An indoor gait monitoring method and system based on distributed optical fiber sensing

Abstract

This invention discloses an indoor gait monitoring method and system based on distributed optical fiber sensing. The method utilizes a two-dimensional grid of sensing optical fibers laid beneath the indoor floor as the sensing component of a distributed optical fiber sensing system. The fiber spacing is less than half the average foot length of the monitored object, enabling blind-spot-free vibration sensing across the entire indoor area. Furthermore, a vibration data matrix containing spatial and temporal dimensions is generated. Through numerical simulation of the distributed optical fiber sensing system, a matching channel consistent with the statistical characteristics of the real system is constructed, generating a high-quality training set to train a multi-layer feedforward neural network to obtain a classification model. The real-time vibration data matrix is ​​input into the trained model to identify fall events and trigger an alarm response. This method achieves radiation-free, privacy-free, and wearable-free full-area gait monitoring, effectively solving the problem of scarce real fall samples and significantly improving the accuracy and generalization ability of fall recognition.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring and sensing technology, and in particular to an indoor gait monitoring method and system based on distributed optical fiber sensing. Background Technology

[0002] Indoor gait monitoring and fall detection have significant applications in elderly care, rehabilitation nursing, and family health management. Traditional monitoring methods mainly include camera-based visual surveillance, wearable sensing devices, and the newly emerging millimeter-wave radar sensing technology. However, these technologies still face significant limitations in practical applications: visual surveillance poses a privacy risk and its effectiveness is limited in low-light or obstructed areas; wearable devices rely on user wearing, leading to issues such as forgetting to wear them, inconvenient charging, and usage gaps; while millimeter-wave radar enables non-contact sensing, its long-term radiation risks and compliance pressures cannot be ignored.

[0003] In recent years, distributed fiber optic sensing technology has been applied in fields such as structural health monitoring and pipeline leak detection due to its advantages such as good concealment, wide coverage, and resistance to electromagnetic interference. For example, existing technologies disclose vibration event localization methods based on distributed fiber optic acoustic sensing, which determine the vibration location by demodulating the backscattered Rayleigh signal. However, such methods are mostly applicable to linear laying scenarios and are difficult to directly apply to gait monitoring in indoor two-dimensional planar areas. In addition, other technologies have proposed methods for determining the lateral distance of vibration sources based on the high-low frequency energy ratio, but these are still limited to direct burial or pipeline monitoring scenarios and fail to solve the problem of indoor gait event identification.

[0004] Currently, applying distributed fiber optic sensing technology to indoor gait monitoring still faces two key technical challenges: first, how to design fiber optic laying structures suitable for indoor floor coverage to effectively capture and locate gait vibrations; and second, how to construct effective gait pattern recognition algorithms for fiber optic vibration signals, especially when real fall samples are difficult to obtain, to achieve highly reliable fall event detection with a low false alarm rate. Existing technologies lack a complete solution that combines distributed fiber optic sensing with artificial intelligence recognition and systematically designs a solution for indoor two-dimensional gait monitoring. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the existing technology cannot achieve full coverage of indoor two-dimensional plane gait monitoring in a non-invasive and radiation-free manner, and lacks a high-precision and high-efficiency identification method for fall events in real-world scenarios.

[0006] To address the aforementioned technical problems, this invention provides an indoor gait monitoring method based on distributed optical fiber sensing, comprising: Two-dimensional grid-like distributed sensing optical fibers are laid under the indoor floor of the monitoring area to form the sensing part of the distributed optical fiber sensing system. The vibration signal of the sensing fiber is collected by the sensing part of the distributed optical fiber sensing system, and the vibration signal is demodulated into a vibration data matrix containing spatial and temporal dimensions. The initial neural network model is trained using the vibration data matrix. The trained neural network classification model is used to identify whether a fall event has occurred. If no fall event is identified, it is considered a normal gait event. If a fall event is identified, an alarm response is triggered.

[0007] In one embodiment of the present invention, the method for training the initial neural network model is as follows: Numerical simulations were performed on the distributed fiber optic sensing system to construct a matching channel. The matching channel The physical and statistical properties are consistent with those of the real system; Based on the matching channel A training set containing fall events and normal gait events is generated through numerical simulation; The initial neural network model is trained using the training set to obtain the neural network classification model.

[0008] In one embodiment of the present invention, based on the matching channel The method for generating a training set containing fall events and normal gait events through numerical simulation is as follows: Based on the matching channel The matching channel was obtained by simulating both fall events and normal gait events. Corresponding spacetime matrix The spatiotemporal matrix is ​​used as a training sample and its corresponding category label is labeled to form the training set.

[0009] In one embodiment of the present invention, the matching channel The physical and statistical properties include noise power spectral density and attenuation coefficient.

[0010] In one embodiment of the present invention, the initial neural network model is a multi-layer feedforward neural network, wherein the first... The mathematical expression for a layer is: , , in, For weighted input sum, For the first The weight matrix of the layer, For the first The layer's bias vector, For the first The output of the layer, For the first The activation function of the layer.

[0011] In one embodiment of the present invention, the spacing between the two-dimensional grid-like distributed sensing optical fibers is less than half the average foot length of the monitored object.

[0012] In one embodiment of the present invention, the method for demodulating the vibration signal into a vibration data matrix containing spatial and temporal dimensions is as follows: For each detection pulse emitted by the distributed optical fiber sensing system, the backscattered Rayleigh signal of the entire sensing fiber is simultaneously acquired to obtain the signal value of each spatial sampling point at that pulse moment. ,in For distance-oriented spatial sampling point index, This refers to the pulse emission time; With pulse period Repeat the detection pulse transmission and data acquisition steps; The signals acquired multiple times are arranged in chronological order to form the vibration data matrix. ,in The pulse ordinal number is represented by the column dimension of the matrix, which corresponds to the spatial sampling points. The row dimension corresponds to the time series. .

[0013] In one embodiment of the present invention, the vibration data matrix ,in This represents the number of spatial sampling points. This represents the number of sampling points within the time window.

[0014] Based on the same inventive concept, this invention also provides an indoor gait monitoring system based on distributed optical fiber sensing, comprising: A sensing fiber optic array, laid in a two-dimensional grid pattern under the indoor floor of the monitoring area, constitutes the sensing part of a distributed fiber optic sensing system. The signal demodulation unit is used to collect the vibration signal of the sensing optical fiber through the distributed optical fiber sensing system and demodulate the vibration signal into a vibration data matrix containing spatial and temporal dimensions. The model training and alarm unit is used to train the initial neural network model using the vibration data matrix. The trained neural network classification model is used to identify whether a fall event has occurred. If no fall event is identified, it is considered a normal gait event. If a fall event is identified, an alarm response is triggered.

[0015] The present invention also provides a fall alarm device, including the indoor gait monitoring system based on distributed optical fiber sensing.

[0016] The technical solution of the present invention has the following advantages compared with the prior art: The indoor gait monitoring method based on distributed optical fiber sensing described in this invention achieves radiation-free, blind-spot-free, and privacy-friendly gait perception across the entire area by laying a two-dimensional grid-like distributed optical fiber under the floor. Through physical link numerical simulation of the distributed optical fiber sensing system, a matching channel consistent with the statistical characteristics of the real system is constructed, generating a high-fidelity gait vibration spatiotemporal matrix as a training set. This effectively solves the problem of scarce real fall samples and significantly improves the accuracy and generalization ability of neural network fall recognition. Simultaneously, the optical fiber laying structure, vibration demodulation, simulation training, and intelligent decision-making are integrated into a complete closed-loop scheme, which is significantly superior to existing technologies in terms of non-intrusive monitoring, high-precision recognition, and engineering feasibility. Attached Figure Description

[0017] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0018] Figure 1 This is a flowchart illustrating the indoor gait monitoring method based on distributed optical fiber sensing provided in this embodiment of the invention. Figure 2 This is a schematic diagram of optical fiber laying in an embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0020] Example 1: like Figure 1 As shown, this invention provides an indoor gait monitoring method based on distributed optical fiber sensing, comprising: Two-dimensional grid-like distributed sensing optical fibers are laid under the indoor floor of the monitoring area to form the sensing part of the distributed optical fiber sensing system. The vibration signal of the sensing fiber is collected by the sensing part of the distributed optical fiber sensing system, and the vibration signal is demodulated into a vibration data matrix containing spatial and temporal dimensions. The initial neural network model is trained using the vibration data matrix. The trained neural network classification model is used to identify whether a fall event has occurred. If no fall event is identified, it is considered a normal gait event. If a fall event is identified, an alarm response is triggered.

[0021] This invention utilizes a two-dimensional grid-like sensing fiber optic cable laid under the floor as the sensing component of a distributed fiber optic sensing system to achieve non-intrusive and radiation-free vibration signal acquisition across the entire indoor area. The acquired vibration signals are demodulated into a vibration data matrix containing spatial and temporal dimensions, and then input into a high-precision neural network classification model trained using numerical simulation and matched channels. This model automatically identifies and distinguishes between normal gait and fall events, and triggers an alarm response immediately upon detecting a fall. This non-intrusive, full-coverage approach solves the privacy blind spots of traditional visual surveillance and the dependence on wearable devices, significantly improving recognition accuracy and generalization ability.

[0022] In this embodiment of the invention, a two-dimensional grid-like distribution of sensing optical fibers is laid under the indoor floor of the monitoring area. These sensing optical fibers serve as the sensing part of a distributed optical fiber sensing system and are used to sense minute vibrations on the ground caused by actions such as walking and falling.

[0023] Specifically, to achieve blind-spot-free coverage of a two-dimensional indoor plane while also considering the economics of fiber optic installation, the grid spacing needs to be optimized. If the fiber spacing is too large, when a footstep falls into the grid gap, it may not be able to effectively excite the fiber to generate a detectable vibration signal, creating a sensing blind spot; if the spacing is too small, it will unnecessarily increase the fiber length and system cost.

[0024] like Figure 2 As shown, the fiber optic cable laying spacing is set based on the characteristics of the contact area between the foot and the ground when a person walks. satisfy: ,in The average foot length of the monitored object is given. This condition ensures that at least one optical fiber is located within the foot projection area at any gait position, thus reliably capturing gait vibrations.

[0025] Preferably, the elderly population in nursing homes is a typical example, with an average foot length typically between 22cm and 26cm. In this embodiment of the invention, taking a woman's size 32 foot (22cm) as an example, the grid spacing is set to 10cm. This spacing allows for sufficient fiber optic vibration response while keeping the amount of fiber optic cable used per unit area within a reasonable range, balancing monitoring sensitivity and workload. The laid sensing fibers extend in a regular grid pattern to various areas of the room, including private spaces such as bedrooms and bathrooms, and are connected to distributed fiber optic demodulation equipment to form a complete whole-room gait sensing network.

[0026] Furthermore, after completing the two-dimensional grid laying of the sensing optical fiber, the vibration signal of the sensing optical fiber is collected by the sensing part of the distributed optical fiber sensing system, and the vibration signal is demodulated into a vibration data matrix containing spatial and temporal dimensions.

[0027] Specifically, in this embodiment of the invention, a phase-sensitive optical time-domain reflectometer (OTDR) architecture based on coherent back Rayleigh scattering is adopted. The basic principle is as follows: continuous coherent light generated by a narrow-linewidth laser is split into probe light and reference light at a ratio of 10:90 by an optical coupler. The probe light is modulated into narrow pulse light by an acousto-optic modulator and amplified, and then injected into the sensing optical fiber laid on the floor through an optical circulator. When external vibrations act on the optical fiber, the phase of the back Rayleigh scattered light in the optical fiber will change linearly with the strain. The scattered light and the reference light are coherently mixed in a mixer, converted into differential electrical signals by a balanced detector, and then converted into analog-to-digital signals and demodulated by a data acquisition card, thereby obtaining the vibration amplitude information at different times along each sensing position of the optical fiber.

[0028] In this embodiment of the invention, for each detection pulse, the acquisition card synchronously records the backscattered Rayleigh digital signal from all spatial sampling points along the entire sensing fiber, denoted as ,in This is a spatial sampling point index along the distance of the optical fiber, which physically corresponds to the equivalent one-dimensional distance encoding of the two-dimensional position coordinates of the underfloor optical fiber network after spatial multiplexing. This is the time when the pulse was emitted.

[0029] The above acquisition process uses a fixed pulse period. Repeated, meaning the system operates at a repetition frequency. Continuous probe pulses are emitted into the sensing fiber, and the backscattered signal of the entire fiber corresponding to each pulse is simultaneously acquired. The signals acquired by the secondary pulses can be arranged sequentially in time to construct a two-dimensional spatiotemporal matrix. ,in The pulse ordinal number represents the number of pulses, and the row dimension of the matrix corresponds to the time series. Column dimensions correspond to spatial sampling points Each element of the matrix Indicates the location place, time The relative change in backscattered light intensity is proportional to the axial strain or vibration amplitude of the optical fiber at that location, thus enabling the mapping of gait vibration events in physical space into a discrete digital signal matrix.

[0030] Furthermore, let the total number of spatial sampling points along the optical fiber be... , The sampling rate, pulse width, and fiber length are all determined by the system sampling rate, pulse width, and fiber length, and are constants when the hardware conditions are fixed; let the number of sampling points within the time window be... And satisfy ,in For the duration of observation, The sampling interval is typically the pulse period. Therefore, the vibration data matrix can be fully represented as follows: That is, one OK, A column of real matrices.

[0031] This matrix simultaneously encodes the spatial distribution characteristics and temporal evolution of gait events: from the column direction, each instant corresponds to a set of instantaneous vibration spatial distributions along the entire length of the optical fiber; from the row direction, each spatial point corresponds to a waveform sequence in which the vibration amplitude changes over time. The two-dimensional spatiotemporal matrix is ​​not only a complete representation of physical vibration signals in the digital domain, but also a direct input to subsequent neural network classification models. Its quality and resolution directly determine the upper limit of gait recognition accuracy, providing a structured data foundation for deep learning-based gait pattern recognition.

[0032] Furthermore, in this embodiment of the invention, an initial neural network model is trained and used for fall event recognition.

[0033] Since real-world falls are low-probability events, and collecting training samples through repeated falls by real people poses ethical and safety risks, it is difficult to obtain sufficient and diverse labeled data for neural network training. To address this technical problem, in this embodiment of the invention, a full physical link numerical simulation is performed on the distributed optical fiber sensing system to construct a matching channel that highly matches the statistical characteristics of the real system. .

[0034] Specifically, by analyzing the optical parameters of the sensing fiber (including the attenuation coefficient) Accurate modeling is performed on factors such as backscattering coefficient, optoelectronic device characteristics (laser linewidth, modulator response, detector noise), demodulation algorithms (phase extraction, spatial resolution), and environmental noise (thermal noise, shot noise, 1 / f noise) to generate a virtual sensing channel that is mathematically equivalent to the real hardware system. This matched channel Its physical and statistical properties are consistent with those of the real system, mainly including the noise power spectral density. and attenuation coefficient This ensures that the simulated data generated on the virtual channel has the same distribution characteristics as the real collected data.

[0035] Furthermore, in this embodiment of the invention, based on the constructed matching channel Numerical simulations were used to generate vibration signals from numerous fall events and normal gait events. During the simulation, the contact force time-history curve between the foot and the ground was set according to a human biomechanical model, and factors such as walking speed, stride length, weight, and ground material were considered to generate diverse gait excitation sources. These gait excitation sources were then applied as input to a matched channel. By solving for the fiber strain and phase response, the spatiotemporal matrix corresponding to the matched channel is obtained. .

[0036] Matching the spatiotemporal matrix of the channel with the vibration data matrix acquired by the actual system Having the exact same dimensions and physical meaning, its elements Indicates the position in the virtual channel place, time The change in backscattered light intensity.

[0037] Furthermore, each spatiotemporal matrix sample is labeled with a corresponding category: a fall event is labeled as "1", and a normal gait event is labeled as "0", thus forming a training set containing tens of thousands of high-fidelity samples. The training set not only covers a wide range of gait variations and noisy conditions, but also has a sufficient number of samples and is balanced between classes, providing a solid data foundation for the learning of the neural network model.

[0038] Furthermore, in this embodiment of the invention, a multi-layer feedforward neural network is used as the initial neural network model. The structural design of the multi-layer feedforward neural network is fully adapted to the characteristics of the vibration spatiotemporal matrix. Since the input layer needs to receive a two-dimensional spatiotemporal matrix... All the information is used to flatten the matrix into a one-dimensional vector before inputting it into the network; therefore, the number of nodes in the input layer is set to [value missing]. The output layer has 2 nodes, corresponding to the two states of falling and normal gait, and uses the Softmax activation function to convert the output into a class probability distribution. The number of hidden layers and the number of neurons in each layer can be adjusted according to the actual task complexity. A typical configuration can contain 2 to 3 fully connected hidden layers, with 128 to 256 neurons in each layer.

[0039] The forward propagation process of a multilayer feedforward neural network consists of linear transformations and nonlinear activations in each layer. For the th... layer( Its mathematical expression is: , , in, For weighted input sum, For the first The weight matrix of the layer, For the first The layer's bias vector, For the first The output of the layer, For the first The activation function of the layer.

[0040] Hidden layers typically use the ReLU function (i.e., To introduce nonlinearity and alleviate the gradient vanishing problem, the output layer uses the Softmax function to achieve binary classification probability output.

[0041] Using the training set constructed above, the network parameters are optimized through backpropagation and gradient descent optimizer. The model is iteratively optimized to minimize the cross-entropy loss function until it converges, thus obtaining the trained neural network classification model.

[0042] During the actual monitoring phase, the vibration data matrix will be acquired and demodulated in real time. After being flattened in the same way, the data is input into a trained neural network classification model. The model performs forward computation and outputs the probability that the current gait event belongs to a fall. A decision threshold is set (e.g., 0.5). If the output probability is greater than the threshold, a fall event is determined, and an alarm response is immediately triggered (e.g., local audible and visual alarm, remote push notification, etc.). Otherwise, it is considered a normal gait event, and monitoring continues. Through this mechanism, this invention achieves all-weather, high-precision, and low-false-alarm intelligent recognition of indoor gait events.

[0043] Example 2: Based on the same inventive concept as Embodiment 1, the present invention also provides an indoor gait monitoring system based on distributed optical fiber sensing, used to implement the steps of the indoor gait monitoring method based on distributed optical fiber sensing described in Embodiment 1, including: A sensing fiber optic array, laid in a two-dimensional grid pattern under the indoor floor of the monitoring area, constitutes the sensing part of a distributed fiber optic sensing system. The signal demodulation unit is used to collect the vibration signal of the sensing optical fiber through the distributed optical fiber sensing system and demodulate the vibration signal into a vibration data matrix containing spatial and temporal dimensions. The model training and alarm unit is used to train the initial neural network model using the vibration data matrix. The trained neural network classification model is used to identify whether a fall event has occurred. If no fall event is identified, it is considered a normal gait event. If a fall event is identified, an alarm response is triggered.

[0044] In this embodiment of the invention, the sensing fiber array uses single-mode or multi-mode optical fibers, laid in a regular grid pattern between the interior floor decoration layer and the structural layer. The spacing of the fiber grid is designed to be less than half the average foot length of the monitored object. Taking a typical application scenario in a nursing home as an example, the average foot length of the elderly is approximately 22cm to 26cm. In this embodiment, the grid spacing is preferably set to 10cm, thereby ensuring that at least one optical fiber is located within the foot projection range at any gait landing point, effectively avoiding sensing blind spots. The laid sensing fiber array covers all indoor areas such as bedrooms, bathrooms, and corridors, and is connected to the back-end demodulation equipment through an optical cable splice box, forming a complete whole-room gait sensing network.

[0045] The signal demodulation unit adopts a phase-sensitive optical time-domain reflectometer architecture based on coherent back Rayleigh scattering. Specifically, the signal demodulation unit includes a narrow-linewidth laser, an optical coupler, an acousto-optic modulator, an erbium-doped fiber amplifier, an optical circulator, a mixer, a balanced detector, and an FPGA-based high-speed data acquisition card and a phase demodulation module. Its operation is as follows: continuous coherent light emitted from the narrow-linewidth laser is split into probe light and reference light at a 10:90 ratio by the optical coupler; the probe light is modulated into narrow pulse light by the acousto-optic modulator and amplified, then injected into the sensing fiber array laid on the floor through the optical circulator; when external gait vibrations act on the fiber, the phase of the back Rayleigh scattered light changes proportionally to the strain; this scattered light returns and is coherently mixed with the reference light in the mixer, then converted into a differential electrical signal by the balanced detector; the data acquisition card performs analog-to-digital conversion on the electrical signal, and the FPGA module demodulates the vibration amplitude and phase information at different times along the fiber in real time. For each detection pulse, signals from all spatial sampling points along the entire sensing fiber are simultaneously acquired and denoted as follows: ,in This is the spatial sampling point index (corresponding to the one-dimensional distance code of the two-dimensional grid under the floor after spatial multiplexing). This refers to the pulse emission time. A fixed pulse period is used. Continuous transmission of detection pulses will continuously The signals collected in this batch are arranged in chronological order to form a vibration data matrix. ,in This represents the number of spatial sampling points. This represents the number of sampling points within the time window. This matrix simultaneously encodes the spatial distribution and temporal evolution characteristics of gait events, serving as the direct data foundation for subsequent intelligent recognition.

[0046] The processing and alarm unit employs an embedded processor or industrial control computer, internally pre-loaded with a neural network classification model trained offline. This model is a multi-layer feedforward neural network, and its training process has been detailed in Example 1: by performing full physical link numerical simulation of the distributed optical fiber sensing system, a matching channel consistent with the statistical characteristics of the real system is constructed. Based on this matching channel, a massive amount of high-fidelity spatiotemporal matrices of fall and normal gait vibrations are generated. As training samples, labeled samples are used to supervise the training of the network, ultimately resulting in a neural network classification model that can be deployed in practical systems. In actual operation, the processing and alarm unit receives the vibration data matrix output by the signal demodulation unit in real time. The gait is flattened into a one-dimensional vector and input into a neural network model. The model calculates the probability that the current gait event is a fall. If the probability exceeds a preset threshold (e.g., 0.5), it is determined that a fall event has occurred, and the processing and alarm unit immediately triggers an alarm response, including activating a local audible and visual alarm, sending an alarm message to the caregiver's mobile phone or management platform via a wireless communication module, and enabling two-way voice communication through an intercom system. If the probability does not exceed the threshold, it is considered a normal gait event, and the system continues to monitor. This unit also has data storage and log recording functions, enabling long-term statistical analysis of gait patterns and providing decision support for health management.

[0047] This embodiment integrates a two-dimensional gridded fiber optic sensing array, a high-precision fiber optic signal demodulation module, and an artificial intelligence gait recognition processor into a single system, achieving full-area, contactless, radiation-free, and highly reliable fall monitoring in indoor environments. It can be widely deployed in typical scenarios such as elderly care institutions, rehabilitation hospitals, and home-based elderly care.

[0048] Example 3: The present invention also provides a fall alarm device, including the indoor gait monitoring system based on distributed optical fiber sensing in Embodiment 2.

[0049] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0050] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0051] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0052] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0053] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. An indoor gait monitoring method based on distributed optical fiber sensing, characterized in that, include: Two-dimensional grid-like distributed sensing optical fibers are laid under the indoor floor of the monitoring area to form the sensing part of the distributed optical fiber sensing system. The vibration signal of the sensing fiber is collected by the sensing part of the distributed optical fiber sensing system, and the vibration signal is demodulated into a vibration data matrix containing spatial and temporal dimensions. The initial neural network model is trained using the vibration data matrix, and the trained neural network classification model is used to identify whether a fall event has occurred. If no fall event is detected, it is considered a normal gait event; if a fall event is detected, an alarm response is triggered.

2. The indoor gait monitoring method based on distributed optical fiber sensing according to claim 1, characterized in that: The method for training the initial neural network model is as follows: Numerical simulations were performed on the distributed fiber optic sensing system to construct a matching channel. The matching channel The physical and statistical properties are consistent with those of the real system; Based on the matching channel A training set containing fall events and normal gait events is generated through numerical simulation; The initial neural network model is trained using the training set to obtain the neural network classification model.

3. The indoor gait monitoring method based on distributed optical fiber sensing according to claim 2, characterized in that: Based on the matching channel The method for generating a training set containing fall events and normal gait events through numerical simulation is as follows: Based on the matching channel The matching channel was obtained by simulating both fall events and normal gait events. Corresponding spacetime matrix The spatiotemporal matrix is ​​used as a training sample and its corresponding category label is labeled to form the training set.

4. The indoor gait monitoring method based on distributed optical fiber sensing according to claim 2, characterized in that: The matching channel The physical and statistical properties include noise power spectral density and attenuation coefficient.

5. The indoor gait monitoring method based on distributed optical fiber sensing according to claim 2, characterized in that: The initial neural network model is a multi-layer feedforward neural network, whose first... The mathematical expression for a layer is: ， ， in, For weighted input sum, For the first The weight matrix of the layer, For the first The layer's bias vector, For the first The output of the layer, For the first The activation function of the layer.

6. The indoor gait monitoring method based on distributed optical fiber sensing according to claim 1, characterized in that: The spacing between the two-dimensional grid-like distributed sensing optical fibers is less than half the average foot length of the monitored object.

7. The indoor gait monitoring method based on distributed optical fiber sensing according to claim 1, characterized in that: The method for demodulating the vibration signal into a vibration data matrix containing spatial and temporal dimensions is as follows: For each detection pulse emitted by the distributed optical fiber sensing system, the backscattered Rayleigh signal of the entire sensing fiber is simultaneously acquired to obtain the signal value of each spatial sampling point at that pulse moment. ,in For distance-oriented spatial sampling point index, This refers to the pulse emission time; With pulse period Repeat the detection pulse transmission and data acquisition steps; The signals acquired multiple times are arranged in chronological order to form the vibration data matrix. ,in The pulse ordinal number is represented by the column dimension of the matrix, which corresponds to the spatial sampling points. The row dimension corresponds to the time series. .

8. The indoor gait monitoring method based on distributed optical fiber sensing according to claim 7, characterized in that: The vibration data matrix ,in This represents the number of spatial sampling points. This represents the number of sampling points within the time window.

9. An indoor gait monitoring system based on distributed optical fiber sensing, characterized in that, include: A sensing fiber optic array, laid in a two-dimensional grid pattern under the indoor floor of the monitoring area, constitutes the sensing part of a distributed fiber optic sensing system. The signal demodulation unit is used to collect the vibration signal of the sensing optical fiber through the distributed optical fiber sensing system and demodulate the vibration signal into a vibration data matrix containing spatial and temporal dimensions. The model training and alarm unit is used to train the initial neural network model using the vibration data matrix. The trained neural network classification model is used to identify whether a fall event has occurred. If no fall event is detected, it is considered a normal gait event; if a fall event is detected, an alarm response is triggered.

10. A fall alarm device, characterized in that, Including the indoor gait monitoring system based on distributed optical fiber sensing as described in claim 9.

Images