Gesture control reconfigurable metasurface system based on wearable device

Abstract

This invention relates to the field of electromagnetic wave manipulation and proposes a reconfigurable metasurface system based on gesture control using wearable devices. The system includes a wearable device, a gesture recognition model, a field-programmable gate array (FPGA), and a reconfigurable metasurface. Specifically, the wearable device acquires and processes surface electromyography (EMG) signals in real time and transmits them to the gesture recognition model. The gesture recognition model performs real-time and efficient recognition of the EMG signals acquired by the wearable device to obtain gesture signals. The FPGA loads the encoding matrix corresponding to the gesture signals recognized by the gesture recognition model onto the reconfigurable metasurface, enabling beam deflection and polarization conversion functions. This invention achieves the combination of real-time manipulation of human physiological signals and electromagnetic waves, possessing significant potential applications in intelligent device control, virtual reality systems, and wireless communication technologies, and providing new technical support for the advancement and innovation of human-computer interaction technology.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic wave manipulation, specifically to a gesture-controlled reconfigurable metasurface system based on wearable devices. Background Technology

[0002] With the development of information technology and wireless communication technology, human-computer interaction methods are constantly evolving. Traditional input devices such as keyboards and mice can no longer meet the ever-increasing demands for human-computer interaction. In recent years, gesture recognition, as a natural and intuitive human-computer interaction method, has received widespread attention. Through gesture recognition technology, users can control electronic devices with simple gestures, greatly improving the convenience of interaction and user experience. Currently implemented gesture recognition methods mainly include computer vision methods, radio frequency methods, and wearable devices. Computer vision methods are sensitive to lighting conditions, and complex backgrounds may affect recognition accuracy. Radio frequency methods also face problems such as signal interference and multipath effects. Wearable device methods can effectively solve these problems, becoming an effective method for gesture recognition.

[0003] Meanwhile, significant progress has been made in the field of electromagnetic wave manipulation, particularly in the application of reconfigurable metasurface technology. Reconfigurable metasurfaces are novel artificial structural materials capable of dynamically adjusting their reflection, refraction, or other electromagnetic properties, thereby enabling flexible manipulation of incident electromagnetic waves. This technology has already been applied in various fields such as wireless communication and radar stealth. However, research on real-time electromagnetic wave manipulation in conjunction with human physiological signals remains limited. Summary of the Invention

[0004] To address the shortcomings of the aforementioned technologies, the present invention aims to provide a gesture-controlled reconfigurable metasurface system based on wearable devices, which combines real-time regulation of human physiological signals and electromagnetic waves. This system has significant potential applications in intelligent device control, virtual reality systems, and wireless communication technologies, and provides new technical support for the advancement and innovation of human-computer interaction technology.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A gesture-controlled reconfigurable metasurface system based on wearable devices includes a wearable device, a gesture recognition model, a field-programmable gate array (FPGA), and a reconfigurable metasurface, wherein:

[0007] The wearable device acquires and processes surface electromyography (EMG) signals in real time and transmits them to the gesture recognition model. The gesture recognition model performs real-time and efficient recognition of the EMG signals acquired by the wearable device to obtain gesture signals. The field-programmable gate array loads the encoding matrix corresponding to the gesture signals recognized by the gesture recognition model onto the reconfigurable metasurface, which realizes beam deflection and polarization conversion functions.

[0008] Specifically, the wearable device is a wearable armband with multiple electromyography (EMG) sensor channels. It amplifies, notch filters, and bandpass filters the collected surface EMG signals and transmits the data to the gesture recognition model via Bluetooth.

[0009] Specifically, the gesture recognition model is a CNN-Transformer hybrid model, which includes an input layer, a CNN module, a Transformer module, and an output layer, wherein:

[0010] The CNN module includes convolutional layers, batch normalization layers, ReLU activation layers, and pooling layers; the Transformer module includes embedding layers and Transformer Encoder layers; the output layer includes fully connected layers and Softmax layers.

[0011] During model training, the collected surface electromyography (EMG) signal data is normalized and then imported into the input layer of the CNN-Transformer hybrid model. The convolutional layers in the CNN module extract local features from the input data by applying multiple convolutional kernels. Batch normalization layers and ReLU activation layers accelerate the training process and introduce non-linearity. Pooling layers perform downsampling to reduce the number of features and retain the most relevant features. The embedding layer in the Transformer module embeds the features extracted by the CNN into a higher-dimensional space, enhancing the model's expressive power. The Transformer Encoder layer performs self-attention modeling to capture long-term dependencies. After processing by the CNN and Transformer modules, the data is finally mapped to various predefined gestures through the output layer.

[0012] Specifically, the input of the field-programmable gate array (FPGA) is connected to the gesture recognition model, and the output pin of the FPGA 3 is connected to the reconfigurable metasurface.

[0013] Specifically, the reconfigurable metasurface is composed of a number of metasurface units arranged periodically; the structure of the metasurface unit, from top to bottom, includes a top square patch, a first dielectric substrate, a ground layer, a second dielectric substrate, and a bottom reflective phase shifter arranged in layers; the top square patch, the ground layer, and the bottom reflective phase shifter are three metal layers.

[0014] The metasurface unit further includes a first metal through-pillar, a second metal through-pillar, a third metal through-pillar, a fourth metal through-pillar, and two bottom-layer DC voltage bias line circuits. The two bottom-layer DC voltage bias line circuits are the first bottom-layer DC voltage bias line circuit and the second bottom-layer DC voltage bias line circuit, respectively. The first metal through-pillar passes through the first dielectric substrate, the ground layer, and the second dielectric substrate, and is connected to the top-layer square patch and the bottom-layer reflective phase shifter, respectively. The second metal through-pillar passes through the first dielectric substrate and is connected to the top-layer square patch and the ground layer, respectively. The third metal through-pillar passes through the second dielectric substrate and is connected to the ground layer and the first bottom-layer DC voltage bias line circuit, respectively. The fourth metal through-pillar passes through the second dielectric substrate and is connected to the ground layer and the second bottom-layer DC voltage bias line circuit, respectively.

[0015] Furthermore, the underlying reflective phase shifter includes a first microstrip transmission line, a first branch microstrip transmission line, a second branch microstrip transmission line, a second microstrip transmission line, a fan-shaped microstrip stub, a first square patch, a second square patch, and three PIN diodes; the three PIN diodes are a first PIN diode, a second PIN diode, and a third PIN diode, respectively.

[0016] One end of the first branch microstrip transmission line and the second branch microstrip transmission line are perpendicularly connected to the first microstrip transmission line; the other ends of the first branch microstrip transmission line and the second branch microstrip transmission line are connected to the second microstrip transmission line through the first PIN diode and the second PIN diode, respectively; the second microstrip transmission line 4-1-5-4 is parallel to the first microstrip transmission line 4-1-5-1, and the two ends of the second microstrip transmission line are connected to the first square patch and the second square patch, respectively; the fan-shaped microstrip stub is connected to one end of the first microstrip transmission line through the third PIN diode.

[0017] Furthermore, the first bottom-layer DC voltage bias line circuit is connected to the fan-shaped microstrip stub of the bottom-layer reflective phase shifter, and the second bottom-layer DC voltage bias line circuit is connected to the first square patch of the bottom-layer reflective phase shifter; the two bottom-layer DC voltage bias line circuits (the first bottom-layer DC voltage bias line circuit and the second bottom-layer DC voltage bias line circuit) control the off or on state of the three PIN diodes (the first PIN diode, the second PIN diode, and the third PIN diode) to achieve different reflection phase responses of the bottom-layer reflective phase shifter.

[0018] Furthermore, the underlying reflective phase shifter has four reflective phase responses: 0°, 90°, 180°, and 270°. Correspondingly, these four reflective phase responses are encoded as four digital coded states: "A", "B", "C", and "D", where:

[0019] The digital coding status "A" indicates that the first PIN diode, the second PIN diode, and the third PIN diode are in the off state; the digital coding status "B" indicates that the first PIN diode and the second PIN diode are in the off state, and the third PIN diode is in the on state; the digital coding status "C" indicates that the first PIN diode and the second PIN diode are in the on state, and the third PIN diode is in the off state; the digital coding status "D" indicates that the first PIN diode, the second PIN diode, and the third PIN diode are in the on state.

[0020] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0021] This invention proposes a reconfigurable metasurface system for gesture control based on wearable devices, used for real-time beam deflection and polarization conversion. In the area of ​​gesture recognition using surface electromyography (EMG) signals, this system innovatively combines the ability of CNNs to extract local spatial features with the advantage of Transformer models in modeling global temporal dependencies through self-attention mechanisms, significantly improving model performance and achieving a high classification accuracy of 98.54%. In the metasurface design, a 2-bit reconfigurable metasurface using only three PIN diodes is proposed. By separating the radiating and phase-shifting components, design complexity and phase quantization error are balanced, and diode integration is facilitated.

[0022] This invention is the first to combine surface electromyography (SEMG) signal gesture recognition technology with reconfigurable metasurface technology, proposing a novel gesture-controlled metasurface system that is expected to bring new breakthroughs to the field of human-computer interaction. Furthermore, this system represents a pioneering integration of cutting-edge technologies from multiple disciplines, including biomedical engineering (surface EMG signal acquisition and processing), artificial intelligence (gesture recognition model), and electromagnetics (metasurface design and control), achieving interdisciplinary collaboration and innovation. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the gesture-controlled reconfigurable metasurface system described in this invention;

[0024] Figure 2 This is a schematic diagram of signal acquisition for the wearable device described in this invention. (a) shows the surface electromyography (EMG) signal acquisition armband and the gesture selected in the embodiment; (b) shows the surface EMG signal corresponding to the first gesture.

[0025] Figure 3 This is a schematic diagram of the CNN-Transformer hybrid model described in this invention. (a) is the network structure diagram of the CNN-Transformer hybrid model, and (b) is the recognition accuracy of the trained CNN-Transformer hybrid model.

[0026] Figure 4 This is a schematic diagram of the reconfigurable metasurface unit described in this invention. (a) is a three-dimensional structural diagram, (b) is a top view, and (c) is a bottom view.

[0027] Figure 5 The experimental results of beam deflection of the reconfigurable metasurface described in this invention are shown. Among them, (a) is the unit coding matrix corresponding to a beam deflection of 10° on the reconfigurable metasurface, (b) is the scattering result diagram of a beam deflection of 10° on the metasurface after recognizing the relevant gesture and loading the coding matrix, (c) is the unit coding matrix corresponding to a beam deflection of 30° on the reconfigurable metasurface, and (d) is the scattering result diagram of a beam deflection of 30° on the metasurface after recognizing the relevant gesture and loading the coding matrix.

[0028] Figure 6 These are the experimental results of polarization conversion of the reconfigurable metasurface described in this invention. (a) is the unit coding matrix for converting the reconfigurable metasurface from linear polarization to left-handed circular polarization; (b) is the axial ratio result of the metasurface converting from linear polarization to left-handed circular polarization after recognizing the relevant gesture and loading the coding matrix; (c) is the unit coding matrix for converting the reconfigurable metasurface from linear polarization to right-handed circular polarization; and (d) is the axial ratio result of the metasurface converting from linear polarization to right-handed circular polarization after recognizing the relevant gesture and loading the coding matrix.

[0029] Figure label:

[0030] 1. Wearable armband; 2. CNN-Transformer hybrid model; 3. Field-programmable gate array (FPGA); 4. Reconfigurable metasurface; 4-1. Metasurface unit; 5. Wearing method illustration; 6. First gesture; 7. Second gesture; 8. Third gesture; 9. Fourth gesture; 10. Fifth gesture; 11. Sixth gesture; 12. Seventh gesture; 13. Eighth gesture; 4-1-1. Top layer square patch; 4-1-2. First dielectric substrate; 4-1-3. Ground layer; 4-1-4. Second dielectric substrate; 4-1-4. Bottom layer reflective phase shifter; 4-1-5. First metal pillar; 4-1-6. Second metal pillar; 4-1-7. Third metal pillar; 4-1- 8, Fourth metal through-pillar 4-1-9, First bottom layer DC voltage bias line circuit 4-1-10, Second bottom layer DC voltage bias line circuit 4-1-11, First microstrip transmission line 4-1-5-1, First branch microstrip transmission line 4-1-5-2, Second branch microstrip transmission line 4-1-5-3, Second microstrip transmission line 4-1-5-4, Fan-shaped microstrip stub 4-1-5-5, First square patch 4-1-5-6, Second square patch 4-1-5-7, First PIN diode 4-1-5-8, Second PIN diode 4-1-5-9, Third PIN diode 4-1-5-10. Detailed Implementation

[0031] The technical solutions provided in this application will be further described below with reference to specific embodiments and accompanying drawings. The advantages and features of this application will become clearer from the following description.

[0032] like Figure 1 As shown, a gesture-controlled reconfigurable metasurface system based on wearable devices for real-time beam deflection and polarization conversion includes:

[0033] Wearable device, specifically wearable armband 1, collects multi-channel surface electromyography signals;

[0034] The gesture recognition model, specifically the CNN-Transformer hybrid model 2, runs on a computer and recognizes various gesture signals.

[0035] Field Programmable Gate Array 3;

[0036] Reconfigurable metasurface 4.

[0037] The wearable armband 1 is connected to the CNN-Transformer hybrid model 2 in the computer via Bluetooth. The CNN-Transformer hybrid model 2 is connected to the field-programmable gate array 3 via USB. The output pins of the field-programmable gate array 3 are connected to the reconfigurable metasurface 4.

[0038] In this embodiment, the wearable armband 1 has 8 electromyography (EMG) sensor channels, which can amplify the collected surface EMG signals by 2000 times, perform 50Hz notch filtering, and 15-550Hz bandpass filtering, and transmit the data to the CNN-Transformer hybrid model 2 via Bluetooth.

[0039] like Figure 2 As shown in (a), the wearable armband 1 for collecting surface electromyography signals is worn in a schematic diagram (5), specifically on the forearm. The eight gestures used for recognition are sequentially designated as gesture 6, gesture 7, gesture 8, gesture 9, gesture 10, gesture 11, gesture 12, and gesture 13. Each gesture has a corresponding electromagnetic wave modulation function. In this embodiment, the eight gestures correspond to eight electromagnetic wave modulation functions of the reconfigurable metasurface. For example, gesture 6 corresponds to the conversion of linear polarization to left-handed circular polarization, gesture 13 corresponds to the conversion of linear polarization to right-handed circular polarization, gesture 7 corresponds to a beam deflection of 10°, and gesture 9 corresponds to a beam deflection of 30°. Figure 2 (b) shows an electromyographic signal acquired during the execution of the first gesture 6 for example.

[0040] like Figure 3As shown, the CNN-Transformer hybrid model includes an input layer, a CNN module, a Transformer module, and an output layer. The CNN module includes convolutional layers, batch normalization layers, ReLU activation layers, and pooling layers; the Transformer module includes an embedding layer and a Transformer Encoder layer; and the output layer includes a fully connected layer and a Softmax layer.

[0041] During model training, the collected surface electromyography (EMG) signal data is normalized and then imported into the input layer of the CNN-Transformer hybrid model. The convolutional layers in the CNN module extract local features from the input data by applying multiple convolutional kernels. Batch normalization layers and ReLU activation layers accelerate the training process and introduce non-linearity. Pooling layers perform downsampling to reduce the number of features and retain the most relevant ones. The embedding layers in the Transformer module embed the features extracted by the CNN into a higher-dimensional space, enhancing the model's expressive power. The Transformer Encoder layer performs self-attention modeling to capture long-term dependencies. After processing by the CNN and Transformer modules, the final output layer maps to eight predefined gestures.

[0042] The CNN-Transformer hybrid model combines the ability of convolutional neural networks to extract local features with the advantage of the Transformer model in modeling global temporal dependencies through self-attention, achieving efficient processing of electromyographic signals. Through multi-level feature extraction and spatiotemporal feature co-learning, the model can simultaneously capture both local fluctuations and global temporal dependencies of the signal, significantly improving the accuracy of gesture recognition.

[0043] In the training process of the example model, three volunteers each repeated eight gestures ten times. The collected surface electromyography (EMG) signal data were normalized and then imported into the input layer of a CNN-Transformer hybrid model. After processing by the CNN and Transformer modules, the data was finally mapped to the eight predefined gestures through the output layer. The model was built and trained using Matlab, and after 100 iterations, it achieved an accuracy of 98.54%.

[0044] The reconfigurable metasurface 4 is composed of a periodically arranged plurality of metasurface units 4-1. As an example, such as... Figure 1 and Figure 4 As shown, the reconfigurable metasurface 4 is composed of 64 metasurface units 4-1 arranged periodically.

[0045] The structure of the metasurface unit 4-1, from top to bottom, includes a top square patch 4-1-1, a first dielectric substrate 4-1-2, a ground layer 4-1-3, a second dielectric substrate 4-1-4, and a bottom reflective phase shifter 4-1-5 arranged in a stacked manner. The top square patch 4-1-1, the ground layer 4-1-3, and the bottom reflective phase shifter 4-1-5 are three metal layers. The width of the top square patch 4-1-1 is 23mm. The size of the ground layer 4-1-3 is 55mm×55mm. The first dielectric substrate 4-1-2 and the second dielectric substrate 4-1-4 are both made of F4b (dielectric constant 2.47, loss tangent 0.002). The thickness of the first dielectric substrate 4-1-2 is h1=3mm, the thickness of the second dielectric substrate 4-1-4 is h2=1mm, and the size of both is 55mm×55mm.

[0046] In addition, the metasurface unit 4-1 also includes a first metal through post 4-1-6, a second metal through post 4-1-7, a third metal through post 4-1-8, a fourth metal through post 4-1-9, a first bottom layer DC voltage bias line circuit 4-1-10, and a second bottom layer DC voltage bias line circuit 4-1-11. The first metal post 4-1-6 passes through the first dielectric substrate 4-1-2, the ground layer 4-1-3, and the second dielectric substrate 4-1-4, and is connected to the top square patch 4-1-1 and the bottom reflective phase shifter 4-1-5, respectively. The distance d1 between the first metal post 4-1-6 and the center of the top square patch 4-1-1 is 4mm. The second metal post 4-1-7 passes through the first dielectric substrate 4-1-2 and is connected to the top square patch 4-1-1 and the ground layer 4-1-3, respectively. The distance d2 between the second metal post 4-1-7 and the center of the top square patch 4-1-1 is 4mm. The third metal post 4-1-8 passes through the second dielectric substrate 4-1-4 and is connected to the ground layer 4-1-3 and the first bottom DC voltage bias line circuit 4-1-10, respectively. The fourth metal post 4-1-9 passes through the second dielectric substrate 4-1-4 and is connected to the ground layer 4-1-3 and the second bottom DC voltage bias line circuit 4-1-11, respectively.

[0047] like Figure 4As shown in (c), the bottom reflective phase shifter 4-1-5 includes a first microstrip transmission line 4-1-5-1, a first branch microstrip transmission line 4-1-5-2, a second branch microstrip transmission line 4-1-5-3, a second microstrip transmission line 4-1-5-4, a fan-shaped microstrip stub line 4-1-5-5, a first square patch 4-1-5-6, a second square patch 4-1-5-7, and three PIN diodes; the three PIN diodes include a first PIN diode 4-1-5-8 and a second PIN diode 4-1-5-9. The third PIN diode is 4-1-5-10; the first microstrip transmission line 4-1-5-1 has a length of 14mm and a width of 2.88mm; the first branch microstrip transmission line 4-1-5-2 and the second branch microstrip transmission line 4-1-5-3 have a length of 6.15mm and a width of 0.5mm; the second microstrip transmission line 4-1-5-4 has a length of 18mm and a width of 2mm; the first square patch 4-1-5-6 and the second square patch 4-1-5-7 have dimensions of 6.5mm × 6.5mm.

[0048] One end of the first branch microstrip transmission line 4-1-5-2 and the second branch microstrip transmission line 4-1-5-3 are perpendicularly connected to the first microstrip transmission line 4-1-5-1; the other ends of the first branch microstrip transmission line 4-1-5-2 and the second branch microstrip transmission line 4-1-5-3 are connected to the second microstrip transmission line 4-1-5-4 through the first PIN diode 4-1-5-8 and the second PIN diode 4-1-5-9, respectively; the second microstrip transmission line 4-1-5-4 is parallel to the first microstrip transmission line 4-1-5-1, and the two ends of the second microstrip transmission line 4-1-5-4 are connected to the first square patch 4-1-5-6 and the second square patch 4-1-5-7, respectively; the fan-shaped microstrip stub 4-1-5-5 is connected to one end of the first microstrip transmission line 4-1-5-1 through the third PIN diode 4-1-5-10.

[0049] The first bottom-layer DC voltage bias line circuit 4-1-10 is connected to the fan-shaped microstrip stub 4-1-5-5 of the bottom-layer reflective phase shifter 4-1-5, and the second bottom-layer DC voltage bias line circuit 4-1-11 is connected to the first square patch 4-1-5-6 of the bottom-layer reflective phase shifter 4-1-5. The two bottom-layer DC voltage bias line circuits (the first bottom-layer DC voltage bias line circuit 4-1-10 and the second bottom-layer DC voltage bias line circuit 4-1-11) control the on / off state of three PIN diodes (the first PIN diode 4-1-5-8, the second PIN diode 4-1-5-9, and the third PIN diode 4-1-5-10) to achieve different reflection phase responses of the bottom-layer reflective phase shifter.

[0050] The underlying reflective phase shifter has four reflective phase responses: 0°, 90°, 180°, and 270°. Correspondingly, the four reflective phase responses are encoded as four digital coded states "A", "B", "C", and "D", where: digital coded state "A" indicates that the first PIN diode, the second PIN diode, and the third PIN diode are in the off state; digital coded state "B" indicates that the first PIN diode and the second PIN diode are in the off state, and the third PIN diode is in the on state; digital coded state "C" indicates that the first PIN diode and the second PIN diode are in the on state, and the third PIN diode is in the off state; and digital coded state "D" indicates that the first PIN diode, the second PIN diode, and the third PIN diode are in the on state.

[0051] The control method of the above-mentioned wearable device-based gesture control reconfigurable metasurface system is as follows: When the user wears the wearable armband 1 and makes a gesture, the wearable armband 1 collects surface electromyography (EMG) signals from 8 channels and then sends the signals to the computer via Bluetooth. The computer uses a pre-trained CNN-Transformer hybrid model 2 to identify the received EMG signals and further controls the output of the field-programmable gate array 3 according to the encoding matrix corresponding to the identified gesture. This controls the opening or closing state of three PIN diodes by driving two underlying DC voltage bias line circuits of the reconfigurable metasurface 4, thereby realizing different electromagnetic wave manipulation functions. When the reconfigurable metasurface performs beam deflection, it reflects the incident electromagnetic waves at different deflection angles; and when the polarization conversion function of the reconfigurable metasurface is activated, it converts linearly polarized incident waves into left-handed or right-handed circularly polarized waves.

[0052] like Figure 5 As shown, when the reconfigurable metasurface 4 implements the beam deflection function, the required encoding matrix, taking deflection of 10° and 30° as examples, is as follows: Figure 5 As shown in (a) and (c) of the present invention, when the wearable armband 1 in the wearable device-based gesture control reconfigurable metasurface system acquires the executed second gesture 7 and fourth gesture 9, it sends the surface electromyography signal data to the trained CNN-Transformer hybrid model 2 via Bluetooth for recognition. Subsequently, the metasurface unit encoding matrix corresponding to the recognized gesture is loaded onto the DC voltage bias line circuit of the reconfigurable metasurface 4 through the field programmable gate array 3, so that the reconfigurable metasurface 4 realizes the metasurface beam deflection function corresponding to the encoding matrix. Figure 5 (b) shows the scattering results after the metasurface beam is deflected by 10° after the second gesture 7 is identified and the coding matrix is ​​loaded. Figure 5 (d) shows the scattering results after the metasurface beam is deflected by 30° after the fourth gesture 9 is identified and the coding matrix is ​​loaded. Figure 5Figures (b) and (d) show that the gesture-controlled reconfigurable metasurface system based on wearable devices of the present invention can effectively realize the beam deflection function of electromagnetic waves controlled by gestures.

[0053] like Figure 6 As shown, when the reconfigurable metasurface 4 realizes the polarization conversion function, the required encoding matrices are as follows: Figure 6 As shown in (a) and (c) of the present invention, when the wearable device-based gesture control reconfigurable metasurface system executes the first gesture 6 corresponding to the linear polarization conversion to left-hand circular polarization, the axial ratio measured in the frequency range of 3.82-3.85 GHz is less than 3 dB, as... Figure 6 As shown in (b); when the linear polarization corresponding to gesture 13 is converted to right-hand circular polarization, the axial ratio measured in the frequency range of 3.68-3.82 GHz is less than 3 dB, as shown in (b). Figure 6 As shown in (d); Figure 6 The results in (b) and (d) demonstrate that the gesture-controlled reconfigurable metasurface system based on wearable devices of the present invention can effectively realize the polarization conversion function of gesture-controlled electromagnetic waves.

[0054] The above description is merely a description of preferred embodiments of this application and is not intended to limit the scope of this application in any way. Any changes or modifications made by those skilled in the art based on the above-disclosed technical content should be considered as equivalent and valid embodiments and fall within the scope of protection of the technical solution of this application.

Claims

1. A gesture-controlled reconfigurable metasurface system based on wearable devices, characterized in that, Including wearable devices, gesture recognition models, field-programmable gate arrays (3), and reconfigurable metasurfaces (4), among which: The wearable device acquires and processes surface electromyography signals in real time and transmits them to the gesture recognition model; the gesture recognition model performs real-time and efficient recognition of the electromyography signals acquired by the wearable device to obtain gesture signals; the field programmable gate array (3) loads the encoding matrix corresponding to the gesture signals recognized by the gesture recognition model onto the reconfigurable metasurface (4), and the reconfigurable metasurface (4) realizes beam deflection and polarization conversion functions; The reconfigurable metasurface (4) is composed of a number of metasurface units (4-1) arranged periodically; The structure of the metasurface unit (4-1) includes, from top to bottom, a top square patch (4-1-1), a first dielectric substrate (4-1-2), a ground layer (4-1-3), a second dielectric substrate (4-1-4), and a bottom reflective phase shifter (4-1-5), which are stacked in sequence. The top square patch (4-1-1), the ground layer (4-1-3), and the bottom reflective phase shifter (4-1-5) are three metal layers. The metasurface unit (4-1) further includes a first metal through-pillar (4-1-6), a second metal through-pillar (4-1-7), a third metal through-pillar (4-1-8), a fourth metal through-pillar (4-1-9), and two bottom-layer DC voltage bias line circuits, namely a first bottom-layer DC voltage bias line circuit (4-1-10) and a second bottom-layer DC voltage bias line circuit (4-1-11), wherein: The first metal post (4-1-6) passes through the first dielectric substrate (4-1-2), the ground layer (4-1-3), and the second dielectric substrate (4-1-4), and is connected to the top square patch (4-1-1) and the bottom reflective phase shifter (4-1-5) respectively; The second metal post (4-1-7) passes through the first dielectric substrate (4-1-2) and is connected to the top square patch (4-1-1) and the ground layer (4-1-3) respectively; The third metal post (4-1-8) passes through the second dielectric substrate (4-1-4) and is connected to the ground layer (4-1-3) and the first bottom layer DC voltage bias line circuit (4-1-10) respectively; The fourth metal post (4-1-9) passes through the second dielectric substrate (4-1-4) and is connected to the ground layer (4-1-3) and the second bottom layer DC voltage bias line circuit (4-1-11), respectively. The bottom reflective phase shifter (4-1-5) includes a first microstrip transmission line (4-1-5-1), a first branch microstrip transmission line (4-1-5-2), a second branch microstrip transmission line (4-1-5-3), a second microstrip transmission line (4-1-5-4), a fan-shaped microstrip stub (4-1-5-5), a first square patch (4-1-5-6), a second square patch (4-1-5-7), and three PIN diodes, namely a first PIN diode (4-1-5-8), a second PIN diode (4-1-5-9), and a third PIN diode (4-1-5-10). One end of the first branch microstrip transmission line (4-1-5-2) and the second branch microstrip transmission line (4-1-5-3) is perpendicularly connected to the first microstrip transmission line (4-1-5-1); the other ends of the first branch microstrip transmission line (4-1-5-2) and the second branch microstrip transmission line (4-1-5-3) are connected to the second microstrip transmission line (4-1-5-4) through the first PIN diode (4-1-5-8) and the second PIN diode (4-1-5-9), respectively; the second microstrip transmission line (4-1-5-4) is parallel to the first microstrip transmission line (4-1-5-1), and the two ends of the second microstrip transmission line (4-1-5-4) are connected to the first square patch (4-1-5-6) and the second square patch (4-1-5-7), respectively; the fan-shaped microstrip stub (4-1-5-5) is connected to one end of the first microstrip transmission line (4-1-5-1) through the third PIN diode (4-1-5-10); The first bottom layer DC voltage bias line circuit (4-1-10) is connected to the fan-shaped microstrip stub (4-1-5-5) of the bottom layer reflective phase shifter (4-1-5), and the second bottom layer DC voltage bias line circuit (4-1-11) is connected to the first square patch (4-1-5-6) of the bottom layer reflective phase shifter (4-1-5). The first bottom layer DC voltage bias line circuit (4-1-10) and the second bottom layer DC voltage bias line circuit (4-1-11) control the off or on state of the first PIN diode (4-1-5-8), the second PIN diode (4-1-5-9), and the third PIN diode (4-1-5-10) to achieve different reflection phase responses of the bottom layer reflective phase shifter.

2. The gesture-controlled reconfigurable metasurface system based on a wearable device as described in claim 1, characterized in that, The wearable device is a wearable armband (1) with multiple electromyography sensor channels. It amplifies, notch filters, and bandpass filters the collected surface electromyography signals and transmits the data to the gesture recognition model via Bluetooth.

3. The gesture-controlled reconfigurable metasurface system based on a wearable device as described in claim 1, characterized in that, The gesture recognition model is a CNN-Transformer hybrid model (2), which includes an input layer, a CNN module, a Transformer module, and an output layer, wherein: The CNN module includes convolutional layers, batch normalization layers, ReLU activation layers, and pooling layers; The Transformer module includes an embedding layer and a Transformer Encoder layer; The output layer includes a fully connected layer and a Softmax layer; During model training, the collected surface electromyography (EMG) signal data is normalized and then imported into the input layer of the CNN-Transformer hybrid model. The convolutional layers in the CNN module extract local features from the input data by applying multiple convolutional kernels. Batch normalization layers and ReLU activation layers accelerate the training process and introduce non-linearity. Pooling layers perform downsampling to reduce the number of features and retain the most relevant ones. The embedding layer in the Transformer module embeds the features extracted by the CNN into a higher-dimensional space, enhancing the model's expressive power. The Transformer Encoder layer performs self-attention modeling to capture long-term dependencies. After processing by the CNN and Transformer modules, the final output layer maps to various predefined gestures.

4. The gesture-controlled reconfigurable metasurface system based on a wearable device as described in claim 1, characterized in that, The input of the field-programmable gate array (3) is connected to the gesture recognition model, and the output pin of the field-programmable gate array 3 is connected to the reconfigurable metasurface (4).

5. The gesture-controlled reconfigurable metasurface system based on a wearable device as described in claim 1, characterized in that, The underlying reflective phase shifter has four reflective phase responses: 0°, 90°, 180°, and 270°. Correspondingly, these four reflective phase responses are encoded as four digital code states: "A", "B", "C", and "D". The digital code "A" indicates that the first PIN diode, the second PIN diode, and the third PIN diode are in the off state; The digital code "B" indicates that the first PIN diode and the second PIN diode are in the off state, and the third PIN diode is in the on state; The digital code "C" indicates that the first PIN diode and the second PIN diode are in the open state, and the third PIN diode is in the closed state. The digital code "D" indicates that the first PIN diode, the second PIN diode, and the third PIN diode are in the open state.

Images