred vigorously at 25 °C to completely disperse the HEC cellulose. 1 g pyrrole was added to the HEC suspension and stirred vigorously for 0.5 h to completely self-assemble onto the cellulose surface. Then 1.89 g of a mixed solution of FeCl3·6H2O and 1 M HCl was added, wherein FeCl3·6H2O was the oxidant catalyst and HCl was the dopant. The reaction was carried out at 4 °C for 8 h. After the reaction, the precipitate was washed with ethanol, deionized water and 1 M HCl to remove the monomers and oxidants. Finally, the washed product was redissolved in 1 M HCl to form a PPy / HEC black suspension.

[0087] The MXene and PPy / HEC solutions were mixed and stirred to obtain a MXene / PPy / HEC solution. The MXene / PPy / HEC composite sensing material was sprayed on the substrate surface and dried to obtain a strain sensing film.

[0088] Figure 1 This is the preparation process diagram of MXene / PPy / HEC composite sensing material.

[0089] Characterization and analysis of strain sensing films:

[0090] SEM, XRD and FT-IR characterization methods were used to analyze the microstructure of MXene / PPy / HEC nanocomposite films.

[0091] Figure 2 X-ray diffraction (XRD) patterns of MXene, PPy / HEC, HEC and MXene / PPy / HEC (Rigaku D / Max 2500PC).

[0092] The main diffraction peak of MXene is at 6.54°, which is attributed to its (002) crystal plane.

[0093] The functional groups and chemical structures of the sensing materials were characterized by Fourier transform infrared spectroscopy.

[0094] Figure 3 Fourier spectra of MXene, PPy, HEC and MXene / PPy / HEC materials.

[0095] The spectrum of HEC shows that pure HEC has peaks at 1016, 1421, 1739, 2950 and 3420 cm -1The characteristic peaks of the spectrum are hydrogen bending of OH, carboxyl and carboxyl salts, double bond stretching of carbonyl, hydrogen stretching of CH and hydrogen stretching of bond OH. -1 The three characteristic absorption peaks at 1471 and 1311 cm-1 are attributed to the out-of-plane and in-plane bending vibrations of CH and CH, respectively. -1 The absorption peak at 1561 cm is caused by the stretching vibration of CN. -1 PPy at 3420cm -1 The characteristic peak at is caused by the stretching vibration of NH.

[0096] The surface morphology of the sensing materials was characterized by scanning electron microscopy.

[0097] Figure 4 a is the SEM image of PPy / HEC material; Figure 4 b is the SEM image of MXene material; Figure 4 c is the SEM image of PPy material; Figure 4 d is the SEM image of MXene / PPy / HEC material.

[0098] like Figure 4 As shown, PPy / HEC is a film with uniform microspheres, while MXene is a stack of sheets, and PPy presents uniform microspheres. It can be observed that PPy microspheres are attached to the flake-like MXene in the composite material. This provides a pathway for the connection between the materials, increasing the response range of the sensor during stretching.

[0099] The present invention designs a simple, time-saving and low-cost method based on spraying to prepare a flexible strain sensor with a wide linear strain detection range (0-94%), high sensitivity (gauge factor up to 357.5), reliable repeatability (>1300 cycles) and ultrafast response-recovery time (300ms / 300ms). The microstructure and elemental composition of MXene, PPy and HEC materials were studied by XRD, SEM and FT-IR characterization techniques.

[0100] Embodiment 2:

[0101] A flexible strain sensor is prepared by depositing MXene / PPy / HEC on a PDMS surface by a spraying method to obtain a flexible strain sensor with a sandwich structure.

[0102] A method for preparing a flexible strain sensor, wherein the prepolymer and curing agent of PDMS are mixed evenly in a ratio of 10:1, and then placed in a vacuum drying oven for 20 minutes to obtain PDMS. When the bubbles in the mixed reagent completely disappear, the PDMS is placed on a glass slide. After heating at 80°C for 2 hours, a transparent and stretchable PDMS is successfully prepared. Subsequently, the MXene / PPy / HEC solution is sprayed on the PDMS substrate and dried in a vacuum drying oven at 60°C. Then, a conductive tape is installed on both sides of the material. Finally, the liquid PDMS is poured into the top layer material and dried in a vacuum drying oven at 80°C to prepare a sandwich structure strain sensor. Figure 5 Structural design flow chart and test platform for flexible strain sensor devices.

[0103] The sandwich packaging method allows the sensor strain sensing film and the sensing electrode to be encapsulated inside the flexible PDMS. During the stretching process, the resistance change of the sensing layer will not be disturbed by the external environment, such as humidity, thus ensuring that the strain sensor has excellent stability.

[0104] The sensing principle of the sandwich structure MXene / PPy / HEC flexible strain sensor is:

[0105] MXene / PPy / HEC strain sensor is an adsorption strain sensor, and the sensing material is adsorbed on the surface of PDMS. The sensing characteristics of the sensor are caused by the geometric effect. The crack mechanism and the fracture mechanism work together. First, the sensitive material of the sensor uses hydroxyethyl cellulose, a polymer that is easy to form a film, which can produce cracks during the stretching process of the device, causing the resistance of the sensor to change. During the stretching process, the length of the sensitive layer of the sensor increases and the cross-sectional area shrinks. According to the relationship between the resistance and its length and cross-section, it can be concluded that the device resistance tends to increase. Figure 6 a is a schematic diagram of the sensor being attached to human skin for physiological and sports testing. Due to the good flexibility of PDMS, it can fit the skin tightly. The molecular structure of MXene and PPy is shown in the enlarged part of the picture. The number of cracks on the sensitive material during stretching and the growth of crack density are shown in the figure. Figure 6 As shown in Figure b, as the length of the sensor increases, the number and degree of cracks increase to varying degrees, causing the resistance of the sensor to increase until the connection is lost.

[0106] Comparison of sensitivity of sandwich structure MXene / PPy / HEC flexible strain sensors:

[0107] According to the method of expressing the sensitivity of strain sensors, the sensitivity of strain sensors can be evaluated by the gauge factor, referred to as GF. GF is defined as:

[0108]

[0109] where R0 is the resistance in the initial state, R is the resistance in the stretched state, and ε is the mechanical strain applied to the sensor.

[0110] In order to prove that composite materials have more advantages than single materials, sensors based on MXene, PPy / HEC and MXene / PPy / HEC were developed. All sensors were sprayed multiple times during the production process. Figure 7 a is the strain-response curve of the sensor device based on a single conductive material and a composite material. The response value is defined as:

[0111]

[0112] It can be found that the strain sensor based on pure MXene has the highest sensitivity, which is much higher than the strain sensor device based on pure PPy / HEC. Although MXene has extremely high conductivity and sensitivity, it is easy to lose connection when stretched, resulting in a narrow sensor detection range. In order to ensure that the strain sensor can meet the requirements of human motion monitoring, the polymer PPy / HEC is introduced to increase the sensing range of the strain sensor. Figure 7 It can be seen from a that the addition of PPy / HEC significantly increases the sensing range of the strain sensor. Due to the good film-forming properties of PPy / HEC, the MXene / PPy / HEC sensor shows excellent sensing properties within the maximum strain range (over 80%). Under the same strain, the sensors of three different sensitive materials were tested for cyclic response, and the results are consistent with Figure 7 The response curves in a are consistent. As the strain increases, the relative resistance of the sensor changes significantly. When the strain is less than 12%, the GF is 118. When the strain increases from 12% to 94%, the GF increases to 357.5. As the crack gap and crack density increase slowly within the 12% strain range, the conductivity decreases slowly. As the applied strain increases (12%-94% strain), the relative resistance increases rapidly. This may be due to the geometric effect of the material. The increase in stretching causes the length of the sensitive material to increase and the cross-sectional area to decrease. At this stage, the crack gap continues to increase and the crack density increases rapidly, resulting in an increase in the specification factor of the sensor. Figure 7 d is the IV characteristic curve of the stress sensor. The IV curve of the sensor presents a perfect straight line relationship, which means that the sensor has perfect ohmic characteristics under different strains.

[0113] Dynamic performance test of sandwich structure MXene / PPy / HEC flexible strain sensor:

[0114] In order to verify whether the sandwich-structured MXene / PPy / HEC strain sensor can be stably applied in the field of human motion detection in the long term, it is necessary to test the dynamic performance of the sensor, such as response time, hysteresis characteristics, and repeatability. By setting the loading program of the test platform, the response and recovery time, repeatability, stretching at different speeds, stretching at different frequencies, and long-term stability of the strain sensor were tested.

[0115] Figure 8 a. Figure 8 b. Figure 8 c is the response diagram of the strain sensor under different strains of 1-60%. For the same strain, the sensor was tested for five cycles. The sensor was able to clearly distinguish different strain sizes and maintained excellent stability within five cycles.

[0116] Figure 8 d is the response and recovery time curve of the sensor. The response speed of the sensor is affected by the set speed in the tensile testing machine. The response recovery time of the sensor can reach the millisecond level.

[0117] Figure 8 e is the response curve of the sensor stretched at different frequencies. The sensor was stretched at different speeds of 10mm / min, 20mm / min, 40mm / min and 80mm / min under 10% strain. The effect of speed on the sensor response can be ignored, and the response cycle of the sensor shortens with the increase of stretching speed.

[0118] Figure 8 f shows the repeatability response of the strain sensor at different stretching speeds. Five repeated responses were tested at different speeds of 10 mm / min, 20 mm / min and 40 mm / min. The response curve shows that the device has good repeatability at different strains and the speed disturbance is negligible.

[0119] Figure 8 g shows the loading and unloading curves of the sensor. The loading and unloading processes are basically corresponding, and the sensor has no drift.

[0120] Figure 8 h shows the hysteresis curve of the sensor. The strain response curves of the sensor were obtained under 20%, 30%, 50%, 65% and 80% strain loading. It can be seen that the MXene / HEC / PPy strain sensor has high linearity and low hysteresis.

[0121] Figure 8i is the long-term stability test curve of the sensor. After a stability test of more than 18,000 seconds, it was found that the sensor could remain stable for more than 1,300 cycles, indicating that the device has excellent stability and repeatability and can meet the life requirements in practical applications.

[0122] Considering that the MXene / PPy / HEC strain sensor has the characteristics of simple preparation, wide detection range and sensitive response, it is compared with the reported flexible strain sensors in terms of strain range, GF value, preparation method and other aspects, as shown in Table 1. Its comprehensive performance is very outstanding compared with many current research works listed in Table 1, thus expanding its application range in detecting the full range of human body movements.

[0123] Table 1 Comparison of sensing performance of MXene / PPy / HEC strain sensor with existing sensors

[0124]

[0125] The sandwich-structured MXene / PPy / HEC sensor of the present invention has excellent sensing performances such as wide linear strain detection range (0-94%), high sensitivity (gauge factor up to 357.5), reliable repeatability (>1300 cycles) and ultrafast response-recovery time (300ms / 300ms). The MXene / PPy / HEC sensor can detect different levels of human physiological activities and shows excellent performance in measuring external strain changes and real-time motion detection.

[0126] Embodiment 3:

[0127] The invention discloses an application of a flexible strain sensor, wherein the flexible strain sensor can be applied to monitoring human motion signals.

[0128] Fig. 9The present invention provides a schematic diagram of the strain sensor preparation of an embodiment of the present invention and sensor response diagrams and mechanism descriptions in different application scenarios (pulse, motion, real-time monitoring of joint motion, and handwritten text recognition). The strain sensor is prepared by spraying MXene / PPy / HEC on the substrate PDMS through a simple and economical spraying process, and then encapsulating it with PDMS to prepare a flexible strain sensor with a sandwich structure. The sensor has great application potential, involving human motion conditions, health conditions, wireless detection, and handwritten text recognition. The motion conditions at different joints of the human body can be accurately captured by the sensor, and combined with Bluetooth technology, real-time detection can be achieved in smart devices. In addition, for weak pulse signals, the sensor also has excellent detection performance, which is reflected in the ability to distinguish different pulse waveforms better to quickly capture weak signals. At the same time, the sensor can be combined with an artificial intelligence algorithm as an intelligent handwriting board. In the present invention, a wavelet scattering network-long short-term memory neural network is used to identify the handwritten text signal collected by the strain sensor. The recognition process includes sensor data collection to input algorithm data processing and classification, and finally achieves high-precision recognition of handwritten text.

[0129] Fig.10 a. Fig.10 b Fig.10 c and Fig.10 d shows the sensors attached to human joints to detect signals generated by body movement. Strain sensors can be installed on wrists, elbows, knees, and fingers to identify a wide range of strains. The sensors can accurately identify different angles of joint bending and maintain stable output.

[0130] Fig.10 e. Fig.10 f is a strain sensor fixed at the finger joint to write letters. Obviously, the strain sensor is sensitive enough to detect the movement trend of the finger and produce corresponding responses to different letter writing signals. This is expected to be applied in the field of text recognition to help disabled people communicate.

[0131] Fig.10 g shows the response captured by the sensor at the joint when the finger performs a mouse click operation. For fast double-clicks and single-clicks, the sensor can sensitively and quickly collect signals.

[0132] Fig.10 h is the signal generated by the bending of the knee when a person is walking. By fixing the sensor at the knee of a person and walking, the sensor can overcome interference and detect a wide range of bending signals.

[0133] Embodiment 4:

[0134] An application of a flexible strain sensor, the flexible strain sensor can be applied to human health signal monitoring.

[0135] Fig.11 a is the pulse signal recorded by the strain sensor. Obviously, the strain sensor is sensitive enough to detect a typical pulse waveform (75 cycles / min), which is expected to help users adjust their exercise intensity and detect heart rate. Fig.11 b is an enlarged view of a single pulse signal within the dotted box in a. It can be seen that every beat of the pulse can be clearly captured as a knock wave, a tidal wave, and a diastolic wave, which provides an important reference for human health monitoring.

[0136] In addition to pulse monitoring, strain sensors can also be installed in the throat to detect voice recognition and swallowing status.

[0137] Fig.11 c is the response curve recorded when saying "Hi", "Sensor" and "Wonderful" respectively. The strain sensor can identify the weak movement in the throat associated with repeating "Hi", "Sensor" and "Wonderful". Since different words are pronounced differently, the vibration frequency and amplitude caused in the throat are also different, so the peak value of the detected curve is also different, which can be used to distinguish different words. Therefore, the excellent sound recognition ability of the strain sensor makes the sensor have great potential in voice rehabilitation exercises and human-computer interaction.

[0138] Fig.11 The figure de is the response diagram of the sensor when detecting human swallowing and chewing. Each swallowing will cause the throat to move, thereby squeezing the strain sensor to generate a pulse signal. In addition, attaching the strain device to the parotid gland can be used to detect the chewing of the human body ( Fig.11 e).

[0139] Fig.11 f is the breathing state of the volunteer. The flexible device is attached to the abdomen of the volunteer. Each breath will cause the abdomen to float slightly. The strain device attached to the abdomen will change its resistance regularly with the change of breathing frequency.

[0140] Embodiment 5:

[0141] Based on the application of MXene / PPy / HEC, the WTSN-LSTM algorithm was used to test the handwritten English words, Arabic numerals and Chinese characters of four volunteers. Combined with the wavelet scattering network-long short-term memory neural network, different handwritten text signals were successfully identified. The recognition accuracy was less affected by the writing habits of different people and could reach 96%. This shows that the device can be used as an intelligent handwriting board for handwritten text recognition and the protection of endangered languages.

[0142] Fig.12a. Fig.12 b. Fig.12 c. Fig.12 d. Fig.12 e and Fig.12 f is the response of the strain sensor to six different letters "A", "B", "C", "D", "P" and "U" written by four volunteers. For different letters, the position and period of the characteristic peaks of the test results are different, and the writing conditions of different volunteers are also different.

[0143] Fig.13 ad are the sensor responses for the remaining 26 letters excluding "A", "B", "C", "D", "P" and "U".

[0144] Fig.14 af is the time-frequency spectrum of the six letters "A", "B", "C", "D", "P" and "U".

[0145] Fig.15 a. Fig.15 b. Fig.15 c. Fig.15 d. Fig.15 e and Fig.15 f and Fig.16 a. Fig.16 b. Fig.16 c. Fig.16 d is the time-frequency spectrum of the ten Arabic numerals “0-9”.

[0146] Fig.17 a is the recognition flow chart of WTSN-LSTM algorithm. First, the response features of different texts are collected through sensors, and two layers of wavelet scattering are used to automatically extract features to reduce the dimension of the data. Then, a five-layer LSTM network with 100 hidden units is established, and the LSTM network is trained using scattering features and the network output is tested. Finally, different types of handwritten texts are successfully distinguished. The wavelet scattering network consists of wavelet convolution, nonlinearity and averaging. The wavelet scattering network automatically extracts features to obtain the time-frequency spectrum of the initial letter detection, and then the processed results are input into the long short-term memory neural network for recognition.

[0147] Fig.17 b is a three-dimensional color mapping diagram of the English letters written by four volunteers. The color of the curve is related to the response value. From this image, it can be seen that there are certain differences in the writing of different volunteers.

[0148] Fig.17c is the confusion matrix recognized by the six letters "A", "B", "C", "D", "P", and "U" written by Volunteer 1. After preprocessing the 360 sets of handwritten letter sequence data collected by the sensor, it is imported into the WTSN-LSTM network for learning and classification. The data recognition accuracy is 96.2%. Some of the letter "C" is recognized as "U", and the recognition accuracy of other letters is close to 100%.

[0149] Six letters, A, B, C, D, P, and U, are selected from the 26 English letters. They have different strokes or the same strokes with different bending arcs. After being handwritten and tested by four volunteers, the experimental results are shown in Fig.12 , and the six letters have different characteristic peaks, and there are also differences in the handwritten letters of the four volunteers. The four volunteers conducted a total of 3,000 sets of handwritten letter experiments, and the data was divided into a training set and a test set to obtain the confusion matrix shown in Figure 16d. From the results of the confusion matrix, it is not difficult to see that even for the differences brought by different people's handwriting, the recognition accuracy can still reach more than 96%.

[0150] Fig.17 e and Fig.17 f are the sensor response diagrams, three-dimensional waterfall diagrams (the color change is related to the response value), and the confusion matrix recognized by the algorithm obtained by a volunteer writing 600 sets of the ten Arabic numerals "0-9". The results show that for the numbers with a single stroke, the sensor can give different responses, and the recognition accuracy of the WTSN-LSTM network we constructed can reach 98%.

[0151] The sensor is no longer an independent unit, and the development of more and more related technologies has improved the intelligent level of the sensor. Fig.18 a is a schematic diagram of the three major application directions of the flexible strain sensor. In the present invention, the sensor can be used for the detection of the pronunciation of English letters, the recognition of handwritten letters, and the real-time detection of joint bending conditions. The sensor has the ability to recognize different English pronunciation situations, which lays a foundation for its later application in voice control. Combining with wireless devices for real-time human motion detection not only tests the long-term stability of the sensor but also poses challenges to its response range and response speed. After testing, the sensor can detect the bending of the human index finger in real time and transmit it to the smartphone APP. Combining with WTSN-LSTM, the handwritten text signals collected by the sensor are recognized, and high-precision detection of multiple languages can be achieved. When the human finger bends at different angles, the change of the sensor resistance signal is displayed in real time on the smartphone APP. In addition to the visual curve change, there is also the real-time change of the device resistance, which is of guiding significance for the real-time detection of the human body state.

[0152] When the volunteers handwrite the Chinese characters "中" and "国", the response change curves captured by the sensor are as shown in Fig.18c and Fig.18 d, the time spectrum of these two Chinese characters is displayed in Fig.18 It is not difficult to see that the signal differences between different Chinese characters are very large, and their frequency spectra can also distinguish the differences, which means that combined with machine learning algorithms, the recognition of Chinese characters is not a problem.

[0153] For the sandwich-structured MXene / PPy / HEC flexible strain sensor, the unique sandwich structure helps to improve the stability of the strain sensor. The sensor has excellent sensing performance such as a wide linear strain detection range (0-94%), high sensitivity (specification factor up to 357.5), reliable repeatability (>1300 cycles) and ultra-fast response-recovery time (300ms / 300ms).

[0154] Based on MXene / PPy / HEC flexible strain sensors, the human body's wide range of motion signals, especially the bending degree of each joint, is tested and clearly distinguished. Secondly, the pulse, sound and breathing conditions related to human health are tested, and the changes in tiny signals are accurately captured, which is expected to be applied to the monitoring of human health status.

[0155] Combined with a unique wavelet scattering network, the characteristics of the input handwritten letter time series signal are automatically identified, and then input into the long short-term memory neural network for classification and recognition. The flexible strain sensor accurately detects the characteristics of different handwritten characters, and is manifested in amplitude, peak width, period, number and position of characteristic peaks. The response results of the sensor are feature extracted to obtain the time-frequency spectrum of different handwritten characters, and then the results are identified. The confusion matrix results show that the flexible strain sensor has the potential to play a role in the field of handwritten letter recognition.

[0156] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for preparing a strain sensing film, wherein the strain sensing film is PPy attached to the surface of HEC to form a uniform PPy / HEC conductive film, and the PPy / HEC conductive film is distributed between MXene layers, the ultra-high conductivity of the MXene increases the sensitivity of the strain sensing film, and the PPy / HEC conductive film connects each MXene under stretching to ensure the connectivity of the conductive network on the strain sensing film, wherein the method is characterized in that: The preparation steps include: (1) Preparation of MXene Slowly add LiF to the hydrochloric acid solution, stir evenly, then slowly add Ti3AlC2 in several times, stir the solution in a 35°C water bath for 24 hours, wash the solution, centrifuge it with deionized water several times until there is no supernatant, ultrasonically treat the solution in an ice bath for 1 hour, then centrifuge it at 3500 rpm, and save the supernatant for later use; (2) Preparation of PPy / HEC solution HEC was added to a dimethylformamide solution and stirred vigorously at 25°C to completely disperse HEC. Pyrrole was added to HEC and stirred vigorously for 0.5 h to allow pyrrole to completely self-assemble on the surface of HEC. A mixed solution of FeCl3·6H2O and HCl was then added and reacted at 4°C for 8 h. After the reaction, the precipitate was washed with ethanol, deionized water, and 1 M HCl to remove monomers and oxidants. The washed product was redissolved in 1 M HCl to form a PPy / HEC solution. (3) Preparation of strain sensing films The MXene and PPy / HEC solutions are mixed and stirred to obtain MXene / PPy / HEC, and the MXene / PPy / HEC is sprayed on the surface of the substrate and dried to obtain a strain sensing film.

2. The method for preparing the strain sensing film according to claim 1, characterized in that: The mass ratio of LiF and Ti3AlC2 in (1) is 1:1, and the concentration of the hydrochloric acid solution is 9M.

3. The method for preparing the strain sensing film according to claim 1, characterized in that: The mass fraction of HEC in (2) is 0.2wt%, the volume ratio of HEC to dimethylformamide is 2:1, the mass ratio of HEC to pyrrole is 1:10; the mass ratio of pyrrole to FeCl3·6H2O is 1:1.89; the concentration of the hydrochloric acid solution is 1M, and the molar ratio of hydrochloric acid to pyrrole is 3:

1.

4. The method for preparing the strain sensing film according to claim 1, characterized in that: The volume ratio of MXene and PPy / HEC solution in (3) is 2:

1.

5. A flexible strain sensor for multi-language handwritten text recognition, wherein the sensor contains a strain sensing film, and the strain sensing film in the sensor is prepared according to the preparation method according to any one of claims 1 to 4, characterized in that: A flexible strain sensor with a sandwich structure was prepared by depositing MXene / PPy / HEC on the surface of the PDMS film by spraying.

6. The flexible strain sensor for multi-language handwritten text recognition according to claim 5 is characterized in that: The sensing principle of the sensor is: The sensing material is adsorbed on the surface of the PDMS film. When the sensor is stretched, cracks appear in the sensing layer, causing an increase in resistance. When the stretching is restored, the cracks disappear and the resistance of the strain sensor returns to its original state. PPy attached to the HEC surface forms a uniform film, which plays a good connecting role. Under small strain conditions, a small separation occurs between the MXenes, resulting in a significant change in the resistance of the sensing layer, causing the strain sensor based on MXene / PPy / HEC to maintain a high sensitivity under small strains. Under larger strains, the MXenes will be completely separated from each other, and the PPy / HEC conductive film acts as a bridge to connect the MXene to ensure the access to the sensing layer, greatly extending the sensing range of the strain sensor.

7. A method for preparing a flexible strain sensor for multi-language handwritten text recognition according to claim 5, characterized in that: (1) PDMS prepolymer and curing agent were uniformly mixed in a ratio of 10:1 and placed in a vacuum drying oven for 20 min to remove bubbles; (2) The mixed reagent is evenly coated on a glass slide and placed in a vacuum drying oven at 80° C. for 2 h to form a PDMS film; (3) Spraying MXene / PPy / HEC onto the surface of the PDMS film by spraying, and spraying and drying multiple times to form a strain sensing film; (4) Cut the PDMS film into long strips, install conductive tape on both sides, and lead out electrodes; (5) A layer of PDMS is further covered on the surface of the PDMS film with the strain sensing film, and the film is placed in an oven at 60°C for curing, thereby preparing a flexible strain sensor with a sandwich structure.

8. An application of a flexible strain sensor for multi-language handwritten text recognition according to any one of claims 5-6, characterized in that: The strain sensors are used to detect swallowing, voice recognition, breathing and joint flexion signals.

9. The application of the flexible strain sensor for multi-language handwritten text recognition according to claim 8 is characterized in that: The sensor measured and obtained the handwritten English words, Arabic numerals and Chinese character signals of four volunteers, and achieved recognition of handwritten text in multiple languages ​​through machine training and learning based on the wavelet time scattering network-long short-term memory neural network (WTSN-LSTM) algorithm, with recognition accuracy higher than 96%.

Images