Optical fiber intelligent carpet gait recognition system based on strain-contour bimodal network

Abstract

The invention discloses an optical fiber intelligent carpet gait recognition system based on a strain-contour bimodal network, and belongs to the technical field of intelligent sensors, and the system comprises an OFDR sensing system which can achieve continuous high-sensitivity strain collection on a long-distance optical fiber; according to the optical fiber sensing carpet, sensing optical fibers are arranged on the carpet to form a dense distributed strain sensing array, and plantar two-dimensional strain distribution is collected in real time; according to the local strain-global contour bimodal attention network, a local strain map attention module is used for modeling a dynamic coupling relation between optical fiber segments and extracting key stress characteristics, and a gradient-guided global contour non-local attention module is used for obtaining a sole structure contour and an overall stress form; and finally, integrating the two kinds of information to recognize gait categories. The optical fiber intelligent carpet gait recognition system based on the strain-contour bimodal network solves the problems that an existing gait recognition method is insufficient in spatial resolution, low in recognition precision, limited to environmental conditions and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent sensor, in particular to a fiber-optic intelligent carpet gait recognition system based on strain-contour dual-modal network. BACKGROUND

[0002] As a technology that realizes identity identification, behavior monitoring and health assessment by analyzing the characteristics of human walking or posture changes, gait recognition is widely used in medical rehabilitation, security monitoring and other fields. Compared with other biometric recognition methods, it has the advantages of non-contact, strong continuity, easy collection, etc.

[0003] Existing gait recognition technologies mainly fall into three categories: vision-based methods rely on cameras to obtain human-related information, but are susceptible to light, occlusion, shooting angle and privacy restrictions, and lack robustness; wearable device-based methods can provide high-precision motion data, but require users to wear additional hardware, affecting comfort and ease of use; and ground pressure or strain sensing-based methods have low spatial resolution, insufficient signal-to-noise ratio, and cannot be continuously sampled, making it difficult to capture fine-grained changes in foot strain, limiting the accuracy and generalization ability of gait recognition.

[0004] With the development of distributed fiber-optic sensing (DFOS) technology, optical frequency domain reflectometry (OFDR) technology, with its millimeter-level spatial resolution and micro-strain-level sensitivity, provides a new direction for building high-precision intelligent carpets. This technology can achieve continuous, high-density, and high-sensitivity strain acquisition on long-distance optical fibers, and has advantages such as electromagnetic interference resistance and flexible laying, enabling the construction of high-resolution two-dimensional strain sensing networks that accurately reflect foot contact areas, pressure distribution, and gait dynamics.

[0005] However, relying solely on local fiber strain signals still has limitations: first, there is a complex mechanical coupling relationship between different fiber segments, and traditional algorithms have difficulty modeling this relationship, requiring integration of relevant information for accurate recognition; second, relying solely on local strain signals makes it difficult to extract global features such as foot pressure contours and overall force structures, and the ability to identify subtle differences in similar gaits is insufficient. Therefore, there is an urgent need for a dual-modal gait recognition technology that integrates local fiber strain information and global foot contour features to fully leverage the sensing advantages of OFDR intelligent carpets and improve the accuracy and robustness of gait recognition. SUMMARY

[0006] The purpose of the present application is to provide a fiber-optic intelligent carpet gait recognition system based on strain-contour dual-modal network, which solves the problems of insufficient robustness, poor comfort and convenience, limited accuracy and generalization ability of existing gait recognition technologies based on vision, wearable devices and traditional ground sensors.

[0007] To achieve the above object, the application provides a fiber intelligent carpet gait recognition system based on strain-profile dual-mode network, which comprises an OFDR sensing system, a fiber sensing carpet and a local strain-global profile dual-mode attention network SCDA-Net. The OFDR sensing system is composed of a combination of a main interferometer and an auxiliary interferometer; the two interferometers share the same tuning laser source and realize light path distribution, reflection and interference signal output through fiber couplers with different splitting ratios and loopers; the main interferometer is used for injecting sweep frequency light into the measured fiber and receiving back Rayleigh scattering signals; the auxiliary interferometer is used for forming K-clock interference signals to compensate for the linearity error of the sweep frequency laser. The fiber sensing carpet uses silica gel as a flexible base, and the whole OFDR strain sensing fiber is laid in parallel on the flexible base at a fixed interval, fixed by silica gel glue, and covered with another layer of silica gel pad, which is integrally pressed with the silica gel glue and the flexible base; the OFDR strain sensing fiber forms a dense distributed strain sensing point array on the whole fiber sensing carpet. The local strain-global profile dual-mode attention network SCDA-Net includes a data preparation part, a strain map attention module SGAM, a pressure profile guided non-local module PGGNM and a final prediction module; the data preparation part pre-processes the original strain signal, inputs the original signal in the form of a two-dimensional matrix into the SGAM, and inputs the high-resolution image data after interpolation of the two-dimensional matrix into the PGGNM; the SGAM extracts and processes the features of the original strain signal, captures the dynamic correlation across the fiber segments and weights the key node features; the PGGNM extracts and processes the features of the image modal data, fuses the gradient information of the image and the non-local attention mechanism; the final prediction module combines the two kinds of features into the final gait recognition result.

[0008] Preferably, the main interferometer comprises a tuning laser source, a fiber coupler one, a fiber coupler two, a fiber coupler three, a fiber loop one, a fiber loop two, a polarization controller, a polarization beam splitter, a photodetector one and a fiber to be measured; the output end of the tuning laser source is connected to the input end of the fiber coupler one, 5% of the light output by the fiber coupler one is introduced into the port one of the fiber loop one, and 95% of the light is output to the fiber coupler two; the fiber coupler two inputs 10% of the light into the polarization controller and inputs 90% of the incident signal into the port one of the fiber loop two; the port two of the fiber loop two is connected to the input end of the fiber to be measured, and the back Rayleigh scattering signal output by the port three enters the fiber coupler three together with the output signal of the port two of the polarization controller and interferes, the two output ends of the fiber coupler three are connected to the two ports of the photodetector one through the polarization beam splitter, and the photodetector one outputs an electrical signal into a data acquisition system.

[0009] Preferably, the auxiliary interferometer comprises a fiber circulator one, a fiber coupler four, a delay fiber, a Faraday rotating mirror one, a Faraday rotating mirror two, a photodetector two and a photodetector two; the port one of the fiber circulator one receives reference light from the 5% output end of the fiber coupler one, and the port two is connected to the port one of the fiber coupler four; the port three of the fiber coupler four is connected to the delay fiber, the end of the delay fiber is connected to the Faraday rotating mirror one, and the long arm of the auxiliary interferometer is formed; the port four of the fiber coupler four is connected to the Faraday rotating mirror two, and the short arm of the auxiliary interferometer is formed; the interference beat frequency signal generated by mixing and interfering the short arm light and the long arm light in the fiber coupler four is output from the port two of the fiber coupler four and received by the photodetector two, and the output electrical signal of the photodetector two enters the data acquisition system as the K-clock time base correction signal of the system.

[0010] Preferably, in the data preparation part, the original strain signal collected by the fiber sensing carpet is processed to remove outliers, the strain data of the fiber segment is spliced into a two-dimensional matrix and input into the SGAM; after interpolation, a high-resolution image data is constructed, and after normalization, the high-resolution image data is input into the PGGNM.

[0011] Preferably, the SGAM constructs the original strain data into a graph structure data, and regards each fiber segment as a node in the graph, and the node feature is the strain sequence of the fiber in the length direction , all nodes are , the connection relationship between the nodes is constructed through K-neighbor, and the graph data is aggregated through the graph attention layer; for the node , the attention coefficient of each neighbor node is: wherein, is the original attention coefficient of the node to the neighbor node ; is a learnable attention parameter, represents vector splicing, is the node feature of the node ; the attention coefficient of the neighbor node is normalized through the Softmax function: wherein, represents the normalized attention coefficient; represents the neighbor set of the node ; the node feature is updated by aggregation: where, denotes the new features of the nodes after aggregation ; Finally, the importance of all nodes is calculated using the Softmax layer and weighted to obtain the output features : where, denotes the total number of nodes of the graph structure.

[0012] Preferably, the PGGNM extracts features and down-samples the input image through a convolutional encoder: where, denotes the input image, which is obtained by interpolating the original strain signal, denotes The data format of is a three-dimensional real array, is the number of channels, is the height, is the width; where, is the horizontal gradient of the image, is the vertical gradient of the image, is the horizontal gradient operator, is the vertical gradient operator, and the Sobel kernel is defined as: , ; Learn the overall gradient factor: where, is the Sigmoid function, is a convolution with a kernel size of ; After obtaining the gradient factor, it is injected into the non-local attention mechanism to map the image features to obtain Query vector, Key vector and Value vector: where, is the Query vector, is the Key vector, Value vector; Add gradient factor to Key vector: wherein, Enhanced Key vector; Flatten three feature maps into vectors with dimensions ; Calculate global attention of image: wherein, attention weight matrix ; Weight image features: ; Finally, dimension transform and residual connection are performed on the features after gradient information enhancement to obtain output features .

[0013] Preferably, the final prediction module outputs gait recognition results; two output features and are obtained after SGAM and PGGNM, which are changed into one-dimensional vectors through dimension transformation operation; after linear layer processing, they are added and then mapped into gait recognition results through a fully connected layer.

[0014] Therefore, the optical fiber intelligent carpet gait recognition system based on the strain-contour dual-modal network has the following beneficial effects: (1) The OFDR technology is adopted to construct a high-density optical fiber intelligent carpet in the application, and the optical fiber is embedded in a flexible base to form a large-area two-dimensional distributed strain sensing base. Compared with traditional pressure arrays or discrete sensing structures, the OFDR optical fiber carpet can realize continuous sampling, has strong anti-interference ability and high stability, and provides a high-quality original data basis for gait recognition; (2) The application proposes a local strain-global contour dual-modal attention network, which extracts the dynamic coupling relationship between optical fiber segments through a local strain map attention module, and strengthens the fine-grained features related to the key areas of gait. At the same time, the non-local module guided by the pressure gradient is used to capture the overall shape and contour structure of the foot bottom, and effectively encode the global information. The gait recognition accuracy is significantly improved; (3) Experiments are performed on 16 types of gaits, and the application achieves a classification accuracy of 97.55% and an AUC of 0.9994, which is significantly better than traditional methods, and the model has only 0.156M parameters, which has real-time recognition capability.

[0015] The technical solutions of the application will be further described in detail below with reference to the drawings and embodiments. DESCRIPTION OF DRAWINGS

[0016] Figure 1 The overall system framework of the embodiment of the present application is shown in the figure; Figure 2 The OFDR system schematic diagram of the embodiment of the present application is shown in the figure; Figure 3 The optical fiber sensing carpet schematic diagram of the embodiment of the present application is shown in the figure; Figure 4 The wiring mode diagram of the embodiment of the present application is shown in the figure; Figure 5 The local strain-global contour dual-modal attention network structure diagram of the embodiment of the present application is shown in the figure; Figure 6 The 16-class gait data signal diagram collected by the OFDR optical fiber carpet of the embodiment of the present application is shown in the figure; Figure 7 The key region extraction diagram of the local strain-global contour dual-modal attention network on two-dimensional strain data and image data of the embodiment of the present application is shown in the figure.

[0017] Reference signs 1, a tuning laser source; 2, a fiber coupler one; 3, a fiber coupler two; 4, a fiber coupler three; 5, a fiber coupler four; 6, a fiber circulator one; 7, a fiber circulator two; 8, a polarization controller; 9, a polarization beam splitter; 10, a delay fiber; 11, a Faraday rotating mirror one; 12, a Faraday rotating mirror two; 13, a photodetector one; 14, a photodetector two; 15, and a fiber to be measured. DETAILED DESCRIPTION

[0018] The following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but only represents selected embodiments of the application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0019] The optical fiber intelligent carpet gait recognition system based on strain-contour dual-modal network of the present application, as shown in the figure, is implemented step by step through the following steps, and the structures, data and network functions involved in each step are described in detail in combination with the accompanying drawings. Figure 1 Figures 1-7 I. Constructing an OFDR sensing system As shown in the figure, Figure 2 ​​As shown, the OFDR sensing system of the present application is composed of a tuning laser source 1, a fiber coupler one 2 (5:95), a fiber coupler two 3 (10:90), a fiber coupler three 4 (50:50), a fiber coupler four 5, a fiber circulator one 6, a fiber circulator two 7, a polarization controller 8, a polarization beam splitter 9, a delay fiber 10, a Faraday rotating mirror one 11, a Faraday rotating mirror two 12, a photodetector one 13, a photodetector two 14, and a fiber to be measured 15, etc. core devices, through the collaborative work of the main interferometer and the auxiliary interferometer to realize high-precision strain signal acquisition.

[0020] (I) Main interferometer setup The output end of the tuning laser source 1 is connected with the input end of the fiber coupler one 2, which provides the swept laser source for the whole system. The fiber coupler one 2 distributes the light signal according to the splitting ratio of 5:95, of which 5% of the light is introduced into the port one of the fiber circulator one 6, which is used for the reference light supply of the auxiliary interferometer; 95% of the light is transmitted to the fiber coupler two 3, which guides most of the optical power into the main path of the main interferometer. The fiber coupler two 3 further distributes the light signal according to the splitting ratio of 10:90, of which 10% of the light is input to the polarization controller 8, which is used for adjusting the polarization state of the reference arm light to ensure the stability of the interference signal; 90% of the incident signal is input to the port one of the fiber circulator two 7, the port two of the fiber circulator two 7 is connected with the input end of the fiber to be measured 15, and the back-end Rayleigh scattering signal reflected by the fiber to be measured 15 is output through the port three of the fiber circulator two 7, and the output signal of the port two of the polarization controller 8 is input into the input end of the fiber coupler three 4 and interferes with each other, which converts the phase information of the Rayleigh scattering light into a measurable interference light intensity signal. The two output ends of the fiber coupler three 4 are connected with the two ports of the photodetector one 13 through the polarization beam splitter 9 respectively, so as to suppress the common mode noise in a balanced detection manner, and the electrical signal output by the photodetector one 13 is connected with the data acquisition system, which is used for collecting the Rayleigh scattering interference signal formed by the main interferometer.

[0021] (II) Auxiliary interferometer setup Port one of the fiber circulator one 6 receives the reference light from the 5% output end of the fiber coupler one 2, and provides a light source for the auxiliary interferometer. Port two of the fiber circulator one 6 is connected with port one of the fiber coupler four 5, port three of the fiber coupler four 5 is connected with the delay fiber 10, the end of the delay fiber 10 is connected with the Faraday rotating mirror one 11, and the fiber circulator one 6, the fiber coupler four 5, the delay fiber 10 and the Faraday rotating mirror one 11 together constitute the long arm of the auxiliary interferometer, introduce a fixed optical path difference, and the reflected light returns to port three of the fiber coupler four 5; port four of the fiber coupler four 5 is connected with the Faraday rotating mirror two 12, and the fiber circulator one 6, the fiber coupler four 5, the delay fiber 10 and the Faraday rotating mirror two 12 together constitute the short arm of the auxiliary interferometer, and the reflected light returns to port four of the fiber coupler four 5. The short arm light and the long arm light are mixed and interfered in the fiber coupler four 5, an interference beat signal with a frequency proportional to the laser sweep speed is generated, the signal is output from port two of the fiber coupler four 5, is received by the photodetector two 14 as an auxiliary interference signal, and an electrical signal output by the photodetector two 14 is connected to a data acquisition system as a K-clock time base correction signal, so that an external trigger clock is provided for the data acquisition system, the main interferometer signal is ensured to be equally spaced sampled in the wave number domain, and the influence of the nonlinearity of the laser sweep is compensated.

[0022] II. Constructing a fiber sensing carpet As shown in Figure 3 , Figure 4 , the fiber sensing carpet of the present application uses flexible silica gel as a base and realizes high-resolution strain sensing by using a dense fiber wiring method.

[0023] The silica gel with a thickness of 1 mm is selected as a flexible base, and the whole OFDR strain sensing optical fiber is laid on the base in parallel at a fixed interval, the longitudinal interval between the optical fibers is set to 1 cm, the optical fiber is fixed on the base by using silica gel glue, and a 0.5 mm thick silica gel pad is covered on the upper layer, and the silica gel glue and the lower silica gel pad are integrally pressed to form an integrated structure. The sensing optical fiber forms a dense distributed strain sensing point array on the whole carpet, the strain sampling resolution of the OFDR host along the optical fiber direction is 2.6 mm, the carpet width is 50 cm, and the overall coverage and fine-grained strain collection of the foot contact area are ensured. In addition, the present application can also use other various dense wiring methods, and the distribution of the optical fiber is adjusted according to the actual application scene to ensure the uniformity and integrity of the strain sensing.

[0024] III. Constructing a local strain-global contour dual-modal attention network As shown in Figure 5 , a local strain-global contour dual-modal attention network SCDA-Net is constructed, the network includes a data preparation part, a strain graph attention module SGAM, a pressure contour guided non-local module PGGNM and a final prediction module, and high-precision gait recognition is realized through dual-modal feature fusion.

[0025] (I) Data preparation The original strain signal collected by the strain sensing optical fiber carpet is a two-dimensional matrix. First, the original strain signal is preprocessed to remove outliers, and the strain data of the optical fiber segment is spliced into a 38x146 two-dimensional matrix, which is directly input into the SGAM. The two-dimensional matrix is interpolated to construct a high-resolution image data of 256x256, which is normalized and input into the PGGNM to provide adaptive data for dual-mode feature extraction.

[0026] (II) Feature extraction of SGAM module SGAM constructs two-dimensional strain data into graph structure data. Since the dimension of two-dimensional strain data is 38x146, the feature length of each node is 146, and the number of nodes is 38. Each segment of optical fiber is regarded as a node in the graph, and the node feature is the strain sequence of the optical fiber in the length direction , all nodes are , the connection relationship between nodes is constructed by K-neighbor, which is set to 5. The graph data is aggregated by the graph attention layer. For node , the attention coefficient of each neighbor node is: wherein, is the original attention coefficient of the neighbor node of node ; is a learnable attention parameter, represents vector concatenation, is the node feature of node , and is a linear layer with a dimension of 256, so the dimension of the node feature is converted to 256; The attention coefficient of the neighbor node is normalized by the Softmax function: wherein, denotes the normalized attention coefficient; denotes the neighbor set of node ; The node feature is aggregated and updated: wherein, denotes the new feature of the aggregated node ; Finally, the importance of all nodes is calculated using the Softmax layer and weighted to obtain the output feature : wherein, the total number of nodes of the graph structure, The dimension of is 38x256.

[0027] (Three) PGGNM module feature extraction PGGNM extracts features and down-samples the input image through a convolutional encoder: wherein, represents the input image, which is obtained by interpolating the original strain signal, and the dimension is 256x256, represents The data format of is a three-dimensional real array, is the number of channels, is the height, is the width, and the specific values are both 64; The Sobel operator is used to calculate the horizontal and vertical pressure edge gradients: wherein, is the horizontal gradient of the image, is the vertical gradient of the image, is the horizontal gradient operator, is the vertical gradient operator, and the Sobel kernel is defined as: , ; Learning the overall gradient factor: wherein, is a Sigmoid function, is a convolution with a kernel size of ; After obtaining the gradient factor, it is injected into the non-local attention mechanism to map the image feature to obtain the Query vector, Key vector and Value vector: wherein, is the Query vector, is the Key vector, is the Value vector; The gradient factor is added to the Key vector: wherein, denotes the enhanced Key vector; The three feature maps are flattened into vectors with dimensions of ; The global attention of the image is calculated: where the attention weight matrix is a 4096x4096 square matrix representing the spatial correlation; ; The image features are weighted: ; Finally, the features after strengthening the gradient information are dimensionally transformed and residual connected to obtain the output features with dimensions of 64x64x64.

[0028] (Four) Prediction module output The two output features and are dimensionally transformed into one-dimensional vectors. After linear layer processing, they are added and mapped to the gait recognition result through the fully connected layer.

[0029] Four, use OFDR sensing system and optical fiber sensing carpet to collect real gait data, set the training network hyperparameters, and start training; Step 1, connect the constructed OFDR sensing system and the constructed optical fiber sensing carpet, and connect the computer for data collection. The collected data is divided into training data and test data according to a certain proportion. Arrange the subjects to perform a variety of specific gait actions on the optical fiber carpet, and collect the contact and force change process on the carpet surface through the OFDR system, thereby obtaining a high-density two-dimensional strain distribution sequence. The actual collected data set contains 2350 gait samples, covering 16 different gait categories, including 8 standing gaits and 8 sitting gaits, and the gait types include double feet together, double feet apart, lifting the back heel, lifting the front toe, inner eight, outer eight, and single foot. Figure 6 The collected 16 gait sample representatives are shown. The data set is divided into 60% training set, 15% validation set, and 25% test set, which can be used for network training, tuning, and performance evaluation.

[0030] Step 2, initialize network parameters and set training process hyperparameters. The training adopts batch iteration method, the batch size is set to 32, and a total of 15 iterations are performed. The learning rate is set to 0.0005, the optimizer is set to Adam optimizer, and the training begins.

[0031] V. The test set data is preprocessed in the same way as the training set, and the trained network is tested. This process does not update the network parameters. The accuracy, specificity, AUC, and other performance indicators of the test results are calculated. Tables 1 and 2 show the comparison of the proposed method and existing methods, and it can be seen that the proposed method has good performance. Figure 7 The proposed method is used to extract key regions in two modal data, and it can be seen that the proposed method can effectively focus on key regions.

[0032] Table 1 Comparison of the proposed method and traditional machine learning methods on 16 gait signals

[0033] Table 2 Comparison of the proposed method and deep learning methods on 16 gait signals

[0034] Therefore, the optical fiber intelligent carpet gait recognition system based on the strain-contour dual-modal network described above is adopted. The OFDR sensing technology is used to construct a high-density and high-sensitivity optical fiber sensing carpet. The dual-modal network composed of a local strain map attention module and a pressure gradient guided non-local module effectively integrates the dynamic coupling relationship between the optical fiber segments and the global contour features of the foot, solves the problems of insufficient robustness, poor comfort, limited accuracy, and lack of global feature extraction in existing gait recognition technology, and finally achieves a high classification accuracy of 97.55% for 16 types of gait and an AUC of 0.9994. The model has only 0.156M parameters, and has high precision, high robustness, and real-time recognition ability. It provides a reliable gait recognition solution for medical rehabilitation, security monitoring, and other fields.

[0035] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can still be modified or replaced by equivalents, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A fiber optic intelligent carpet gait recognition system based on strain-profile dual-mode network, characterized in that, This includes OFDR sensing systems, fiber optic sensing carpets, and the local strain-global profile dual-modal attention network SCDA-Net; The OFDR sensing system is composed of a main interferometer and an auxiliary interferometer. The two interferometers share the same tuned laser source and achieve optical path distribution, reflection and interference signal output through fiber couplers and circulators with different splitting ratios. The main interferometer is used to inject swept light into the fiber under test and receive backscattered Rayleigh signals. The auxiliary interferometer is used to generate K-clock interference signals to compensate for the linearity error of the swept laser. The fiber optic sensing carpet uses silicone as a flexible base. The entire OFDR strain sensing fiber is laid parallel on the flexible base at a fixed interval and fixed with silicone glue. Another layer of silicone pad is covered on top and pressed together with the silicone glue and the flexible base. The OFDR strain sensing fiber forms a dense distributed strain sensing point array on the entire fiber optic sensing carpet. The Local Strain-Global Contour Dual-Modal Attention Network (SCDA-Net) comprises a data preparation section, a strain map attention module (SGAM), a pressure profile-guided nonlocal attention module (PGGNM), and a final prediction module. The data preparation section preprocesses the raw strain signal, inputting the original signal in two-dimensional matrix form into SGAM and the high-resolution image data after two-dimensional matrix interpolation into PGGNM. SGAM extracts and processes features from the raw strain signal, capturing dynamic correlations across fiber segments and weighting key node features. PGGNM extracts and processes features from the image modal data, fusing gradient information from the image with the nonlocal attention mechanism. The final prediction module maps the combination of these two feature sets to the final gait recognition result.

2. The fiber optic intelligent carpet gait recognition system based on strain-profile dual-mode network according to claim 1, characterized in that, The main interferometer includes a tuned laser source, fiber coupler 1, fiber coupler 2, fiber coupler 3, fiber circulator 1, fiber circulator 2, polarization controller, polarization beam splitter, photodetector 1, and the fiber under test. The output of the tuned laser source is connected to the input of fiber coupler 1. 5% of the light output from fiber coupler 1 is introduced into port 1 of fiber circulator 1, and 95% of the light is output to fiber coupler 2. Fiber coupler 2 inputs 10% of the light to the polarization controller and 90% of the incident signal to port 1 of fiber circulator 2. Port 2 of fiber circulator 2 is connected to the input of the fiber under test. The Rayleigh scattering signal output from port 3 and the output signal from port 2 of the polarization controller enter fiber coupler 3 together and interfere. The two outputs of fiber coupler 3 are connected to the two ports of photodetector 1 through the polarization beam splitter. The electrical signal output from photodetector 1 enters the data acquisition system.

3. The fiber optic intelligent carpet gait recognition system based on strain-profile dual-mode network according to claim 2, characterized in that, The auxiliary interferometer includes fiber optic circulator I, fiber optic coupler IV, delay fiber, Faraday rotator mirror I, Faraday rotator mirror II, photodetector II, and photodetector II. Port 1 of fiber optic circulator I receives reference light from the 5% output end of fiber optic coupler I, and port 2 is connected to port 1 of fiber optic coupler IV. Port 3 of fiber optic coupler IV is connected to the delay fiber, and the end of the delay fiber is connected to Faraday rotator mirror I, forming the long arm of the auxiliary interferometer. Port 4 of fiber optic coupler IV is connected to Faraday rotator mirror II, forming the short arm of the auxiliary interferometer. The interference beat frequency signal generated by the mixing and interference of the short arm light and the long arm light in fiber optic coupler IV is output from port 2 of fiber optic coupler IV and received by photodetector II. The output electrical signal of photodetector II enters the data acquisition system as the K-clock time base correction signal of the system.

4. The fiber optic intelligent carpet gait recognition system based on strain-profile dual-mode network according to claim 1, characterized in that, In the data preparation section, the raw strain signals collected by the fiber optic sensing carpet are processed to remove outliers. The strain data of the fiber segments are stitched into a two-dimensional matrix and input into SGAM. After interpolation, the two-dimensional matrix is ​​used to construct a high-resolution image data, which is then normalized and input into PGGNM.

5. The fiber optic intelligent carpet gait recognition system based on strain-profile dual-mode network according to claim 4, characterized in that, SGAM constructs graph-structured data from the raw strain data, treating each fiber segment as a node in the graph, with each node characterized by the strain sequence along its length. All nodes are The graph data is aggregated through a graph attention layer after the K-nearest neighbor layer to build connections between nodes. Each of its neighboring nodes The attention coefficient is: in, For nodes For neighboring nodes The original attention coefficient; It is a learnable attention parameter. This represents vector concatenation. For nodes Node characteristics; For neighboring nodes The attention coefficients are normalized using the Softmax function: in, This represents the normalized attention coefficient; Represents a node The set of neighbors; Aggregate and update node features: in, Represents the node after aggregation New features; Finally, a Softmax layer is used to calculate the importance of all nodes and weight them to obtain the output features. : in, This represents the total number of nodes in the graph structure.

6. The fiber optic intelligent carpet gait recognition system based on strain-profile dual-mode network according to claim 5, characterized in that, PGGNM extracts and downsamples features from the input image using a convolutional encoder: in, This indicates that the input image is obtained by interpolation from the original strain signal. express The data format is a three-dimensional array of real numbers. For the number of channels, For height, Width; Calculate the transverse and longitudinal pressure edge gradients using the Sobel operator: in, The horizontal gradient of the image. The vertical gradient of the image. For the horizontal gradient operator, For the vertical gradient operator, the Sobel kernel is defined as: ， ； Learn the overall gradient factor: in, For the Sigmoid function, The kernel size is Convolution; After obtaining the gradient factor, it is injected into the nonlocal attention mechanism to process image features. The mapping process yields the Query vector, Key vector, and Value vector: in, For the query vector, For the key vector, For the Value vector; Add the gradient factor to the Key vector: in, This represents the enhanced key vector; Flatten the three feature maps into vectors with dimensions of . ; Calculate global attention for the image: Among them, the attention weight matrix ; Weighting of image features: ； Finally, the gradient-enhanced features are subjected to dimensionality transformation and residual connection to obtain the output features. .

7. The fiber optic intelligent carpet gait recognition system based on strain-profile dual-mode network according to claim 6, characterized in that, The final prediction module outputs gait recognition results; after SGAM and PGGNM, two types of output features are obtained. and The two vectors are transformed into one-dimensional vectors through a dimension transformation operation; after being processed by a linear layer, they are added together and then mapped to the gait recognition result through a fully connected layer.

Images