Potential conduction type flexible pressure sensor based on multi-scale hierarchical electronic channel and application of potential conduction type flexible pressure sensor

Abstract

The invention discloses a potential conduction type flexible pressure sensor based on a multi-scale hierarchical electronic channel. The potential conduction type flexible pressure sensor comprises a metal anode layer, a PVDF-HFP-GO electrolyte layer, a multi-gradient MXene / rGO foaming film cathode layer and graphite paper loaded with a CNF isolation layer which are stacked from top to bottom. The invention relates to the technical field of flexible wearable electronics. The MXene / rGO foaming film with the three-dimensional gradient porous structure is of a layered structure in the microscopic aspect, the pore size is increased layer by layer, but the MXene / rGO foaming film is integrated in the macroscopic aspect, so that the MXene / rGO foaming film has the characteristics of lightness, thinness, small size, high flexibility and the like, and is very suitable for being applied to development of flexible wearable high-performance electronic devices. The artificial ear can be used as an artificial ear to detect sound waves and can also monitor mechanical stimulation generated by laryngeal muscle movement, and therefore accurate collection of voice signals is achieved. And in combination with machine learning, voice text content and emotion recognition can be realized, and the accuracy rates respectively reach 95.24% and 97.26%. And a man-machine voice emotion interaction interface is established on the basis.

Description

Technical Field

[0001] The present invention relates to the technical field of flexible wearable electronics, and in particular to a potential conductive pressure sensor (PTPS) based on a multi-scale hierarchical electron channel (MHEC) and its application as a flexible wearable sensor in realizing a human-machine intelligent emotional interaction system. Background Art

[0002] With the development of artificial intelligence, "perceptual intelligence" has gradually transformed into "cognitive intelligence" with the ability to understand and express, and giving machines emotions has become an inevitable trend. How to achieve machine recognition, understanding and further feedback of human emotions is the key to achieving "cognitive intelligence". At present, people have achieved preliminary recognition of human emotions by machines through facial expression recognition. However, it is difficult to collect and process facial expressions, and the amount of data is large and the equipment cost is high. In comparison, voice, as a carrier of human information exchange and emotional expression, is an extremely accurate and fast human-computer interaction tool, and humans can always easily judge the emotional state of others through their voices, which makes detecting voice signals and extracting emotional features an effective means to achieve "cognitive intelligence". However, traditional voice detection technology cannot meet the needs of special groups (deaf-mute people, Parkinson's disease patients, laryngeal surgery patients, etc.) and cannot cope with high noise and lack of sound media. It has hindered the further development of human-computer voice emotional interaction.

[0003] In fact, phonation involves mechanical movement of laryngeal muscles, which are less susceptible to external interference. Therefore, speech recognition can be made suitable for patients with dysphonia and extreme environments by realizing the perception of this mechanical stimulus. For the hearing-impaired, the detection of air pressure waves (sound waves) caused by sound transmission is helpful for the development of artificial ears. However, the realization of the above functions requires the development of high-performance sensors. Devices with flexible passive design are conducive to being worn on the throat. The detection of tiny sound waves requires the sensor to have high sensitivity within a small pressure range. The detection of laryngeal muscle movement requires a wide detection range under large pressure. The detection of phrases and long sentence signals in speech requires the sensor to have the ability to detect dynamic and static signals respectively. The linearity of the sensor response facilitates the qualitative, processing and analysis of speech signals. Therefore, there is an urgent need for sensors that combine the above advantages to realize speech recognition and human-computer emotional interaction. However, it is difficult for traditional sensors to perfectly meet the high requirements of the above applications in terms of response mechanism and structural design. Therefore, further innovation in the mechanism and structure of the sensor is urgently needed. In the field of speech research, some acoustic detectors have made preliminary explorations into speech emotion extraction, but there is no sufficient evidence to show that the mechanical stimulation produced by laryngeal muscle movement can extract information about the six major human emotions (happiness, anger, surprise, fear, disgust and sadness), and there is no experiment to prove its feasibility and completeness as a means of human-computer emotional intelligence interaction. Summary of the invention

[0004] The purpose of the present invention is to provide a potential conductive pressure sensor with a multi-scale hierarchical electronic channel and its application as a flexible wearable sensor in a human-machine intelligent emotional interaction system. Figure 1 ). The electron-ion type potential conduction mechanism gives the sensor the advantages of self-power, compatibility with dynamic and static pressure detection, fast response / recovery time (71ms), and high cycle stability (100,000 cycles). What is particularly critical is that in view of the problem that ion transport channels are difficult to achieve large-scale and high-precision regulation, the present invention proposes and proves that this multi-scale hierarchical electronic channel design can effectively regulate the coupled electron-ion transport under mechanical stimulation, and ultimately achieve codable electrical signal output. With the assistance of machine learning, the sensor's recognition accuracy for human voice content and emotions reached 95.24% and 97.26%, respectively. On this basis, a voice and emotion interaction interface between humans and robots was built, and the robot was controlled to make corresponding action feedback through short phrase instructions and long sentence instructions with different emotions. Specifically, the present invention is implemented using the following technical solutions:

[0005] The present invention discloses a potential-conducting flexible pressure sensor based on a multi-scale hierarchical electron channel, comprising a metal anode layer, a PVDF-HFP-GO electrolyte layer, a cathode layer, and graphite paper loaded with a CNF isolation layer stacked from top to bottom, wherein the cathode layer is a multi-gradient MXene / rGO foaming film material.

[0006] For the potential conduction type flexible pressure sensor, preferably, the metal anode layer is a Zn sheet.

[0007] For the potential conductive flexible pressure sensor, preferably, the PVDF-HFP-GO electrolyte layer is poly (vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP), zinc trifluoromethanesulfonate (Zn (OTf) 2 ) and graphene oxide (GO) to form a solid electrolyte membrane.

[0008] For the potential conductive flexible pressure sensor, preferably, the multi-gradient MXene / rGO foaming film material is formed by stacking a MXene / GO composite film I with a MXene to GO mass ratio of 30:6, a MXene / GO composite film II with a MXene to GO mass ratio of 30:12, and a MXene / GO composite film III with a MXene to GO mass ratio of 30:18 to obtain a multi-gradient MXene / GO composite film and then foaming; wherein the MXene mass ratio in the MXene / GO composite film I, the MXene / GO composite film II, and the MXene / GO composite film III is 1:1:1. The mass dosage of MXene is the same, and the pore size of the foaming film is determined by the content of GO. By controlling the same MXene and increasing the GO content, the gradient structure of the pore size increasing layer by layer as described in the present invention can be obtained.

[0009] Preferably, the MXene / GO composite film I is obtained by distributing the MXene / GO mixed solution I on the substrate by a casting method and drying; the MXene / GO composite film II is obtained by distributing the MXene / GO mixed solution II on the MXene / GO composite film I by a casting method and drying; the MXene / GO composite film III is obtained by distributing the MXene / GO mixed solution III on the MXene / GO composite film II by a casting method and drying. Specifically, after the MXene / GO composite film I with a mass ratio of 30:6 is dried, a layer of 30:12 MXene / GO composite film II is continuously cast thereon, and after drying, a layer of 30:18 MXene / GO composite film III is cast again, thereby obtaining a multi-gradient MXene / GO composite film, and then foaming is performed to obtain a multi-gradient MXene / rGO foaming film. The foaming is a common method, such as placing the obtained multi-gradient MXene / GO composite film in a glass culture dish, adding hydrazine hydrate, and sealing it at 85-95°C for 4-7h.

[0010] For the potential conductive flexible pressure sensor, the MXene / GO mixed solution is obtained by mixing a MXene solution and a GO solution. Preferably, the MXene solution can be prepared by using a commercial product or a common method. For example, the preparation method can be: first, add HCl solution (45 mL, 12 M) to deionized water (15 mL) to prepare HCl solution (9 M). Then, LiF (4.5 g) and MAX phase (Ti 3 AlC 2 ) powder (3 g) were weighed separately. Then, lithium fluoride was slowly added into the hydrochloric acid solution (9 M). After sufficient stirring, Ti 3 AlC 2The powder (3 g) was added to the mixed solution in several portions and then stirred at 35 °C for 24 h. The obtained sample was washed several times with deionized water until the pH value of the supernatant reached ≈6. The clay-like sediment obtained after centrifugation was treated by low-temperature ultrasonication for 1 h under flowing argon to delaminate the MXene nanosheets. Finally, the above mixture was centrifuged at low speed for 1 h to remove the unexfoliated MXene. The collected dark green suspension is the MXene nanosheet dispersion.

[0011] Preferably, the GO solution can be prepared by commercial products or by common methods. For example, the preparation method can be as follows: add 46 ml of concentrated H 2 SO 4 Adding NaNO 3 (1g) was stirred for 10 min until NaNO 3 After the solution is completely dissolved, add graphite (2 g) and continue stirring for 10 min. Then slowly add KMnO 4 (6g is added in 6 times, once every 10 minutes), and the temperature of the suspension is kept below 20℃, and stirring is continued for 30 minutes. React in a water bath at 35℃ for 4~5 hours until a gray thick paste or a cracked and hardened state is formed. Use a glass rod to assist the rotation of the magnet. Under the condition of vigorous stirring, continuously and slowly add deionized water (92mL) to stir the thick paste again. Place the beaker containing the solution in a water bath at 98℃ for 10 minutes. Then pour the solution into 300 ml (0.997moL / L) H 2 O 2 The remaining KMnO 4 The reaction was complete. Add H 2 O 2 After stirring evenly, add deionized water to about one liter, seal the beaker with plastic wrap and let it stand until it is completely precipitated, and remove the supernatant. Add deionized water to 240 ml and ultrasonicate for 2 minutes. The obtained sample is washed several times with deionized water and then ultrasonicated at low temperature for 5 minutes. Finally, the above mixture is centrifuged at low speed for 3 minutes, and the collected yellow suspension is the GO solution. There is no special requirement for the concentration of MXene solution and GO solution, as long as the mass amount of MXene and GO can meet the requirements. MXene can be obtained from MXene solution by the commonly used vacuum filtration and drying method, and GO can be obtained from GO solution by the commonly used vacuum drying method.

[0012] As a preferred method, the potential conductive flexible pressure sensor is composed of a metal anode Zn sheet layer, a PVDF-HFP-GO electrolyte layer, a cathode layer of a multi-gradient MXene / rGO foaming film material, and a graphite paper loaded with a CNF isolation layer, which are stacked and packaged from top to bottom to form a sandwich structure. The graphite paper serves as the lead-out electrode of the cathode.

[0013] In addition, the potential conduction flexible pressure sensor described in the present invention can be used in sound wave detection, and it can also be used as an artificial eardrum. Further, it can be used in voice signal monitoring and recognition. Furthermore, it can be used in human-computer emotional intelligent interaction systems.

[0014] Furthermore, the present invention also provides a human-computer emotional intelligent interaction system, which includes:

[0015] Potentially conductive flexible pressure sensors (PTPS), which can collect speech signals by detecting sound waves or by monitoring laryngeal muscle vibrations;

[0016] The microprocessor processes the voice signals collected by the PTPS and recognizes the input voice signals through a pre-trained algorithm;

[0017] The robot arm completes the instructions corresponding to the voice content and emotional information recognized by the microprocessor, thereby realizing human-machine emotional intelligent interaction;

[0018] A speaker module, used for providing sound wave signals;

[0019] A signal filtering and amplifying module, used for amplifying and filtering the signal;

[0020] The power module is used to provide electrical energy to the signal filtering and amplifying module, the microprocessor and the manipulator.

[0021] In order to facilitate a more detailed understanding of the present invention, the mechanism of the present invention is explained as follows:

[0022] On the one hand, the present invention provides a multi-scale hierarchical electronic channel, which can achieve the output of encodable electrical signals by regulating the electron-ion coupled transport. First, Figure 2 The response mechanism of the sensor is demonstrated. When Zn, MXene / rGO electrode and PVDF-HFP-GO electrolyte are in contact, Zn as the anode loses electrons and is converted into Zn 2+ , gather on the anode surface, attracting OTf in the electrolyte - Electrons are transferred to the anode surface through the external circuit to the MXene / rGO foam film electrode. The MXene / rGO foam film is used as the cathode, and Zn 2+Combined with electrons, the sensor can output a stable electrical signal due to the existence of this redox reaction. Figure 3 As shown in a, under no stress, the electron transmission channel is sparse and lengthy, the electron transmission speed is very difficult and slow, the reaction on the anode surface will be suppressed, and the cathode has fewer electrons, and its surface Zn 2+ The combination of Zn and electrons is almost interrupted. It can be seen that too slow electron transmission will reduce the interface reaction rate of the electrode. 2+ The distribution of electrons on the cathode surface is small, and the driving ability for ions is weak. - Transport to the anode surface and Zn 2+ The transport process to the cathode surface becomes extremely slow, indicating that the slow transmission of electrons will also slow down the transmission of ions. Figure 3 As shown in b and c, when the stress applied to the sensor gradually increases, the multi-scale hierarchical electron channel structure changes, the electron channels increase, and the electron transmission path shortens. This results in faster transmission of electrons in the external circuit. At the anode, Zn loses electrons to produce Zn 2+ The reaction is accelerated. At the cathode, more electrons react with Zn 2+ Reaction, Zn 2+ The embedding speed of Zn 2+ The number of electrons distributed on the cathode surface increases, the driving ability of ions is enhanced, and the transmission of ions is accelerated. In summary, it is the process of coupled electron-ion transport. Therefore, when the sensor detects voltage, the equivalent circuit diagram is shown in 3d. The blue part is the internal schematic diagram of the sensor, and the yellow part is the schematic diagram of the detection circuit. The measured voltage V oc It will change with the change of the overall internal resistance of the circuit. The specific formula is as follows:

[0023]

[0024] Where E 1 is the intrinsic potential difference between the two electrodes of Zn and MXene / rGO foam film. Thermodynamically, once the ion concentration of the electrolyte is fixed, E 1 It is fixed. v is the internal resistance of the voltmeter, V oc is the voltage obtained from the test, R c is the contact resistance between the MXene / rGO foam film and the extraction electrode, R foam is the resistance of MXene / rGO foam film, R ion is the impedance of ion transport. From the coupled electron-ion transport, we know that R ion Will follow R foam and R cWhen there is no stress, the internal resistance of the sensor is close to infinity, so the measured voltage V oc is 0. Under stress, due to the effect of coupled electron-ion transport, the overall impedance inside the sensor decreases under pressure, and the detected voltage V is obtained from formula (1): oc It will increase with the increase of pressure. When the internal resistance of the sensor is much smaller than R v When the measured voltage approaches E 1 .

[0025] On the other hand, in order to understand the regulation of mechanical stimulation on coupled electron-ion transport, electrochemical impedance spectroscopy (EIS) analysis was performed on the sensor (in order to make the coupling effect of electron transport on ion transport more obvious, no isolation layer was added to the test device). Figure 3 e is the Nyquist diagram of the sensor under different pressure conditions and the corresponding equivalent circuit diagram. Ω Usually refers to the intercept of the high-frequency curve on the horizontal axis, representing the internal resistance of the device. Figure 3 It can be seen that R Ω The value decreases with the increase of external pressure. This is mainly due to the increase of electron transmission channels and the acceleration of electron speed due to the existence of multi-scale hierarchical electron channels under pressure, and the acceleration of ion diffusion due to the existence of coupled electron-ion transport. ct represents the arc area, which refers to the charge transfer resistance in the electrode / electrolyte interface. As the pressure increases, the micro-arc length decreases, indicating that under the stimulation of external pressure, due to the influence of coupled electron-ion transport, the electron transfer speed accelerates the transfer of ions at the electrode / electrolyte interface. This process can be represented by a parallel circuit of resistance and capacitance. This also increases the ionic conductivity σ of the device. The ionic conductivity σ is calculated as follows:

[0026]

[0027] Where d is the thickness of the sample to be tested (cm), S is the effective area of ​​the electrode (cm 2 The calculation result is as follows. Figure 3 f. The oblique line appears in the low frequency region, which is related to the diffusion process of ions in the electrolyte. This process can be expressed by the Warburg impedance Z W As the pressure increases, the slope of the oblique line increases, which verifies that the diffusion process of ions in the electrolyte is accelerated due to the faster electron transmission speed. Figure 3As shown in Figure g, from the potential balance, it can be seen that when the current is 0, the IV curve intersects at one point, indicating the existence of the intrinsic potential difference between the two electrodes of Zn and MXene / rGO foam film, and it is fixed. When the voltage input to both ends of the sensor is 0, the current in the circuit still exists and increases with the increase of pressure. It confirms the existence of redox reaction and the regulation of coupled ion-electron transport under mechanical stimulation to achieve the output of encodable electrical signals. As the pressure increases, the slope of the IV curve increases, indicating that the overall impedance in the circuit decreases with the increase of pressure. This is consistent with the change in the overall circuit impedance caused by coupled electron-ion transport modulated by multi-scale hierarchical electronic channels.

[0028] Thirdly, the present invention realizes precise control of the density and pore size of the three-dimensional structural system of the composite material. Figure 4 is the thickness of the composite film before and after foaming. Compared with the film before foaming, the film thickness is mainly concentrated in 7.12-13.2 μm, the thickness after foaming can be adjusted between 93.87-152.16 μm, and the gradient film is from 67.4 μm to 183.16 μm. The density of MXene / rGO foam film gradually decreases with the increase of GO (from 407 mg cm −3 Reduced to 224 mg cm −3 )( Figure 5 ), the change in pore size showed an opposite trend (from 4.33 μm to 8.27 μm) ( Figure 6 With the addition of GO, the content of oxygen-containing functional groups in the MXene / GO composite membrane increased, and the foaming property of the composite membrane increased significantly, while maintaining a high porosity of more than 66.24% ( Figure 7 ). Therefore, by simply adjusting the content of oxygen-containing functional groups in the composite film, the density and pore size of MXene / rGO can be freely adjusted while maintaining high porosity. Combined with actual needs, we can prepare foaming films that meet specific requirements.

[0029] Fourthly, in the absence of pressure, the CNF nanofiber network prevents electron transport. Under slight mechanical stimulation, the MXene / rGO foam film electrode contacts the lead-out electrode (graphite paper), producing a certain high-sensitivity response output. This allows accurate detection of small signals, but cannot detect higher pressures. The MXene / rGO foam film has a micron-scale three-dimensional (3D) porous structure, which can generate a large number of electron channels under mechanical stimulation. However, the sensor based on the single-ratio MXene / rGO foam film has a narrow detection range and poor linearity, which cannot meet the application requirements. Therefore, we designed a three-dimensional gradient porous structure MXene / rGO foam film (GPMF) as the electrode material of the sensor, and combined it with the CNF nanofiber network to construct a multi-scale hierarchical electron channel MHEC. GPMF (the mass ratios of MXene / GO are 30:6, 30:12 and 30:18, respectively) is a three-dimensional gradient structure, and the pore size increases layer by layer ( Figure 8 ). The equivalent circuit is shown in Figure 9, where R 1 , R 2 and R 3 represents the resistance of each layer in the gradient film (from bottom to top, the mass ratios of MXene / GO are 30:6, 30:12, and 30:18, respectively). In addition, V 1 、V 2 and V 3 Respectively represent the voltage response of sensors using single ratio (30:6, 30:12 and 30:18) MXene / rGO foam films under pressure. First, due to the presence of the isolation layer, high sensitivity can be achieved at low pressure, and Rc decreases. Second, under pressure, the pores of the single ratio MXene / rGO foam films are closed synchronously, and V 1 、V 2 and V 3 The pores in the gradient structure will gradually close from small to large, continuously and linearly generating electron channels, R 1 , R 2 and R 3 Decreases successively, and finally V reaches saturation, and the response range is very wide ( Fig.10 ). In short, the sensor with MHEC has high sensitivity at low pressure and a wide linear response range at high pressure (Figure 11a), thus meeting the requirements of voice signal detection and human-computer emotional interaction.

[0030] The fifth aspect, Fig.11 b shows the response of MHEC-based PTPS under different pressures, proving that it can accurately and sensitively identify the response of different pressures. The sensor can effectively identify different frequencies ( Fig.11c) and the signal is stable. The sensor's response recovery speed is also very fast ( Fig.11 d). When no force is applied, the electron transmission in the sensor circuit is blocked and the energy consumption is extremely low. Therefore, the sensor has excellent cycle stability and can be cycled 105,000 times at a pressure of 300 kPa with almost no attenuation ( Fig.11 e), which ensures its long-term use as a sound detector.

[0031] Sixthly, human language contains not only basic text information, but also human emotions. People often express their emotions when communicating with others. The same sentence may have different meanings depending on the personal emotions mixed in it. Based on machine learning, PTPS can detect throat muscle movements and realize voice content and emotion recognition. On this basis, an interactive system between humans and robots is built, and the robot can give feedback based on the voice content and emotions.

[0032] The beneficial effects of the present invention are as follows:

[0033] 1. The current potential conduction mechanism is to achieve the output of coded electrical signals by regulating the ion transmission channel, so it is difficult to achieve large-scale and high-precision regulation. The present invention uses the nano-scale CNF fiber network as an isolation layer to achieve high sensitivity under low pressure, and the micron-scale three-dimensional gradient porous structure MXene / rGO foam film can achieve a wide linear detection range under high pressure. This meets the needs of speech detection.

[0034] 2. The integrated three-dimensional gradient porous structure MXene / rGO foam film of the present invention is different from the previous gradient porous materials. The gradient MXene / rGO foam film is a layered structure at the microscopic level, and the pore size increases layer by layer, but it is integrated at the macroscopic level. Therefore, it has the characteristics of lightness, small size and high flexibility, and is very suitable for the development of flexible wearable high-performance electronic devices.

[0035] 3. The present invention can be used as an artificial ear to detect sound waves and monitor the mechanical stimulation generated by the movement of the throat muscles, thereby realizing the accurate collection of voice signals. Combined with machine learning, it can realize the recognition of voice text content and emotion, with an accuracy rate of 95.24% and 97.26% respectively. On this basis, a human-computer voice emotion interaction interface was built. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The composition structure diagram of PTPS. In the figure, 1. Zn sheet; 2. PVDF-HFP-GO solid electrolyte; 3. 3D gradient structure MXene / rGO foam film; 4. CNF nanofiber network; 5. Graphite paper.

[0037] Figure 2 Diagram of the response mechanism of PTPS.

[0038] Figure 3 a is the principle of coupled electron-ion transport. Figure 3 b is the coupled electron-ion transport process under low pressure. Figure 3 c is the coupled electron-ion transport process under high pressure. Figure 3 d is the equivalent circuit diagram of the sensor using MXene / rGO foam film when detecting pressure. Figure 3 e is the electrochemical impedance spectroscopy (EIS) of the sensor under different pressures. Figure 3 f is R Ω and σ (ionic conductivity) change with increasing pressure. Figure 3 g is the change of IV curve of the sensor under different pressures.

[0039] Figure 4 Comparison of the thickness of MXene / GO composite films with different mass ratios and gradient structures before and after foaming.

[0040] Figure 5 is the mass density of MXene / GO film and MXene / rGO foam film with different mass ratios.

[0041] Figure 6 The average pore size and pore distribution of MXene / rGO foam films with different mass ratios.

[0042] Figure 7 Calculation of the porosity of MXene / rGO foam films with different mass ratios and gradient structures.

[0043] Figure 8 The cross-sectional SEM image of the gradient MXene / GO foam film and the enlarged local SEM images of each layer.

[0044] Fig. 9 This is the equivalent circuit diagram of the sensor using MHEC when detecting pressure.

[0045] Fig.10 It is the sensor's regulation mechanism for the output voltage.

[0046] Fig.11a is the electromechanical response curve of MXene / rGO foaming film with different mass ratios and gradient MXene / GO film before and after foaming. 11b is the vt curve of MHEC-based PTPS under different pressures. 11c is the sensor recognition of different frequencies at 300kPa. 11d is the fast response and recovery curve of MHEC-based PTPS at 2kPa and 400kPa. 11e is 100,000 cycles of MHEC-based PTPS at 300kPa.

[0047] Fig.12 The response signal corresponding to each word.

[0048] Fig.13 Schematic diagram for building a human-computer emotional intelligent interaction system.

[0049] Fig.14 Sound detection system built for PTPS devices.

[0050] Fig.15 Schematic diagram of using PTPS device as sound detector to detect sounds of animals and nature. DETAILED DESCRIPTION

[0051] The following embodiments are further descriptions of the content of the present invention as an explanation of the technical content of the present invention, but the essential content of the present invention is not limited to the following embodiments. Ordinary technicians in this field can and should know that any simple changes or replacements based on the essential spirit of the present invention should fall within the scope of protection required by the present invention.

[0052] Example 1

[0053] A method for preparing a PTPS device based on multi-scale hierarchical electron channels comprises the following steps:

[0054] Step (1) Preparation of MXene solution:

[0055] First, HCl solution (9 M) was prepared by adding HCl solution (45 mL, 12 M) into deionized water (15 mL). Then, LiF (4.5 g) and MAX phase (Ti 3 AlC 2 ) powder (3 g) were weighed separately. Then, lithium fluoride was slowly added into the hydrochloric acid solution (9 M). After sufficient stirring, Ti 3 AlC 2The powder (3 g) was added to the mixed solution in several portions and then stirred at 35 °C for 24 h. The obtained sample was washed several times with deionized water until the pH value of the supernatant reached ≈6. The clay-like sediment obtained after centrifugation was treated by low-temperature ultrasonication for 1 h under flowing argon to delaminate the MXene nanosheets. Finally, the above mixture was centrifuged at low speed for 1 h to remove the unexfoliated MXene. The collected dark green suspension is the MXene nanosheet dispersion.

[0056] Step (2) Preparation of GO solution:

[0057] Under stirring in an ice-water bath, add 46 ml of concentrated H 2 SO 4 Adding NaNO 3 (1g) was stirred for 10 min until NaNO 3 After the solution is completely dissolved, add graphite (2 g) and continue stirring for 10 min. Then slowly add KMnO 4 (6g is added in 6 times, once every 10 minutes), and keep the temperature of the suspension below 20℃, and continue stirring for 30 minutes. React in a water bath at 35℃ for 4~5 hours until a gray thick paste or a cracked and hardened state is formed. Use a glass rod to assist the rotation of the magnet. Under the condition of vigorous stirring, continuously and slowly add deionized water (92 mL) to stir the thick paste again. Place the beaker containing the solution in a water bath at 98℃ for 10 minutes. Then pour the solution into 300 ml (0.997 mol / L) H 2 O 2 The remaining KMnO 4 The reaction was complete. Add H 2 O 2 Stir evenly, add deionized water to about one liter, seal the beaker with plastic wrap and let it stand until it is completely precipitated, then remove the supernatant. Add deionized water to 240 ml and ultrasonicate for 2 minutes. Wash the obtained sample with deionized water several times and then ultrasonicate at low temperature for 5 minutes. Finally, centrifuge the above mixture at low speed for 3 minutes, and the collected yellow suspension is the GO solution.

[0058] Step (3) The MXene solution obtained in step (1) and the GO solution obtained in step (2) are mixed according to the MXene to GO mass ratio (30:6, 30:12, 30:18, in mg) to obtain MXene / GO mixed solution I, MXene / GO mixed solution II, and MXene / GO mixed solution III, respectively.

[0059] Step (4) The MXene / GO mixed solution I obtained in step (3) is evenly distributed on a glass substrate by a casting method and naturally dried to obtain a MXene / GO composite film I.

[0060] Step (5) Preparation of multi-gradient MXene / rGO foam film:

[0061] On the MXene / GO composite membrane I with a mass ratio of 30:6 obtained in step (4), a MXene / GO mixed solution II with a mass ratio of 30:12 was evenly distributed by a casting method. After natural drying, a MXene / GO mixed solution III with a mass ratio of 30:18 was added by the same method and dried to obtain a multi-gradient MXene / GO composite membrane. The composite membrane was then placed in a glass culture dish, hydrazine hydrate was added, and the mixture was sealed and foamed at 90°C for 6 hours to obtain a gradient MXene / rGO foamed membrane.

[0062] Step (6) Preparation of CNF isolation layer:

[0063] The original CNF solution was diluted with a mixture of alcohol and water in a volume ratio of 2:1 and shaken for 30 min. Then, the diluted CNF was evenly coated on the graphite paper and dried on a heating table at 50 °C to form a CNF isolation layer.

[0064] Step (7) Preparation of PVDF-HFP-GO electrolyte:

[0065] Poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP), graphene oxide (GO) and Zn(OTF) 2 Dissolved in dimethylformamide (DMF) at a mass ratio of 10:1:1. The mixed solution was stirred at 40°C for 6 hours, ultrasonicated for 10 minutes, then cast on a glass plate and dried at room temperature. A PVDF-HFP-GO (solid electrolyte) membrane was obtained.

[0066] Step (8) Preparation of PTPS;

[0067] The multi-gradient MXene / rGO foam film is used as the cathode, PVDF-HFP-GO is used as the electrolyte, the Zn sheet is used as the anode, and the graphite paper with a CNF isolation layer is used as the lead-out electrode of the cathode. The four are stacked and packaged from top to bottom to form a "sandwich" structured potential transduction pressure sensor PTPS.

[0068] In Example 1, the assembled sensor is attached to the throat of the experimenter, and the experimenter says different words: 'wearable', 'electronic', 'self-powered', 'flexible', 'potentiometric', 'pressure', 'sensor'. The speech signals corresponding to the detected different words are preprocessed to remove clutter, and then the signal features are extracted and input into the neural network, and the recognition results are obtained by combining machine learning technology. It is worth noting that when a word is pronounced repeatedly, the signal pattern generated is highly stable and repeatable. When different words are pronounced, the corresponding sensor response signals show obvious differences. This is because different word pronunciations will cause unique vocal cords. , which results in different pressure changes in the sensor. More specifically, the intensity of the signal change is related to the syllable In addition, the signal of each word is composed of several characteristic peaks. The peak value of the characteristic peak corresponds to the volume of the syllable. The interval and peak width between the characteristic peaks indicate the pause and duration of each syllable during pronunciation. Since the pronunciation of each word is different, the response signal of each word has its own unique characteristic peak-to-peak value, peak width, and peak interval ( Fig.12 ). In addition to the qualitative analysis of the above sensor signals, a BP neural network based on a machine learning framework is further used to parse, classify and predict the pronunciation signal patterns obtained from different words. The BP neural network is divided into two parts: BP (back propagation) and neural network. It mainly uses back propagation to continuously adjust the weights and thresholds of the neural network to minimize the sum of square errors of the network and ultimately achieve accurate recognition. In order to establish a training data set for the BP neural network model, we collected response data from 1,400 repeated utterances of the above seven words, and extracted three characteristic parameters including the characteristic peak-to-peak value, peak width, and peak interval of the signal. Due to the inconsistent length of the extracted feature data, a zero-filling mode is used. The data set is divided into a training data set and a (Training dataset: test dataset = 8:2), the training dataset is used to train the BP neural network machine learning framework. Finally, the test dataset is used to predict and observe the network performance.

[0069] Example 2

[0070] PTPS device based on multi-scale hierarchical electronic channels to build a human-computer emotional interaction system ( Fig.13), the PTPS device in Example 1 has a good detection and recognition ability for voice signals, so its detection of voice emotion signals is further explored, and based on this, emotional intelligent interaction between people and manipulators is realized. Using 'three', 'two', and 'fist' as phrase instructions, the manipulator recognizes the text instructions contained in them and makes corresponding actions. Using 'please, open the door' as a long sentence instruction, with two emotions of anger and happiness, the manipulator can make a gesture of rejection for instructions in an angry tone, and a gesture of consent for instructions in a happy tone. First, the sensor is attached to the throat of the experimenter, and then the two output electrodes of the sensor are connected to the Filter-Amplifier. Then connect the Filter-Amplifier, Arduino development board, and computer. Then, the experimenter repeats each instruction 200 times, thereby collecting a large number of voice signals as labeled data sets. These labeled data sets are cleaned and transmitted to the neural network model, and the data sets are analyzed, randomly split, and input and output are assigned in the neural network. Because the sensor has high sensitivity and linear working range, the time domain features of these data sets can be well extracted by the neural network. In order to change the linear operation of pure convolution, an activation function is added after each convolution layer. Neural activity in nature is almost achieved by activating some neurons and inhibiting some neurons. At the same time, the activation and inhibition states of neurons can switch with each other under certain conditions. The activation function is equivalent to bringing nonlinear features to the neural network, so that the features extracted by the neural network can adapt to the fluctuations of complex environments. Finally, the fully connected layer uses the Softmax activation function to integrate all feature information so that one output result corresponds to one input. According to the above steps, a total of 1,000 voice data were collected, and 5 categories ('three', 'two', 'fist', 'please, open the door (happy)', 'please, open the door (angry)') were divided. After the five types of gesture data were extracted and classified in the hidden layer of the neural network, the gradient descent optimization algorithm Adam (adaptive moment estimation) was used to reduce the micro-processing memory requirements and reduce gradient noise interference, because the deep learning application is limited to a certain extent by the learning rate hyperparameter update strategy. After the neural network model is trained, a real-time recognition platform is built based on the Arduino development board. When the experimenter says one of the sentences, the signal is transmitted to the Arduino development board through the Filter-Amplifier. The text instructions and emotions contained in the sentence are recognized according to the previously trained neural network model, so as to control the robot to make corresponding actions in response, thus achieving the purpose of human-computer interaction.

[0071] Example 3

[0072] Numerous studies have shown that natural sounds are very important for human mental health. For example, the sound of running water can increase positive emotions, and the chirping of birds can reduce stress. This means that human health and life experience can be improved by managing natural sounds. The principle of using MHEC-based PTPS to distinguish natural sounds is the same as that of the eardrum. Using the PTPS device in Example 1 to build a sound detection system, as shown in Figure 1, Fig.14 , the PTPS device sensor was connected to the computer's speakers and its performance was tested by playing audio signals. The sensor's response signals to different biological sounds and physical sounds showed its excellent detection performance as a sound detector, which can identify animal acoustic activities and monitor natural sound changes. Different biological sounds (birds, cats and dogs) and some physical sounds (wooden fish, thunder and rain), their response signals are as follows (ac) and (df) shown. Fig.15 In the figure, the blue curve is the acoustic wave electrical signal detected by using the PTPS device of the present invention as a sound sensor. It is not difficult to find that the sensor can respond quickly to both dense and noisy biological signals and constantly changing natural physical sounds, and the characteristic peaks are easy to identify. Recently, natural sounds have been shown to help alleviate the increasingly serious mental health and anxiety problems of modern people, and it is easy to infer that future research will focus on exploring the impact of sound on health. Therefore, the sound detector lays the foundation for the realization of the above preliminary scheme.

[0073] It should be noted that the above-mentioned technical contents of the present invention are only for explanation and clarification to enable those skilled in the art to understand the technical essence of the present invention, so the technical contents are not used to limit the substantial protection scope of the present invention. The substantial protection scope of the present invention shall be subject to the claims. Those skilled in the art should know that any modification, equivalent substitution and improvement based on the substantial spirit of the present invention shall be within the substantial protection scope of the present invention.

Claims

1. A potential-conducting flexible pressure sensor based on multi-scale hierarchical electron channels, comprising a metal anode layer, a PVDF-HFP-GO electrolyte layer, a cathode layer, and a graphite paper loaded with a CNF isolation layer stacked from top to bottom, wherein the cathode layer is a multi-gradient MXene / rGO foam film material.

2. The potential conductive flexible pressure sensor according to claim 1, characterized in that: The metal anode layer is a Zn sheet.

3. The potential conductive flexible pressure sensor according to claim 1, characterized in that: The PVDF-HFP-GO electrolyte layer is a solid electrolyte membrane formed by poly (vinylidene fluoride-co-hexafluoropropylene), zinc trifluoromethanesulfonate, and graphene oxide.

4. The potential conductive flexible pressure sensor according to claim 1, characterized in that: The multi-gradient MXene / rGO foaming film material is formed by stacking a MXene / GO composite film I with a MXene to GO mass ratio of 30:6, a MXene / GO composite film II with a MXene to GO mass ratio of 30:12, and a MXene / GO composite film III with a MXene to GO mass ratio of 30:18 to obtain a multi-gradient MXene / GO composite film and then foaming; the MXene mass ratio in the MXene / GO composite film I, the MXene / GO composite film II, and the MXene / GO composite film III is 1:1:

1.

5. The potential conductive flexible pressure sensor according to claim 4, characterized in that: The MXene / GO composite membrane I is obtained by distributing the MXene / GO mixed solution I on a substrate by a tape casting method and drying; the MXene / GO composite membrane II is obtained by distributing the MXene / GO mixed solution II on a MXene / GO composite membrane I by a tape casting method and drying; the MXene / GO composite membrane III is obtained by distributing the MXene / GO mixed solution III on a MXene / GO composite membrane II by a tape casting method and drying.

6. The potential conductive flexible pressure sensor according to claim 4, characterized in that: The foaming step is to place the obtained multi-gradient MXene / GO composite film in a culture dish, add hydrazine hydrate, and then seal the dish and foam it at 85-95° C. for 4-7 hours.

7. The potential conductive pressure sensor according to any one of claims 1 to 6, which is a sandwich structure formed by stacking and encapsulating a metal anode Zn sheet layer, a PVDF-HFP-GO electrolyte layer, a cathode layer of a multi-gradient MXene / rGO foam film material, and a graphite paper loaded with a CNF isolation layer from top to bottom.

8. Application of the potential conduction flexible pressure sensor according to claim 1 in acoustic wave detection.

9. Use of the potential conduction flexible pressure sensor according to claim 1 as an artificial eardrum.

10. Application of the potential conductive flexible pressure sensor according to claim 1 in voice signal monitoring and recognition.

11. Application of the potential conductive flexible pressure sensor described in claim 1 in a human-computer emotional intelligent interaction system.

12. A human-computer emotional intelligent interaction system, comprising: The potential conductive flexible pressure sensor of claim 1 can collect speech signals by detecting sound waves or by monitoring laryngeal muscle vibrations; A microprocessor processes the voice signal collected by the potential conductive flexible pressure sensor and recognizes the input voice signal through a pre-trained algorithm; The robot arm completes the instructions corresponding to the voice content and emotional information recognized by the microprocessor, thereby realizing human-machine emotional intelligent interaction; A speaker module, used for providing sound wave signals; A signal filtering and amplifying module, used for amplifying and filtering the signal; The power module is used to provide electrical energy to the signal filtering and amplifying module, the microprocessor and the manipulator.

Images