A device, system and method for blind positioning and yawing early warning based on optical fiber vibration signals

Abstract

The application provides a device, system and method for blind positioning and yaw early warning based on an optical fiber vibration signal. The device comprises an optical time domain reflectometer module, a signal preprocessing module, a blind person / non-blind person discrimination module and a positioning and yaw early warning module. The system provides a high blind user recognition rate and high positioning accuracy without the need for video monitoring equipment and wearable sensing equipment, and more comprehensively and effectively provides application support for the field of blind travel. The system enhances the adaptability for different human body recognition, identifies the position of a visually impaired user at a low time delay and a high accuracy, and provides early warning for the user in a timely manner when the visually impaired user deviates from walking, so as to provide related services for the visually impaired user in the future.

Description

Technical Field

[0001] This invention belongs to the field of intelligent mobility technology, and in particular relates to a device, system and method for blind positioning and yaw warning based on fiber optic vibration signals. Background Technology

[0002] Despite the rapid development of internet and wireless communication technologies, the mobility of visually impaired users remains significantly limited. Currently, visually impaired individuals rely on tactile paving, guide canes, guide dogs, and professional guide assistants for assistance when traveling. However, canes and guide dogs can only distinguish obstacles on the ground, not tree branches extending into the air. Furthermore, many public places in cities are unfriendly to guide dogs, and the environment itself can interfere with their work. In complex urban environments, the lack of maintenance for infrastructure such as tactile paving often renders the use of canes ineffective. Additionally, the labor costs for guide assistants are high, and the number of professional guide assistants is limited, making this solution impractical for visually impaired users in non-first-tier cities. Indoor walking environments are even more complex.

[0003] Recently, many vision-based mobility solutions for the visually impaired have emerged, employing object recognition algorithms. However, these solutions have limited adaptability to unknown outdoor environments, and the computational time and energy costs severely restrict their application scenarios. For the vast visually impaired population, a low-cost, low-latency, and high-performance mobility solution remains lacking. Summary of the Invention

[0004] The purpose of this invention is to address the problems in existing technologies by proposing a device, system, and method for blind person positioning and yaw warning based on fiber optic vibration signals. This method overcomes the limitations of existing vision- or sensor-based human body recognition applications and the inconvenience of using wearable devices. It enhances adaptability to different environments, especially unfamiliar ones, and calculates the location of visually impaired users at the current target location with low latency and high accuracy, thus paving the way for future services for visually impaired users.

[0005] The present invention is achieved through the following technical solution: The present invention proposes a device for blind positioning and yaw warning based on optical fiber vibration signal. The device includes an optical time domain reflectometer module, a signal preprocessing module, a blind / non-blind person discrimination module, and a positioning and yaw warning module.

[0006] The optical time-domain reflectometer module is used to collect photoelectric signal intensity data generated by walking vibrations;

[0007] The signal preprocessing module is used for preprocessing the extracted two-dimensional intensity signal, including filtering out the bottom noise of the collected data, removing the non-target vibration signal static frame, calculating the number of pedestrians contained in the input signal, and separating the vibration signals of different pedestrians.

[0008] The blind person / non-blind person discrimination module is used for predicting whether each pedestrian in the signal is a blind person, interpolating and denoising the one-dimensional maximum amplitude time domain signal of each pedestrian, calculating the gait feature, inputting the feature vector into a classification algorithm for judgment.

[0009] The positioning and yawing early warning module calculates the walking position of the visually impaired pedestrian in the original signal and tracks it; if the position of the pedestrian deviates from the fiber blind path, an alarm is given to remind the pedestrian of the correct walking direction.

[0010] The application provides a method for positioning and yawing early warning of blind people based on fiber vibration signals.

[0011] Step 1: Start the optical time domain reflectometer and record the data S(z, t) of the target user walking, obtain the generated signal intensity data and output it in the form of a TXT document;

[0012] Step 2: Preprocess the obtained two-dimensional signal; the preprocessing includes noise reduction, removal of static frames, calculation of the number of pedestrians and separation of the vibration signals of different pedestrians; when reducing noise, search for a noise threshold X z >150 threshold , set the threshold to X threshold , consider the signal with an amplitude lower than the threshold as noise and directly assign it a value of zero, and define the useful signal interval as the average interval of the non-zero value signals after noise reduction; when removing static frames, an energy threshold E threshold is needed, the signal is divided into multiple frames in the time dimension, the energy E r of each frame is calculated, and it is judged whether the current frame is a static frame, finally the case of "walking-stop-walking" is checked, and the damage of the removed static frame to the original step interval is corrected; finally, the start and end points of each frame after correction are calculated, the frames are spliced, and the two-dimensional signal of the combined multiple frames is output; when calculating the number of pedestrians, an observation window is set, the number of symmetric peak groups in each observation window is calculated, since the left and right laying of the measuring optical fiber has symmetry, the number of symmetric peak groups is the number of pedestrians N; since the number of pedestrians may change dynamically, the period when the number of pedestrians is stable is counted, and the two-dimensional signal of the period is extracted for separation; when separating the vibration signals of different pedestrians i, the position point where the amplitude of each symmetric wave group falls to zero is recorded According to cut the two-dimensional signal;

[0013] Step 3: Interpolation and denoising of the vibration signal of each pedestrian, calculation of gait features, input of the feature vector into a classification algorithm for judgment; if a pedestrian is determined to be a blind user, step 4 is entered;

[0014] Step 4: Positioning and yaw warning of the blind user; first, the step point peak time coordinates obtained when calculating the gait features are used to calculate the position of the user in the current step Search for the position of the step point peak in the two-dimensional signal Then, the position estimation value calculated by the square weighting of the signal amplitude is used The step point peak position is combined The position z of the kth step of the pedestrian is determined together k ; the left and right fiber measurements of the energy of each step when calculating the gait features are used to determine if yaw is needed, if the energy measured by one side fiber is greater than the energy measured by the other side fiber for three consecutive steps, the user is given a yaw warning and guidance on the correct walking direction.

[0015] Further, whether each pedestrian in the predicted signal is blind, specifically:

[0016] Step 3.1, interpolation and denoising of the one-dimensional maximum amplitude time domain signal X0(t) of each pedestrian, in the one-dimensional maximum amplitude time domain signal, the vibration signal of each step may have a small amplitude part removed in the denoising process, and a large amplitude noise may be ignored and retained in the denoising process; set the interpolation interval threshold D threshold , calculate the time interval d of the signal from the non-zero signal before and after each sampling time, if d≤D threshold and the current time signal is 0, interpolation is performed, and the amplitude value is inserted as the average of the amplitudes of the non-zero signals before and after; if d≥D threshold and the current time signal is not 0, denoising is performed, and the amplitude of the point is set to 0; the maximum amplitude time domain signal obtained after interpolation and denoising is denoted as X(t);

[0017] Step 3.2, gait feature extraction of the signal of each pedestrian, extract the following 5 groups of features, the feature vector [f i,1 , f i,2 , f i,3 , f i,4 , f i,5 ] is defined as follows:

[0018] (1) f i,1 : step frequency, i.e. the estimated value of the number of steps per minute;

[0019] (2) f i,2 : step length, i.e. the average value of the single step distance;

[0020] (3) f i,3: Energy average value of vibration signal;

[0021] (4)f i,4 : Zero rate of maximum amplitude time domain signal X(t);

[0022] (5)f i,5 : Spatial span average value of vibration signal;

[0023] Step 3.3, combine all features into a feature vector, and normalize the feature vector; each time a sample is randomly taken from the training sample set, then a sample is found from the same sample set, and a sample is found from each different sample set, then the weight of each feature is updated, the ReliefF algorithm is used to screen the feature weight, all features are input into the CNN network for convolution classification, and it is judged whether the pedestrian is a visually impaired user.

[0024] Further, the step frequency is defined as the number of step points per unit time, and the calculation method of the step frequency is specifically:

[0025] 1.1, record the monotonicity change of short-time accumulation of correlation function R(m) by using sliding window method; first, set the sliding window length L, single sliding distance D, step point peak time coordinate array T, observation window length L w ; calculate the sliding number M, initialize the flag one-dimensional array, initialize the dynamic array G for storing autocorrelation peak value, calculate the signal length n, and calculate the short-time accumulation of correlation function The mean value in each sliding window is compared with the mean value in the previous and next windows, and the flag value is set to record the autocorrelation and change trend;

[0026] 1.2, initialize the step point peak time coordinate array, search for the peak coordinate, and the corresponding coordinate is the time position coordinate corresponding to the step point; initialize the window number j as 1, and when j

[0027] flag(j)+flag(j+1)+flag(j+2+p)=3+p,p≥0

[0028] Then, the judgment is performed again, and if

[0029] flag(j+p+1)+flag(j+p+1+q)=-2-q,q≥0,

[0030] Then, the maximum value coordinate of the window j+2+p is stored in G;

[0031] 1.3, calculate the length L G of the array G step N G-1, initialize the peak time coordinate array T of the step points. p L before X(t) W Intra-point search peak coordinate T P (1) For the kth value, it can be obtained through formula T P (k)=T P (k-1)+G(k)-G(k-1) is calculated;

[0032] 1.4 The output step frequency is: f i,1 =F=60×f s ×N step / [T P (L G )-T P (1)].

[0033] Furthermore, the stride length is the distance between adjacent step points, and its specific calculation method is as follows:

[0034] Input the number of steps N, the signal in a single frame, and the peak coordinates. For the k-th step, when k does not exceed L... G When, calculate P k (z), and estimate the position. k exceeds L G When, calculate S1, S2, S; where,

[0035] P k (z) is calculated by the following formula:

[0036]

[0037] In the formula, t range =200;

[0038] Calculated by the following formula:

[0039]

[0040] The formulas for calculating S1, S2, and S are as follows:

[0041] Unit: cm

[0042] Unit: cm

[0043]

[0044] Furthermore, the average energy of the signal can be calculated using the following formula:

[0045]

[0046] Furthermore, the zeroing rate can be calculated using the formula given below:

[0047]

[0048] In the formula, II (·) is an indicator function, and if · is true, the value is 1, otherwise the value is 0.

[0049] Further, the space span average value can be calculated by the following formula:

[0050] W k =z kmax -z kmin , P k (z) > 0, z [z kmin , z kmax ]

[0051]

[0052]

[0053] The space span reflects the strength and propagation ability of the vibration signal, for the input two-dimensional signal, the length L of the observation window is set OB , the number N of the calculation window is calculated W , after calculating the space span of the current window, the next window is calculated, and finally the average value of the space span of each window is taken as the output.

[0054] The application provides a system for blind positioning and yawing early warning based on an optical fiber vibration signal.

[0055] The application has the following beneficial effects:

[0056] The application provides a device, system and method for blind positioning and yawing early warning based on an optical fiber vibration signal, which provides a higher blind user recognition rate and higher positioning accuracy without video monitoring equipment and wearable sensing equipment, and more comprehensively and effectively provides application support for the field of blind travel. The system enhances the adaptability to different human body recognition, recognizes the position of the visually impaired user at a lower time delay and a higher accuracy, and provides early warning for the user in time when the visually impaired user deviates from the walking direction, so as to provide related services for the visually impaired user in the future. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 is a process block diagram of the device.

[0058] Figure 2 is a schematic diagram of a two-dimensional intensity-time signal collected by the optical time domain reflectometer in the device.

[0059] Figure 3is a schematic diagram of a one-dimensional maximum amplitude time domain signal obtained by a pre-processing module in the device of the present application.

[0060] Figure 4 is a comparison schematic diagram of a two-dimensional signal before and after noise reduction in the device of the present application.

[0061] Figure 5 is a comparison schematic diagram of separation of vibration signals of different pedestrians in the device of the present application.

[0062] Figure 6 is a comparison schematic diagram of interpolation and noise reduction of a one-dimensional maximum amplitude time domain signal in the device of the present application.

[0063] Figure 7 is a physical diagram of an optical time domain reflectometer used in the device of the present application. DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0065] Referring to Figures 1-7 , the present application proposes a device for positioning and yawing early warning of blind people based on optical fiber vibration signals, which comprises an optical time domain reflectometer module (OTDR, Optical Time Domain Reflectometer), a signal pre-processing module, a blind person / non-blind person discrimination module and a positioning and yawing early warning module.

[0066] The optical time domain reflectometer module is used for collecting photoelectric signal intensity data generated by walking vibration.

[0067] The signal pre-processing module is used for pre-processing the extracted two-dimensional intensity signal, including filtering the collected data to remove the noise, removing the stationary frame of the vibration signal without target, calculating the number of pedestrians contained in the input signal and separating the vibration signals of different pedestrians.

[0068] The blind person / non-blind person discrimination module is used for predicting whether each pedestrian in the signal is a blind person, interpolating and denoising the one-dimensional maximum amplitude time domain signal of each pedestrian, calculating the gait feature, inputting the feature vector into a classification algorithm for judgment.

[0069] The positioning and yawing early warning module calculates the walking position of the visually impaired pedestrian in the original signal and tracks it for the blind person target. If the position of the pedestrian deviates from the possibility of the optical fiber blind path, an alarm is given to remind and indicate the correct walking direction of the pedestrian.

[0070] This invention proposes a method for blind person localization and yaw warning based on fiber optic vibration signals. The method is implemented using the aforementioned device for blind person localization and yaw warning based on fiber optic vibration signals. Specifically, the method comprises:

[0071] Step 1: Start the optical time domain reflectometer and record the target user's walking data S(z,t), obtain the signal intensity data generated by it and output the data in the form of a TXT document; use a data processing tool to read the two columns of data in the TXT document and store them in the time and intensity arrays respectively;

[0072] Step 2: Preprocess the obtained two-dimensional signal; preprocessing includes noise reduction, removal of still frames, calculation of pedestrian count, and separation of vibration signals from different pedestrians; during noise reduction, the useful signal interval D is searched. z Noise threshold X > 150 threshold And set the threshold to X threshold The filter considers signals with amplitudes below a certain threshold as noise and assigns them a value of zero. The useful signal interval is defined as the average interval of the non-zero signals after noise reduction. When removing still frames, an energy threshold E needs to be set. threshold The signal is divided into multiple frames in the time dimension, and the energy E of each frame is calculated. r The algorithm then determines whether the current frame is a still frame. If the energy is higher than a threshold, it is considered a non-still frame; otherwise, it is considered a still frame. Finally, it checks for "walk-stop-walk" patterns. For a non-still frame with frame number r, it checks whether its K adjacent frames are still frames. If a still frame appears in the adjacent frames of the non-still frame, it indicates that the experimental data exhibits a "walk-stop-walk" phenomenon. The step frequency algorithm is used to extract the step interval Q0 of the original non-still frame and the step interval Q after the still frame. If Q0-Q is greater than 0, then Q0-Q zeros are padded between the two frames after removing the still frame. Otherwise, delete Q-Q0 zeros between the two frames after removing the still frame to correct the disruption of the original step interval caused by discarding the stop frame; finally, calculate the start and end points of each frame after correction, stitch the frames together and output them; when calculating the number of pedestrians, set an observation window and calculate the number of symmetrical peak groups in each observation window. Since the measurement fiber is laid symmetrically on both sides, the number of symmetrical peak groups is the number of pedestrians N; since the number of pedestrians may change dynamically, statistically analyze the time period when the number of pedestrians is stable, extract the two-dimensional signal of the time period for separation; when separating the vibration signals of different pedestrians i, record the position point where the amplitude of each symmetrical wave attenuates to zero. according to Cutting two-dimensional signals;

[0073] Step 3: Interpolate and denoise the vibration signal of each pedestrian, calculate gait features, and input the feature vector into the classification algorithm for judgment; if a pedestrian is judged to be a blind user, proceed to step 4;

[0074] Step 4: Localization and yaw alarm for blind users; first, based on the peak time coordinates of gait points obtained when calculating gait characteristics. Searching for peak positions of step points in a two-dimensional signal The position estimate is then calculated using the weighted average of the squared signal amplitudes. Combined with peak position of step point Together determine the position z of the pedestrian at step k. k Determining yaw requires calculating the energy of each step measured by the left and right optical fibers when calculating gait characteristics. If the energy measured by one optical fiber is greater than that measured by the other optical fiber in three consecutive steps, a yaw warning will be given to the user, and guidance will be provided to the user in the correct walking direction.

[0075] Whether each pedestrian in the predicted signal is blind is specifically determined as follows:

[0076] Step 3.1: Interpolate and denoise the one-dimensional maximum amplitude time-domain signal X0(t) for each pedestrian. In the one-dimensional maximum amplitude time-domain signal, the vibration signal of each step may have small amplitude components that are removed during the denoising process, while large amplitude noise may be ignored and retained. Set the interpolation interval threshold D. threshold Calculate the time interval d between the signal at each sampling moment and the preceding and following non-zero signals. If d ≤ D threshold If the signal is 0 at the current moment, then interpolation is performed, and the magnitude of the interpolated amplitude is the average of the amplitudes of the non-zero signals before and after it; if d ≥ D threshold If the signal is not zero at the current time, then denoising is performed and the amplitude at that point is set to zero; the maximum amplitude time-domain signal obtained after interpolation and denoising is denoted as X(t);

[0077] Step 3.2: Extract gait features from the signal of each pedestrian, extracting a total of 5 sets of features, feature vector [f] i,1 f i,2 f i,3 f i,4 f i,5 The elements are defined as follows:

[0078] (1)f i,1 : Step frequency, which is the estimated number of steps per minute;

[0079] (2)f i,2 Stride length, which is the average distance of a single step;

[0080] (3)f i,3 The average energy of the vibration signal;

[0081] (4)f i,4 The zeroing rate of the time-domain signal X(t) with maximum amplitude;

[0082] (5)f i,5 : Average spatial span of the vibration signal;

[0083] Step 3.3: Combine all features into a feature vector and standardize the feature vector. Each time, randomly select a sample from the training sample set, then find a nearest neighbor samples from the same class sample set, and find a nearest neighbor samples from each different class sample set. Then update the weight of each feature, use the ReliefF algorithm to filter the feature weights, input all features into the CNN network for convolutional classification, and determine whether the pedestrian is a visually impaired user.

[0084] The step frequency is defined as the number of steps per unit time, and the step frequency is calculated as follows:

[0085] 1.1. Record the monotonicity of the short-time cumulative quantity R(m) of the relevant function using the sliding window method; first, set the sliding window length L, the single sliding distance D, the peak time coordinate array T, and the observation window length L. w Calculate the number of slides M, initialize the one-dimensional array flag, initialize the dynamic array G storing the autocorrelation peaks, calculate the signal length n, and calculate the short-time cumulative value of the correlation function. The mean value within each sliding window is compared with the mean value within the previous and next windows. A flag value is set to record the autocorrelation and trend of change.

[0086] 1.2 Initialize the coordinate array within the peak time of each step point, search for the peak coordinates, and the corresponding coordinates are the time position coordinates of the step point; initialize the window number j to 1, and when j < M-6, perform a judgment; if...

[0087] flag(j)+flag(j+1)+flag(j+2+p)=3+p,p≥0

[0088] Then perform the judgment again, if

[0089] flag(j+p+1)+flag(j+p+1+q)=-2-q,q≥0,

[0090] Then store the coordinates of the maximum value of window j+2+p into G;

[0091] 1.3 Calculate the length L of array G G Number of steps N step =L G -1, initialize the peak time coordinate array T of the step points. p L before X(t) W Intra-point search peak coordinate T P (1) For the kth value, it can be obtained through formula T P (k)=TP (k-1)+G(k)-G(k-1) is calculated;

[0092] 1.4 The output step frequency is: f i,1 =F=60×f s ×N step / [T P (L G )-T P (1)].

[0093] The stride is the distance between adjacent step points. Due to the limitations of OTDR equipment, a joint analysis of the signal's spatial center and peak position is used. The specific calculation method is as follows:

[0094] Input the number of steps N, the signal in a single frame, and the peak coordinates. For the k-th step, when k does not exceed L... G When, calculate P k (z), and estimate the position. k exceeds L G When, calculate S1, S2, S; where,

[0095] P k (z) is calculated by the following formula:

[0096]

[0097] In the formula, t range =200;

[0098] Calculated by the following formula:

[0099]

[0100] The formulas for calculating S1, S2, and S are as follows:

[0101] Unit: cm

[0102] Unit: cm

[0103]

[0104] The average energy of the signal can be calculated using the following formula:

[0105]

[0106] The zeroing rate can be calculated using the formula given below:

[0107]

[0108] In the formula, Ⅱ(·) is an indicator function. If · is true, it takes the value 1; otherwise, it takes the value 0.

[0109] Since the noise reduction process sets signal values ​​no higher than the noise threshold to zero, and the interpolation process recovers part of the filtered-out signal, the zeroing rate is essentially the proportion of time occupied by the noise signal. Furthermore, intermittent single pulses are considered noise above the noise threshold.

[0110] The average spatial span can be calculated using the following formula:

[0111] W k =z kmax -z kmin P k (z)>0, z∈[z kmin ,z kmax ]

[0112]

[0113]

[0114] Spatial span reflects the intensity and propagation capability of vibration signals. For an input two-dimensional signal, the length L of the observation window is set. OB Calculate the number of windows N W After calculating the spatial span of the current window, the calculation moves to the next window, and finally the average spatial span of each window is output.

[0115] This invention proposes a system for blind positioning and yaw warning based on fiber optic vibration signals, which is implemented by the aforementioned device for blind positioning and yaw warning based on fiber optic vibration signals.

[0116] This system provides a high recognition rate and high positioning accuracy for blind users without the need for video surveillance equipment or wearable sensors, offering more comprehensive and effective application support for fields such as security and mobility for the blind. The system enhances its adaptability to different human body types, identifying the location of visually impaired users with low latency and high accuracy, and providing timely warnings when visually impaired users deviate from their designated path, paving the way for future services tailored to visually impaired users.

[0117] Symbol Explanation Table

[0118]

[0119]

[0120]

[0121]

Claims

1. A method for blind person localization and yaw warning based on fiber optic vibration signals, characterized in that: The apparatus of the method includes an optical time domain reflectometer module, a signal preprocessing module, a blind / non-blind person discrimination module, and a positioning and yaw warning module; The optical time-domain reflectometer module is used to collect photoelectric signal intensity data generated by walking vibrations; The signal preprocessing module is used to preprocess the extracted two-dimensional intensity signal, including filtering out background noise from the acquired data, removing static frames of vibration signals without targets, calculating the number of pedestrians contained in the input signal, and separating the vibration signals of different pedestrians. The blind / non-blind person discrimination module is used to predict whether each pedestrian in the signal is blind. It performs interpolation and noise reduction on the one-dimensional maximum amplitude time domain signal of each pedestrian, calculates gait features, and inputs the feature vector into the classification algorithm for judgment. The positioning and yaw warning module is designed for blind individuals. It calculates and tracks the walking position of visually impaired pedestrians in the original signal. If the pedestrian's position deviates from the fiber optic tactile paving, it issues an alarm and indicates the correct walking direction to the pedestrian. The method is specifically as follows: Step 1: Start the optical time domain reflectometer and record the target user's walking data. It acquires the generated signal strength data and outputs it in the form of a TXT document; Step 2: Preprocess the obtained two-dimensional signal; preprocessing includes noise reduction, removal of still frames, calculation of pedestrian count, and separation of vibration signals from different pedestrians; during noise reduction, an adaptive noise threshold is searched. And set the threshold to The filter considers signals with amplitudes below a certain threshold as noise and assigns them a value of zero; when removing still frames, an energy threshold needs to be set. The signal is divided into multiple frames in the time dimension, and the energy of each frame is calculated. The system determines whether the current frame is a still frame and checks for "walk-stop-walk" patterns, correcting the disruption to the original step interval caused by discarding still frames. Finally, it calculates the start and end points of each corrected frame, stitches the frames together, and outputs a two-dimensional signal combining multiple frames. When calculating the number of pedestrians, an observation window is set, and the number of symmetrical peak groups within each observation window is calculated. Due to the symmetrical layout of the measuring optical fiber, the number of symmetrical peak groups represents the number of pedestrians. Because the number of pedestrians may change dynamically, we statistically analyze periods when the number of pedestrians is stable and extract the two-dimensional signal for separation during those periods; this allows us to separate different types of pedestrians. When detecting vibration signals, record the points where the amplitude of each group of symmetrical waves decays to zero. ,according to Cutting two-dimensional signals; Step 3: Interpolate and denoise the vibration signal of each pedestrian, calculate gait features, and input the feature vector into the classification algorithm for judgment; if a pedestrian is judged to be a blind user, proceed to step 4; Step 4: Localization and yaw alarm for blind users; first, based on the peak time coordinates of gait points obtained when calculating gait characteristics. Search for the peak position of the step point in the two-dimensional signal. The position estimate is then calculated using a weighted average of the squared signal amplitudes. Combined with the peak position of the step point Jointly determine the pedestrian number Step position ; Determining yaw requires calculating the energy of each step measured by the left and right optical fibers when calculating gait characteristics. If the energy measured by one optical fiber is greater than that measured by the other optical fiber in three consecutive steps, a yaw warning will be given to the user, and guidance will be provided to walk in the correct direction.

2. The method according to claim 1, characterized in that, Whether each pedestrian in the predicted signal is blind is specifically determined as follows: Step 3.1: For the one-dimensional maximum amplitude time-domain signal of each pedestrian... Interpolation and denoising are performed. In the one-dimensional maximum amplitude time domain signal, the vibration signal at each step may have small amplitude parts that are removed during the denoising process, and may have large amplitude noise that is ignored and retained during the denoising process. Set interpolation interval threshold Calculate the time interval between the signal at each sampling point and the preceding and following non-zero signals. ,like If the signal is 0 at the current moment, then interpolation is performed, and the magnitude of the interpolated amplitude is the average of the amplitudes of the non-zero signals before and after it; if If the signal at the current moment is not zero, then denoising is performed, and the amplitude at that sampling point is set to zero; the maximum amplitude time-domain signal obtained after interpolation and denoising is denoted as... ; Step 3.2: Extract gait features from the signal of each pedestrian, extracting a total of 5 sets of features, and feature vectors. The elements are defined as follows: (1) : Step frequency, which is the estimated number of steps per minute; (2) Stride length, which is the average distance of a single step; (3) The average energy of the vibration signal; (4) Maximum amplitude time domain signal The rate of zeroing; (5) : Average spatial span of the vibration signal; Step 3.3: Combine all features into a feature vector and standardize the feature vector; Each time, a sample is randomly selected from the training sample set, and then samples from the same type of sample set are selected. Find the nearest neighbor samples from each different class sample set. The algorithm iterates through the nearest neighbor samples, updates the weight of each feature, uses the ReliefF algorithm to filter the feature weights, and then inputs all features into a CNN network for convolutional classification to determine whether the pedestrian is a visually impaired user.

3. The method according to claim 2, characterized in that, The step frequency is defined as the number of steps per unit time, and the step frequency is calculated as follows: 1.

1. Use the sliding window method to record the short-term cumulative values ​​of related functions. Monotonicity changes; first set the sliding window length. Single sliding distance Step point peak time coordinate array Observation window length ; Calculate the number of slides Initialize a one-dimensional array of flag bits. Initialize the dynamic array that stores the autocorrelation peaks. Calculate the signal length Calculate the short-time cumulant of the correlation function. The mean value within each sliding window is set by comparing the mean values ​​within the previous and next windows. Values ​​are used to record autocorrelation and trends of change; 1.

2. Initialize the coordinate array within the peak time of each step point, search for the peak coordinates, and the corresponding coordinates are the time position coordinates of the step point; initialize the window number. When it is 1, Make a judgment in time, if Then perform the judgment again, if ， Then the window Store the coordinates of the maximum value ; 1.3 Calculate the array length Steps Initialize the peak time coordinate array of the step points ,exist The former Peak coordinates in point search For the A value, which can be obtained through the formula Calculated; 1.4 The output step frequency is: .

4. The method according to claim 2, characterized in that, The stride is the distance between adjacent step points, and its specific calculation method is as follows: Enter the number of steps N Single-frame signal and peak coordinates, for the first Step, when Not exceeding At that time, calculate And estimate the location , Exceeding At that time, calculate ;in, Calculated by the following formula: In the formula, ; Calculated by the following formula: The calculation formula is as follows: 。 5. The method according to claim 2, characterized in that, The average energy of the signal can be calculated using the following formula: 。 6. The method according to claim 2, characterized in that, The zeroing rate can be calculated using the formula given below: In the formula, , It is an indicator function, if , If true, the value is 1; otherwise, the value is 0.

7. The method according to claim 2, characterized in that, The average spatial span can be calculated using the following formula: Spatial span reflects the intensity and propagation capability of vibration signals. For input two-dimensional signals, the length of the observation window is set. Calculate the number of windows After calculating the spatial span of the current window, the calculation moves to the next window, and finally the average spatial span of each window is output.

8. A system for blind positioning and yaw warning based on fiber optic vibration signals, characterized in that: The system is used to implement the method for blind positioning and yaw warning based on fiber optic vibration signals according to any one of claims 1-7.

Images