A gesture recognition sensor based on a triboelectric nanogenerator and a preparation method thereof

Abstract

The present invention discloses a gesture recognition sensor based on a triboelectric nanogenerator and a preparation method thereof, which relates to the technical field of gesture recognition and solves the technical problem of poor wearing comfort of the gesture recognition sensor. The key points of the technical solution are that the sensor includes an Ecoflex layer, a metal electrode layer, and an insulating support layer. Based on the principle of electrostatic induction and friction, the contact and separation of the back-of-hand skin and the sensor can generate electrical signals of a certain mode. The sensor can obtain multiple groups of different electrical signals generated by the deformation of the skin at different positions on the back of the hand through an array arrangement on the back of the hand. The electrical signals can accurately identify specific gestures after being processed and analyzed by machine learning. The application of the array arrangement method not only reduces the size of a single sensor and improves the wearing comfort, but also can obtain more motion signals from different positions, greatly improving the accuracy of gesture recognition.

Description

Technical Field

[0001] The present application relates to the technical field of gesture recognition, specifically to flexible self-powered intelligent hardware sensing technology, and particularly to a gesture recognition sensor based on a triboelectric nanogenerator and a preparation method thereof. Background Art

[0002] With the development of society and technology, gesture recognition sensors have unique requirements in various scenarios. Classified according to the driving form, they can be divided into two categories: battery-driven type and self-powered type. For battery-driven sensors, due to the need for an additional power supply, their volume is often large. For self-powered sensors, although they do not require battery power supply, in order to obtain relatively accurate measurement data, the required installation form often still causes certain obstacles to the movement of human body parts, and gesture recognition is mostly achieved in the form of complex gloves, finger cots, etc., which is not convenient for the free bending movement of fingers. Therefore, at present, there is still a large room for improvement in the wearing comfort of wearable gesture recognition sensors. Summary of the Invention

[0003] The present application provides a gesture recognition sensor based on a triboelectric nanogenerator and a preparation method thereof, and its technical purpose is to provide a simple preparation method, low cost, little influence on hand movement, and easy to be widely applied in gesture recognition scenarios.

[0004] The above technical purpose of the present application is achieved through the following technical solutions:

[0005] A gesture recognition sensor based on a triboelectric nanogenerator, the gesture recognition sensor is a sensor array, including at least 3 sensors, each sensor includes an Ecoflex layer, a metal electrode layer, a connecting electrode layer, a sealing film layer, an insulating support layer and a PTFE layer. The Ecoflex layer is a "concave" structure with protrusions at both ends, and the bottom of the Ecoflex layer is connected to the top of the insulating support layer; the metal electrode layer includes 2 metal electrode layers distributed left and right and are both located on the top of the Ecoflex layer. One end of the 2 metal electrode layers close to each other is connected through the connecting electrode layer, and one end of the 2 metal electrode layers far from each other is arranged in the protrusions at both ends of the Ecoflex layer; the bottom of the connecting electrode layer is provided with a sealing film layer, and both ends of the sealing film layer are respectively connected to one end of the 2 metal electrode layers close to each other; the bottom of the insulating support layer is provided with a PTFE layer, and the bottom of the insulating support layer is directly pasted on the skin, and the PTFE layer is also pasted on the skin.

[0006] A method for preparing a gesture recognition sensor as described above includes: mixing the A and B bottles of Ecoflex in a ratio of 1:1 and stirring well, then pouring the Ecoflex mixture into a mold and waiting for 4 to 6 hours to cure it; subsequently, sticking two copper foils of the same size side by side on the surface of the cured Ecoflex, placing a fixing block slightly smaller than the total covered area of the two copper foils in the center on the surface of the copper foils, and pouring a small amount of Ecoflex mixture around it and waiting for it to cure; after curing, removing the fixing block, taking out the whole sensor from the mold and cutting it into a suitable size; then, taking another longer copper foil, sticking a sealing film in the middle area of its back, and sticking two copper foils at both ends of this copper strip respectively to connect them; finally, sticking two slender Kapton or other insulating supports on the bottom surface of the Ecoflex layer, and the preparation of the sensor can be completed.

[0007] The beneficial effects of this application are as follows:

[0008] (1) The gesture recognition sensor proposed in this application is used for the array arrangement on the back of the hand, has little impact on hand movement after installation. And this device is simple to prepare, convenient to install, small in volume, can be self-driven without an additional power supply. Without increasing the structural complexity of the sensor itself, the gesture recognition accuracy can be greatly improved; by applying machine learning algorithms, the accuracy rate of this sensor in recognizing the gestures corresponding to the numbers 1 to 9 can reach more than 95%.

[0009] (2) The Ecoflex layer material used in this application is a flexible material with good stretchability, has little hindrance to hand movement, and can improve the comfort of the sensor. At the same time, through the two-piece electrode structure (2 metal electrode layers), its ductility can be further ensured.

[0010] (3) Common gesture recognition sensors mainly rely on devices that require external power supply, or self-powered devices with complex structures embedded in media such as gloves and finger cots, and their disadvantages are all that they are easy to cause hindrance to normal movement, but the contact-separation mechanism of this application does not require an additional power supply, and the array arrangement method can increase the amount of information obtained by measuring a single gesture action, and can greatly reduce the hindrance to hand movements on the premise of maintaining high gesture recognition accuracy. Description of the Drawings

[0011] Figure 1 is a schematic structural diagram of the sensor described in this application;

[0012] Figure 2 is a schematic diagram of the array structure of the gesture recognition sensor;

[0013] Figure 3 is a schematic diagram of the preparation process of the gesture recognition sensor;

[0014] Figure 4 Schematic diagram of pasting sensors at measurement points

[0015] Figure 5 Schematic diagram of the voltage response of the sensor at test position 1 to different finger movements

[0016] Figure 6 Schematic diagram of the voltage response of the sensor at test position 2 to different finger movements

[0017] Figure 7 Schematic diagram of the voltage change when the palm makes three times of fist clenching, full opening and then fist clenching at different sensor positions

[0018] Figure 8 Schematic diagram of the voltage waveforms of different fingers corresponding to different sensors in different gestures

[0019] Figure 9 Flowchart of the program processing method for gesture recognition implemented by machine learning

[0020] Figure 10 Schematic diagram of the clustering analysis results of the original data after machine learning

[0021] Figure 11 Confusion matrix after machine learning training of the gesture recognition sensor described in this application when recognizing gestures corresponding to numbers 1 to 9

[0022] In the figure: 1 - Ecoflex layer; 2 - Metal electrode layer; 3 - Connecting electrode layer; 4 - Sealing film layer; 5 - Insulating support layer; 6 - PTFE layer Specific embodiments

[0023] The technical solution of this application will be described in detail below with reference to the accompanying drawings

[0024] The embodiments of this application disclose a gesture recognition sensor based on a triboelectric nanogenerator. During the finger movement, the muscle groups on the back of the hand will be driven to move together, resulting in the deformation and surface undulation changes of the skin on the back of the hand. Based on the principles of electrostatic induction and friction, the contact and separation between the skin on the back of the hand and the sensor can generate electrical signals with a certain waveform. The sensor can obtain multiple groups of different electrical signals generated by the deformation of the skin at different positions on the back of the hand through the array arrangement on the back of the hand. The electrical signals can accurately identify specific gestures after being processed and analyzed by machine learning

[0025] As Figure 1As shown, each sensor includes an Ecoflex layer 1, a metal electrode layer 2, a connecting electrode layer 3, a sealing film layer 4, an insulating support layer 5, and a PTFE layer 6. The Ecoflex layer 1 has a "concave" structure with protrusions at both ends, and the bottom of the Ecoflex layer 1 is connected to the top of the insulating support layer 5. The metal electrode layer 2 includes two metal electrode layers distributed left and right and are both located on the top of the Ecoflex layer 1. One end of the two metal electrode layers close to each other is connected by the connecting electrode layer 3, and the other ends of the two metal electrode layers away from each other are arranged in the protrusions at both ends of the Ecoflex layer 1. The bottom of the connecting electrode layer 3 is provided with a sealing film layer 4, and both ends of the sealing film layer 4 are respectively connected to one end of the two metal electrode layers close to each other. The bottom of the insulating support layer 5 is provided with a PTFE layer 6. The bottom of the insulating support layer 5 is directly pasted on the skin, and the PTFE layer 6 is also pasted on the skin at the same time. The area of the PTFE layer 6 is smaller than that of the insulating support layer 5. The PTFE layer 6 is first pasted on the back of the hand skin, and then the insulating support layer 5 is covered on the PTFE layer 6 and pasted to the back of the hand skin.

[0026] The Ecoflex layer 1 serves as a carrier for the collected gesture signals, and its material can also be polydimethylsiloxane. This layer is made of a flexible material, which is convenient for improving the comfort of wearing the sensor during hand movement. At the same time, the Ecoflex layer 1 is also a carrier for the metal electrode layer 2. The protruding parts at both ends of the Ecoflex layer 1 respectively wrap the two metal electrode layers, which can achieve a good fixing effect. The two metal electrode layers are arranged left and right, and the two metal electrode layers are connected by the connecting electrode layer 3. The main purpose is to improve the ductility of the sensor so that it can still fit well with the skin and work after a certain degree of extension.

[0027] Since the metal electrode often has adhesiveness on one side, the sealing film layer 4 is added to prevent the connecting electrode layer 3 from adhering to the lower metal electrode layer 2 during the extension process, which affects the ductility. The purpose of the insulating support layer 5 is to lift the sensor, so that there is a certain distance between the lower surface of the Ecoflex layer 1 and the back of the hand skin. The back of the hand skin deforms during finger movement, and the skin continuously contacts and separates from the Ecoflex layer 1, generating electrical signals. The signals are transmitted to the processing device for processing through the wires connected to the metal electrode layer 2. Since the back of the hand skin is relatively flat and has little undulation, the additional PTFE layer 6 can make the voltage response waveform more significant.

[0028] With Figure 2For example, the sensor array includes 3 sensors, and their arrangement includes: 2 sensors are sequentially arranged on the back skin between the index finger and the middle finger, and 1 sensor is arranged on the back skin between the ring finger and the little finger. After making the gesture of "2" with the left hand, the sensor in the upper right corner generates a relatively strong specific electrical signal due to the large-amplitude movement of the skin and muscles driven by the index finger and the middle finger, while the thumb does not move. Similarly, the sensor below also generates a certain electrical signal due to the movement of the index finger and the middle finger, while the sensor in the upper left corner generates a smaller electrical signal fluctuation due to its relatively slight movement. Through the joint judgment of the electrical signals obtained by the three sensors and machine learning training on the data obtained from multiple previous tests, the subsequent program can complete the recognition of the gesture "2".

[0029] The preparation method of the sensor is as Figure 3 shown, specifically including: First, mix the A and B bottles of Ecoflex in a ratio of 1:1 and stir well. Pour the Ecoflex mixture into a mold and wait for 4 to 6 hours to cure it. Subsequently, paste two copper foils of the same size side by side on the surface of the cured Ecoflex, place a fixing block slightly smaller than the total covered area of the two copper foils in the center on the surface of the copper foil, and pour a small amount of Ecoflex mixture around it and wait for it to cure. After curing, remove the fixing block, take out the whole sensor from the mold and cut it into a suitable size. Then, take another longer copper foil, paste a sealing film in the middle area of its back, and paste two copper foils at both ends of this copper foil to connect them. Finally, paste two slender Kapton or other insulating supports on the bottom surface of the Ecoflex layer to complete the preparation of the sensor.

[0030] In a specific embodiment, taking a sensor with a size of 1mm×2mm as an example, first paste the sensor at the measurement point, as Figure 4 shown. Select the position 2 cm along the internal vein from the index finger joint on the back of the left hand as test position 1, and the position 2 cm along the internal vein from the ring finger joint on the back of the left hand as test position 2. When starting the measurement, the initial gesture is a fist, and the voltage changes are respectively tested when each finger changes from a fist to full extension and then back and repeated several times, and the action strength gradually increases. The measured curves are as Figure 5 and Figure 6 shown.

[0031] Comprehensive comparative analysis Figure 5 and Figure 6, when the sensor is at test position 1, the sensor has a relatively obvious voltage response to the movements of the thumb, index finger, and middle finger, and the voltage response curves of the three different fingers are significantly different. For example, when the index finger extends and then retracts, the voltage change curve roughly shows a "W"-shaped decline and then rise, with a decline of about 1 to 1.5 V; when the middle finger extends and then retracts, the voltage change curve roughly shows a trend of slightly declining first, then rising significantly, and then declining to the origin, with a decline of about 0.5 V, and then a subsequent rise of about 1.5 V; when the thumb extends and then retracts, the voltage change curve is roughly similar to one period of a trigonometric function, first slightly declining, then rising significantly, then declining significantly, and then rising to the origin, and the overall voltage change is between 1 and 1.5 V. When the sensor is at test position 2, the sensor also has significantly different voltage response curves to the movements of the little finger and ring finger. It can be seen that this sensor array can comprehensively obtain and distinguish the motion state information of each finger.

[0032] The morphology of the back of the hand has large individual differences, and it is difficult to find the best measurement points among different individuals. If the measurement target of a single sensor is only limited to certain fingers, the range of points that can be used for precise measurement will be greatly increased. At the same time, the smaller size of the sensor in this application ensures that even if several sensors are applied simultaneously, it will not hinder the normal movement of the hand and will not make the wearing uncomfortable. Therefore, several identical finger motion sensors can be simultaneously arranged in an array on the back of the hand. Each sensor mainly measures the motion state of a certain or several specific fingers respectively, and the motion information of each finger is completely covered by sufficient sensors. The array arrangement method can also enable the motion state information of the same finger to be recorded simultaneously by multiple sensors, and the data measured by each sensor form a complement, greatly improving the accuracy of gesture data, which is a preferred application solution for realizing gesture recognition.

[0033] Further measure the response to the same action when the sensor is attached to different positions. The voltage change curve when the hand clenches into a fist, fully opens, and then clenches again is as Figure 7 shown, which can verify that this sensor array form can make the data of different sensors complementary and further improve the accuracy of gesture recognition. Figure 8 That is, the schematic diagram of the voltage waveforms of different fingers corresponding to different sensors in different gestures.

[0034] Taking 15 groups of original data, 3 sensors in each group, and 501 voltage data for each sensor as an example, the specific program processing process is as Figure 9 shown.

[0035] In the process of applying the LDA algorithm for machine learning, it is necessary to appropriately select the program parameters.

[0036] For the solver, it is divided into three types: svd (singular value decomposition method), lsqr (least squares method), and eigen (eigenvalue decomposition method). Given that the number of features in this gesture recognition application scenario is small, it is suitable to select eigen as the solver.

[0037] For the regularization parameter shrinkage, it can enhance the generalization ability of LDA classification and can be ignored in this application.

[0038] For the class weights priors, since this application is for digit recognition and the weights of all classes should be exactly the same, the default value is selected for this item and it is ignored.

[0039] For the dimensionality reduction dimension n_components, the set dimension must be between 1 and "number of classes - 1". Since the number of classes in this application is 9 and the original data has 21 dimensions (3 sensors, each sensor extracts 7 eigenvalue features), the default value of "number of classes - 1", which is 8 dimensions, is sufficient to complete the work. Therefore, this item is also ignored.

[0040] The clustering analysis results of the original data after machine learning are as Figure 10 shown, Figure 10 In it, each digit is represented by the points within the circle. The results of this clustering analysis show that after machine learning, the differences between different gestures corresponding to each digit can be accurately extracted, and each digit can be clearly and unambiguously classified.

[0041] As can be seen from Figure 11 it, after the sensors are arranged in an array as shown in Figure 2 and through machine learning training, it has a very high accuracy rate for recognizing the gestures corresponding to digits 1 to 9, and can reach an identification accuracy rate of 96.67% in 500 tests, enabling high-accuracy gesture recognition.

[0042] In the embodiment of this application, the ideal thickness of the Ecoflex layer 1 is 1 mm, and its material can be other materials such as polydimethylsiloxane (PDMS), polytetrafluoroethylene (PTFE), etc., but the voltage response effect and ductility of Ecoflex are relatively good.

[0043] In the embodiment of this application, the electrode materials of the metal electrode layer 2 and the connecting electrode layer 3 can be aluminum (Al), gold (Au), copper (Cu), silver (Ag), etc. Materials such as aluminum foil and copper foil can be directly used, or they can be prepared by electron beam evaporation or magnetron sputtering. The thickness of the electrode can be 5 nm - 2 μm.

[0044] In the embodiment of the present application, the material of the insulating support layer 5 can be an insulating material with a certain thickness, such as an acrylic board, Kapton, etc., and it is tightly pasted to the Ecoflex layer 2 and the skin through glue. The ideal thickness of the insulating material is 0.5 mm - 2 mm. To achieve a good contact separation effect, the thickness should not be too large.

[0045] In the embodiment of the present application, the ideal thickness of the PTFE layer 6 is 100 μm.

[0046] In the embodiment of the present application, the array arrangement form can be the way in the example, or other arrangement forms or arrangement points can be adopted. The basic principle is that after the program is trained with a certain amount of data, it can achieve a good distinction of different gestures. At the same time, the number of sensors required to obtain gesture information is not limited to three, but should not be less than three. Four or more sensors can be used to obtain more accurate motion information, but it should be noted that when using more sensors, the size of the sensors should be reasonably controlled to avoid being too densely pasted and hindering the normal movement of the fingers.

[0047] The above is a demonstration embodiment of the present application, and the protection scope of the present application is defined by the claims and their equivalents.

Claims

1. A gesture recognition sensor based on a triboelectric nanogenerator, the gesture recognition sensor being a sensor array including at least three sensors, characterized in that, Each sensor includes an Ecoflex layer, a metal electrode layer, a connecting electrode layer, a sealing film layer, an insulating support layer, and a PTFE layer. The Ecoflex layer has a "concave" structure with protrusions at both ends, and the bottom of the Ecoflex layer is connected to the top of the insulating support layer. The metal electrode layer includes two metal electrode layers arranged left and right and both are located on the top of the Ecoflex layer. One end of the two metal electrode layers close to each other is connected by the connecting electrode layer, and the ends of the two metal electrode layers far from each other are arranged in the protrusions at both ends of the Ecoflex layer. The bottom of the connecting electrode layer is provided with a sealing film layer, and both ends of the sealing film layer are respectively connected to one end of the two metal electrode layers close to each other. The bottom of the insulating support layer is provided with a PTFE layer.

2. The gesture recognition sensor according to claim 1, wherein The thickness of the Ecoflex layer is 1 mm.

3. The gesture recognition sensor according to claim 1, wherein The electrode materials of the metal electrode layer and the connecting electrode layer are any one of aluminum, gold, copper, silver, and copper, and the thickness of the electrode is 5 nm - 2 μm.

4. The gesture recognition sensor according to claim 1, wherein The material of the insulating support layer is any one of acrylic plate and Kapton, and its thickness is 0.5 mm - 2 mm.

5. The gesture recognition sensor according to claim 1, wherein The thickness of the PTFE layer is 100 μm.

6. The gesture recognition sensor according to claim 1, wherein The sensor array includes three sensors, and its arrangement includes: two sensors are sequentially arranged on the back skin between the index finger and the middle finger, and one sensor is arranged on the back skin between the ring finger and the little finger.

7. A method for preparing a gesture recognition sensor according to any one of claims 1-6, characterized in that, Including: Mix the A and B bottles of Ecoflex in a ratio of 1:1, and stir well. Then pour the Ecoflex mixture into a mold and wait for 4 to 6 hours to cure it. Subsequently, paste two copper foils of the same size side by side on the surface of the cured Ecoflex, place a fixing block slightly smaller than the total covered area of the two copper foils in the center on the surface of the copper foils, and pour a small amount of Ecoflex mixture around it and wait for it to cure. After curing, remove the fixing block, take out the whole sensor from the mold and cut it into a suitable size. Then, take another longer copper foil, paste a sealing film in the middle area of its back, and paste both ends of this copper foil to the two copper foils respectively to connect them. Finally, paste two slender Kapton or other insulating supports on the bottom surface of the Ecoflex layer to complete the preparation of the sensor.

Images