A gait recognition device, system and method based on optical fiber vibration signals

Abstract

The application provides a gait recognition device, system and method based on an optical fiber vibration signal. The device comprises an optical time domain reflectometer module, a signal preprocessing module, a signal cutting module, a signal feature extraction module and a signal classification module. The method provides a high recognition rate without video monitoring equipment and wearable sensing equipment, more comprehensively and effectively provides application support for security, monitoring and other fields, and recognizes the corresponding user identity of the vibration signal received by the current device at a low time delay and a high accuracy, so as to provide related services for security and monitoring services in the future. In the signal feature extraction algorithm, a long-time and short-time feature joint recognition mode is adopted, and the accuracy is further improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of human recognition, and particularly relates to a gait recognition device, system and method based on an optical fiber vibration signal. BACKGROUND

[0002] With the vigorous development of the Internet and wireless communication technology, users have put forward higher and higher requirements for identity authentication technology, mainly in accuracy, real-time, non-contact and other characteristics. In theory, biometric authentication is the most reliable identity authentication method, because it uses the unique biological characteristics of each person to represent the digital identity of each person and performs corresponding verification, which is almost impossible to be imitated. Common biometric recognition includes fingerprint scanning, face recognition, iris recognition and behavior recognition. Among them, the biometric behavior recognition technology is a method of recognizing a person by walking, typing or handling a certain device. Users using biometric identity verification usually have a smoother and faster experience, because it does not require password recall, and attackers are more difficult to cheat.

[0003] Gait recognition technology, as a new biometric recognition technology, aims to realize automatic identity recognition through the differences in human walking methods. Among them, gait refers to the way people walk and the various features it contains, such as the posture of each joint, step, step force, step speed, etc. Compared with other biometric technologies, gait recognition has the advantages of non-contact, long-distance observation and difficult disguise. The non-contact of gait recognition is reflected in that the whole recognition process does not require the active cooperation of the person; the long-distance observation of gait recognition is reflected in that the input information required for recognition can be obtained, transmitted and analyzed from a distance; the difficult disguise of gait recognition is reflected in that everyone has a different walking posture. As early as the 1960s, Murray's medical research has shown that the gait of each person is different, and it is very difficult to disguise the walking posture. Because there are subtle differences in muscle strength, tendon and bone length, bone density, visual sensitivity, coordination ability, past walking experience, weight, center of gravity, degree of muscle or bone damage, and personal walking habits.

[0004] Gait recognition technology has broad prospects in the fields of intelligent security monitoring, public transportation, smart city construction, etc., and can be used for security control, cardless travel, crowd density monitoring, and over-flow early warning. Obviously, the related research on gait recognition has important significance. In the above background, how to improve the existing gait recognition system or propose a new and more efficient gait recognition device and method has become a problem to be solved and has attracted the attention of many scholars. SUMMARY

[0005] The application aims to solve the problems in the prior art, and provides a gait recognition device, system and method based on an optical fiber vibration signal.

[0006] The application is implemented by the following technical solutions, and provides a recognition method of a gait recognition device based on a distributed optical fiber sensing device.

[0007] The optical time domain reflectometer module is used to collect photoelectric signal intensity data generated by walking vibration signals.

[0008] The signal preprocessing module is used to preprocess the extracted two-dimensional intensity signals, filter out the noise, calculate the number of people recognized in a period, separate the collected different human body signals, and remove the static frames without target user signals.

[0009] The signal cutting module is used to cut the signals after preprocessing into segments, and perform frame processing on the signals with long length.

[0010] The signal feature extraction module is used to extract features from the segmented signal, and output a set of gait feature vectors.

[0011] The signal classification module is used to classify the gait feature vectors extracted from each set of signals, and output the identity label.

[0012] The method specifically comprises the following steps.

[0013] 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.

[0014] Step 2: Use the signal preprocessing module to preprocess the obtained intensity two-dimensional signal, first set a certain threshold X threshold , and consider that the signal with energy lower than the intensity is noise and is directly filtered out. Remove the static frame, that is, first set an energy threshold E threshold , cut the signal into multiple frames, calculate the frame energy E i, judge whether the frame is a still frame; finally check whether there is "walk-stop-walk", for a non-stationary frame j, judge whether its adjacent K frames belong to still frames, if there is a still frame in the adjacent frames of the non-stationary frame, it means that the experimental data has "walk-stop-walk" phenomenon, using the step frequency algorithm of the signal feature extraction module to extract the step interval D of the original non-stationary frame and the step interval Q after the still frame, using the difference between the two, if D-Q interpolation is greater than the threshold value, then fill D-Q zeros between j frame and j+K+1 frame, otherwise fill Q-D zeros between the two frames; correct the damage of the discarded stop frame to the original step interval, finally calculate the start and end points of each frame after correction, splice each frame and output; the two-way sensing signals measured by the left and right optical fibers are separated to obtain double-channel signals, the maximum amplitude value of each channel of the two-dimensional signal of the double-channel signal is taken in unit time, and a one-dimensional signal X(t) of time-maximum amplitude is obtained;

[0015] Step 3: complete signal framing by using signal cutting module; first set the frame number, divide the whole signal length into frame number equal length modules, decide whether to extend the frame head or frame tail by detecting whether there is an incomplete step in the frame, if there is an incomplete step, extend the frame head or frame tail at the corresponding position, complete the output of each frame;

[0016] Step 4: extract features from the signal in each frame, extract the following 11 groups of features, the i-th frame single frame feature vector [f i,1 ,f i,2 ,f i,3 ,f i,4 ,f i,5 ,f i,6 ,f i,7 ,f i,8 ,f i,9 ,f i,10 ,f i,11 ] each element is defined as follows:

[0017] (1)f i,1 : step frequency, i.e. the estimated value of steps per minute;

[0018] (2)f i,2 : step length, i.e. the average value of single step distance;

[0019] (3)f i,3 : average value of vibration signal energy;

[0020] (4)f i,4 : variance mean ratio of maximum amplitude time domain signal X(t);

[0021] (5)f i,5 : peak value coefficient of maximum amplitude time domain signal X(t);

[0022] (6)f i,6 : The rate of over-mean deviation of the time-domain signal X(t) with maximum amplitude;

[0023] (7)f i,7 : The average duration of the time-domain signal X(t) with maximum amplitude;

[0024] (8)f i,8 : Sample entropy of the time-domain signal X(t) with maximum amplitude;

[0025] (9)f i,9 The peak frequency of the spectrum of the time-domain signal X(t) with maximum amplitude;

[0026] (10)f i,10 : Power spectral entropy of the time-domain signal X(t) with maximum amplitude;

[0027] (11)f i,11 : Average spatial span of the vibration signal;

[0028] Step 5: Combine all features into a feature vector, standardize the feature vector, and use an extreme gradient boosting tree as the classifier. XGBoost uses CART as the weak classifier. XGBoost integrates the CART classifier by continuously adding CART and splitting the features.

[0029] Furthermore, the step frequency is defined as the number of steps per unit time, and its calculation method is as follows:

[0030] 1.1 Set the sliding window length L, single sliding distance D, step point peak time coordinate array T, and observation window length L. w Calculate the number of sliding steps K, initialize the one-dimensional array flag, initialize the dynamic array G storing the autocorrelation peaks, and calculate the autocorrelation sum. The mean value within each sliding window is compared with the mean value within the previous and next windows, and a flag value is set to record the autocorrelation and trend of change.

[0031] 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; set the flag j to 1, and when j < K-6, perform a judgment; if...

[0032] flag(j)+flag(j+1)+flag(j+2+p)=3+p,p≥0

[0033] Then perform the judgment again, if

[0034] flag(j+p+1)+flag(j+p+1+q)=-2-q,q≥0,

[0035] Then store the coordinates of the maximum value of window j+2+p into G;

[0036] 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)=T P (k-1)+G[k]-G[k-1] is calculated to obtain;

[0037] 1.4 The output step frequency is: f i,1 =F=60×f s ×N step / [T P (L G )-T P (1)].

[0038] Furthermore, the stride length is the distance between adjacent step points, and the number of step points N is input. step For a single frame signal and peak coordinates, for the i-th step, when i does not exceed the maximum step limit, calculate P. i (z), and estimate the position z. i When i exceeds the maximum number of steps, calculate S1, S2, and S; where,

[0039] P i (z) is calculated by the following formula:

[0040]

[0041] In the formula, t range =200;

[0042] z i Calculated by the following formula:

[0043]

[0044] The formulas for calculating S1, S2, and S are as follows:

[0045]

[0046]

[0047]

[0048] Furthermore, the average energy is calculated using the following formula:

[0049]

[0050] The variance-to-mean ratio is calculated using the following formula:

[0051]

[0052] Furthermore, the peak value coefficient is calculated using the following formula:

[0053]

[0054] in:

[0055] X max =max{X(t)}

[0056]

[0057] The over-mean rate is calculated using the following formula:

[0058]

[0059] Furthermore, the average duration is calculated using the following formula:

[0060]

[0061] In the formula, Ⅱ(·) is an indicator function. If · is true, it takes the value 1; otherwise, it takes the value 0.

[0062] Furthermore, the sample entropy E s The non-stationarity and complexity of the signal are characterized, and the calculation process is as follows:

[0063] (1) Set the reconstruction dimension m = 2; distance threshold

[0064] (2) Reconstruct a vector sequence Y of dimension m in time sequence. m (1),…,Y m (n-m+1);

[0065] Y m (i)={X(i),X(i+1),…,X(i+m-1)}, 1≤i≤n-m+1

[0066] (3) Define vector Y m (i) and Y m The distance between (j) is:

[0067]

[0068] (4) For a given Y m (i), Statistics Ym (i) and Y m B is the number of j whose distance to r is less than or equal to r. i , 1≤j≤nm and j≠i,

[0069]

[0070] (5) Increase the dimension to m+1 and calculate Y. m (i) and Y m The number A of j whose distance to each other is less than or equal to r. i , 1≤j≤nm and j≠i; similarly, define

[0071]

[0072]

[0073] (6) Approximate calculation of the sample entropy:

[0074]

[0075] Furthermore, the peak frequency of the spectrum is calculated using the following formula:

[0076]

[0077] Where f s The sampling frequency of the OTDR.

[0078] The power spectral entropy is calculated using the following formula:

[0079] SPD(f) = |FFT(R) x [X(t)])

[0080]

[0081]

[0082] Furthermore, the spatial span reflects the intensity and propagation capability of the vibration signal. For the input two-dimensional signal, the length L of the observation window is set. OB The number of windows, M, is calculated. After the spatial span of the current window is calculated, the window completes its displacement. The spatial span within the window is defined as:

[0083] W k =z kmax -z kmin P k (z)>0, z∈[z kmin ,z kmax ]

[0084]

[0085]

[0086] The beneficial effects of this invention are as follows:

[0087] This invention provides a gait recognition device, system, and method based on fiber optic vibration signals. This method offers a high recognition rate without requiring video surveillance equipment or wearable sensing devices, providing more comprehensive and effective application support for security, monitoring, and other fields. With low latency and high accuracy, it identifies the user identity corresponding to the vibration signals received by the device, enabling future services for security and monitoring operations. The signal feature extraction algorithm employs a joint recognition approach combining long-term and short-term features, further improving accuracy. Attached Figure Description

[0088] Figure 1 This is a schematic diagram of the process of the method described in this invention.

[0089] Figure 2 This is a schematic diagram of the two-dimensional intensity-time signal acquired by the optical time-domain reflectometer in the method described in this invention.

[0090] Figure 3 This is a schematic diagram of a one-dimensional intensity-time signal obtained by the preprocessing module in the method described in this invention.

[0091] Figure 4 This is a schematic diagram of the channel splitting of a two-dimensional signal in the method described in this invention.

[0092] Figure 5 This is a physical image of the optical time domain reflectometer used in this invention. Detailed Implementation

[0093] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0094] This invention proposes a gait recognition method based on a distributed optical fiber sensing device. The device includes an optical time domain reflectometer (OTDR) module, a signal preprocessing module, a signal segmentation module, a signal feature extraction module, and a signal classification module.

[0095] The optical time-domain reflectometer module is used to collect photoelectric signal intensity data generated by walking vibration signals;

[0096] The signal preprocessing module is used to preprocess the extracted two-dimensional intensity signal, process the collected data, filter out background noise, calculate the number of people identified in a certain period of time, separate the collected different human body signals, and remove still frames without target user signals.

[0097] The signal segmentation module is used to segment the preprocessed signal, divide long signals into frames, and separate the two signals within each frame.

[0098] The signal feature extraction module is used to extract features from the segmented signal fragments and output a set of gait feature vectors;

[0099] The signal classification module is used to classify the gait feature vectors extracted from each group of signals and output their identity labels (i.e., the predicted value of the user's identity).

[0100] The method is specifically as follows:

[0101] 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 it 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.

[0102] Step 2: Preprocess the obtained two-dimensional intensity signal using the signal preprocessing module. First, set a certain threshold X. threshold The filter considers signals with energy below a certain intensity as noise and filters them out directly; still frames are removed, i.e., an energy threshold E is first set. threshold The signal is divided into multiple frames, and the frame energy E is calculated for each frame. iThe process involves summing the squares of all signals within a frame to determine if it is a stationary frame. If the energy within the frame exceeds a threshold, it is considered a non-stationary frame; otherwise, it is considered a stationary frame. Finally, it checks for "walk-stop-walk" patterns. For a non-stationary frame j, it checks if its K adjacent frames are stationary frames. If a stationary frame appears in the adjacent frames of the non-stationary frame, it indicates that the experimental data exhibits a "walk-stop-walk" phenomenon. The step frequency algorithm of the signal feature extraction module is used to extract the step interval D of the original non-stationary frame and the step interval Q after the stationary frame. The difference between the two is then used... If the DQ interpolation is greater than the threshold, pad with DQ zeros between frame j and frame j+K+1; otherwise, pad with QD zeros between these two frames. 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, splice the frames and output them. Separate the two sensing signals simultaneously measured by the left and right optical fibers of the device to obtain dual-channel signals. Obtain the two positions of the maximum amplitude per unit time and record them as arrays of the two channels. Take the maximum amplitude value of the two-dimensional signal of each channel of the dual-channel signal per unit time to obtain the one-dimensional signal X(t) of time-maximum amplitude.

[0103] Step 3: Use the signal cutting module to complete the signal framing; First, set the number of frames and divide the entire signal length into the number of equal-length modules. By detecting whether there is an incomplete step in the frame, determine whether it is necessary to extend the frame header or frame tail. If there is an incomplete step, extend the frame header or frame tail at the corresponding position to complete the output of each frame.

[0104] Step 4: Extract features from the signal within each frame, extracting a total of 11 sets of features, including the single-frame feature vector [f] of the i-th frame. i,1 ,f i,2 ,f i,3 ,f i,4 ,f i,5 ,f i,6 ,f i,7 ,f i,8 ,f i,9 ,f i,10 ,f i,11 The elements are defined as follows:

[0105] (1)f i,1 : Step frequency, which is the estimated number of steps per minute;

[0106] (2)f i,2 Stride length, which is the average distance of a single step;

[0107] (3)f i,3 : The average energy of the vibration signal;

[0108] (4)f i,4 : The ratio of the variance to the mean of the time-domain signal X(t) with maximum amplitude;

[0109] (5)f i,5 Peak value of the time-domain signal X(t) with maximum amplitude;

[0110] (6)f i,6 : The rate of over-mean deviation of the time-domain signal X(t) with maximum amplitude;

[0111] (7)f i,7 : The average duration of the time-domain signal X(t) with maximum amplitude;

[0112] (8)f i,8 : Sample entropy of the time-domain signal X(t) with maximum amplitude;

[0113] (9)f i,9 The peak frequency of the spectrum of the time-domain signal X(t) with maximum amplitude;

[0114] (10)f i,10 : Power spectral entropy of the time-domain signal X(t) with maximum amplitude;

[0115] (11)f i,11 : Average spatial span of the vibration signal;

[0116] Step 5: Combine all features into a feature vector and standardize it to accelerate the convergence of the classification algorithm. Use Extreme Gradient Boosting Tree (XGBoost) as the classifier. Boosting is a family of algorithms that enhance weak learners into strong learners. XGBoost uses CART (Classification And Regression Tree) as the weak classifier. CART is a binary tree that continuously splits the features of the sample set from top to bottom. XGBoost integrates the CART classifier by continuously adding CART classes and splitting the features.

[0117] The step frequency is defined as the number of steps per unit time, and its calculation method is as follows:

[0118] 1.1 Set the sliding window length L, single sliding distance D, step point peak time coordinate array T, and observation window length L. w Calculate the number of sliding steps K, initialize the one-dimensional array flag, initialize the dynamic array G storing the autocorrelation peaks, and calculate the autocorrelation sum. The mean value within each sliding window is compared with the mean value within the previous and next windows, and a flag value is set to record the autocorrelation and trend of change.

[0119] 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; set the flag j to 1, and when j < K-6, perform a judgment; if...

[0120] flag(j)+flag(j+1)+flag(j+2+p)=3+p,p≥0

[0121] Then perform the judgment again, if

[0122] flag(j+p+1)+flag(j+p+1+q)=-2-q,q≥0,

[0123] Then store the coordinates of the maximum value of window j+2+p into G;

[0124] 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)=T P (k-1)+G[k]-G[k-1] is calculated to obtain;

[0125] 1.4 The output step frequency is: f i,1 =F=60×f s ×N step / [T P (L G )-T P (1)].

[0126] The stride is the distance between adjacent step points. Due to the limitations of OTDR devices, a joint analysis of the signal's spatial dimension and peak position is used, with the number of input step points N. step For a single frame signal and peak coordinates, for the i-th step, when i does not exceed the maximum step limit, calculate P. i (z), and estimate the position z. i When i exceeds the maximum number of steps, calculate S1, S2, and S; where,

[0127] P i (z) is calculated by the following formula:

[0128]

[0129] In the formula, t range =200;

[0130] z iCalculated by the following formula:

[0131]

[0132] The formulas for calculating S1, S2, and S are as follows:

[0133]

[0134]

[0135]

[0136] The average energy value is calculated using the following formula:

[0137]

[0138] The variance-to-mean ratio is calculated using the following formula:

[0139]

[0140] The peak value coefficient is calculated using the following formula:

[0141]

[0142] in:

[0143] X max =max{X(t)}

[0144]

[0145] The over-mean rate is calculated using the following formula:

[0146]

[0147] The average duration is calculated using the following formula:

[0148]

[0149] In the formula, Ⅱ(·) is an indicator function; it takes the value 1 if · is true, and 0 otherwise. Since the noise reduction process sets signal values ​​not higher than the noise threshold to zero, the zeroing rate is essentially the proportion of time occupied by the noise signal. In addition, intermittent single pulses are considered noise higher than the noise threshold.

[0150] The sample entropy E s The non-stationarity and complexity of the signal are characterized, and the calculation process is as follows:

[0151] (1) Set the reconstruction dimension m = 2; distance threshold

[0152] (2) Reconstruct a vector sequence Y of dimension m in time sequence. m (1),…,Y m (n-m+1);

[0153] Y m (i)={X(i),X(i+1),…,X(i+m-1)}, 1≤i≤n-m+1

[0154] (3) Define vector Y m (i) and Y m The distance between (j) is:

[0155]

[0156] (4) For a given Y m (i), Statistics Y m (i) and Y m B is the number of j whose distance to r is less than or equal to r. i , 1≤j≤nm and j≠i,

[0157]

[0158] (5) Increase the dimension to m+1 and calculate Y. m (i) and Y m The number A of j whose distance to each other is less than or equal to r. i , 1≤j≤nm and j≠i; similarly, define

[0159]

[0160]

[0161] (6) Approximate calculation of the sample entropy:

[0162]

[0163] The peak frequency of the spectrum is calculated using the following formula:

[0164]

[0165] Where f s The sampling frequency of the OTDR.

[0166] The power spectral entropy is calculated using the following formula:

[0167] SPD(f) = |FFT(R) x [X(t)])

[0168]

[0169]

[0170] 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 The number of windows, M, is calculated. After the spatial span of the current window is calculated, the window completes its displacement. The spatial span within the window is defined as:

[0171] W k =z kmax -z kmin P k (z)>0, z∈[z kmin ,z kmax ]

[0172]

[0173]

[0174] This invention provides a gait recognition device, system, and method based on fiber optic vibration signals. The method offers a high recognition rate without requiring video surveillance equipment or wearable sensing devices, providing more comprehensive and effective application support for security, monitoring, and other fields. It identifies the source user of human walking signals received by the device with low latency and high accuracy, enabling future services for security and monitoring. The signal feature extraction algorithm employs a joint recognition approach combining long-term and short-term features, further improving accuracy.

[0175] Table 1. Symbol Explanation Table

[0176]

[0177]

[0178]

Claims

1. A recognition method of a gait recognition device based on a distributed optical fiber sensing device, characterized by: The device comprises an optical time domain reflectometer module, a signal preprocessing module, a signal cutting module, a signal feature extraction module and a signal classification module. The optical time domain reflectometer module is used to collect photoelectric signal intensity data generated by the walking vibration signal. The signal preprocessing module is used to preprocess the extracted two-dimensional intensity signal, filter out the noise, calculate the number of people identified in a period, separate the collected different human body signals, and remove the static frames without target user signals. The signal cutting module is used for segment cutting of the preprocessed signal, frame processing of the long length signal, and segmentation of the two-way signal in each frame. The signal feature extraction module is used to extract features from the segmented signal and output a set of gait feature vectors. The signal classification module is used to classify the gait feature vectors extracted from each group of signals and output their identity labels. The method specifically comprises: Step 1 : Start the optical time domain reflectometer and record data of the target user walking Obtain the generated signal strength data and output in the form of a TXT document; Step 2: The obtained intensity two-dimensional signal is preprocessed by a signal preprocessing module, and a certain threshold is first set The filter is considered as noise and is directly filtered out if the signal energy is lower than the intensity. Remove still frames, i.e., first set the energy threshold. The signal is divided into multiple frames, and the frame energy is calculated for each frame. First, determine if the frame is a still frame; finally, check if there is a "walk-stop-walk" pattern. For a non-still frame... Determine its adjacent If a non-stationary frame is identified as a still frame, and a still frame appears in the adjacent frame, it indicates that the experimental data exhibits a "walk-stop-walk" phenomenon. The step frequency algorithm of the signal feature extraction module is then used to extract the step interval of walking within the original non-stationary frames. and the step interval after the still frame. Using the difference between the two, if D-Q If the difference is greater than the threshold, then... Frame and Inter-frame padding If there are zeros, padding is used between these two frames; otherwise, padding is used. 0; 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, splice the frames and output; The two-way sensing signals measured simultaneously by the left and right optical fibers of the device are separated to obtain double-channel signals, and the maximum amplitude value of the two-dimensional signals of each channel of the double-channel signals in a unit time is taken to obtain a one-dimensional signal of time-maximum amplitude ; Step 3: The signal cutting module is used to complete signal frame cutting. First, set the frame number, divide the entire signal length into frame number equal length modules, and determine whether the frame head or frame tail needs to be extended by detecting whether there is an incomplete step in the frame. If there is an incomplete step, the frame head or frame tail is extended at the corresponding position to complete the output of each frame. Step 4: Feature extraction is performed on each intra-frame signal, extracting a total of 11 groups of features, the first 8 groups of which are as follows: i Frame Frame Feature Vector Each element is defined as follows: (1) : step frequency, i.e. the estimated number of steps per minute; (2) : stride, i.e. the average of the single step distances; (3) : average value of energy of vibration signal; (4) : maximum amplitude time domain signal : variance mean ratio (5) : peak coefficient of the maximum amplitude time domain signal ; (6) : maximum amplitude time domain signal : over mean rate; (7) : maximum amplitude time-domain signal average duration of time; (8) : maximum amplitude time-domain signal sample entropy of (9) : maximum amplitude time-domain signal : spectral peak frequency of the maximum amplitude time-domain signal (10) : power spectrum entropy of the maximum amplitude time domain signal ; (11) : average value of spatial span of vibration signal; Step 5: Combine all features as a feature vector, normalize the feature vector, use an extremely gradient boosting tree as a classifier, and use CART as the weak classifier of XGBoost. XGBoost continuously adds CART to continuously split the features to complete the integration of the CART classifier.

2. The method of claim 1, wherein, The step frequency is defined as the number of steps per unit time, and the calculation method is specifically as follows: 1.1, Set the length of sliding window , Single sliding distance , Peak time coordinate array of step point , Set the length of observation window , Calculate the number of sliding , Initialize the flag bit one-dimensional array , Initialize the dynamic array to store autocorrelation peaks , Calculate autocorrelation sum , Mean value in each sliding window, set the value by comparing the size of the mean value in the previous and next window , Record the autocorrelation sum trend 1.2, initialize the coordinate array in the peak time of steps, search for the peak coordinate, and the corresponding coordinate is the time position coordinate corresponding to the step point; Setting a flag bit is 1, a determination is made whether if Then, the judgment is performed again. If ， The maximum coordinates of the window are stored in ; 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 k A value, which can be obtained through the formula Calculated; 1.4, the output step frequency is: .

3. The method of claim 2, wherein, The step is the distance between adjacent steps, the number of input steps , single frame signal and peak coordinate, for the first i step, when i the maximum step limit is not exceeded, calculate , and estimate the position , i when the maximum step is exceeded, calculate ; wherein, Calculated from the formula: In the formulae, ; Calculated from the formula: The calculation formula is as follows: 。 4. The method of claim 3, wherein, The energy mean value is calculated by the following formula: The variance mean ratio is calculated by the following formula: 。 5. The method of claim 4, wherein, The peak value coefficient is calculated by the following formula: Wherein: The over mean rate is calculated by the following formula: 。 6. The method of claim 5, wherein, The average duration is calculated by the following formula: wherein is an indicator function that takes the value 1 if is true and 0 otherwise.

7. The method of claim 6, wherein, The sample entropy The non-stationarity and complexity of the signal are characterized, and the calculation process is as follows: (1) setting a reconstruction dimension ; a distance threshold ; (2) reconstructing a sequence of vectors with dimensionality in time order ; (3) defining a vector the distance between is: (4) For a given , the number of j's for which the distance between and is less than or equal to r , and , (5) increase dimension to , compute distance between and , ; Similarly, define (6) The value of the approximate calculation sample entropy is: 。 8. The method of claim 7, wherein, The spectral peak frequency is calculated by the following formula: wherein is the sampling frequency of the OTDR, ; The power spectrum entropy is calculated by the following formula: 。 9. The method of claim 8, wherein, The space span reflects the strength and propagation ability of the vibration signal. For the input two-dimensional signal, the length of the observation window is set , the number of calculation windows M is calculated, and the window displacement is completed after the space span of the current window is calculated. The space span in the window is defined as: 。

Images