Indoor pedestrian navigation method and system based on behavior probability analysis

Abstract

The application discloses an indoor pedestrian navigation method and system based on behavior probability analysis, wherein the method steps comprise the following steps: based on the measurement result of an IMU sensor, a Kalman filter is used to predict the current state of an indoor pedestrian, and a prediction result is obtained; a gait detection method is used to extract a maximum probability zero speed point of the indoor pedestrian; a stair step probability detection method is used to estimate the height of the pedestrian climbing stairs, and an estimation result is obtained; and the prediction result is corrected by using the maximum probability zero speed point and the estimation result, so that indoor pedestrian navigation is completed. The application extracts the zero speed point based on the maximum probability, is suitable for different gaits, and reduces the influence of unstable barometer output on height correction. Through experimental verification, the application can realize high positioning accuracy in an indoor scene.

Description

Technical Field

[0001] This application relates to the field of indoor navigation, specifically to an indoor pedestrian navigation method and system based on behavioral probability analysis. Background Technology

[0002] With the advancement of technology, people's demand for location services is becoming increasingly prominent, which places higher demands on the range and service quality of navigation and positioning. However, satellite-based navigation and positioning systems can only provide high-precision, long-endurance, and stable navigation services in open outdoor environments, and may fail in complex indoor environments.

[0003] For indoor environments, current solutions for pedestrian navigation mainly include two types: wireless positioning technology and inertial navigation technology. Wireless positioning technology involves deploying wireless communication technologies (Wi-Fi, Bluetooth, UWB, etc.) to measure environmental information in indoor settings, thereby achieving pedestrian navigation and positioning. In contrast, inertial navigation technology requires no additional information assistance and no prior deployment, making it a completely autonomous navigation and positioning solution. With the rapid development of Micro-Electro-Mechanical Systems (MEMS), pedestrian inertial navigation systems primarily utilize MEMS inertial sensors, which offer advantages such as small size, low power consumption, and low cost.

[0004] However, MEMS devices have low precision and introduce significant noise during inertial data acquisition. During inertial navigation, system errors accumulate over time, leading to decreased positioning accuracy. Since pedestrians' feet move periodically, there is a zero-velocity interval when the foot is fully in contact with the ground within each gait cycle. To suppress the divergence of navigation errors during long-term continuous integration in pedestrian navigation systems, a zero-velocity correction method is commonly used to correct navigation errors within this zero-velocity interval. Pedestrians in indoor environments typically engage in various motion states, including walking, running, and climbing stairs. However, due to the different dynamic characteristics of these motion states, a fixed zero-velocity threshold results in a high misjudgment rate. Therefore, selecting appropriate zero-velocity thresholds for various motion states and extracting accurate zero-velocity intervals for error correction becomes crucial to navigation and positioning accuracy. Summary of the Invention

[0005] To address the technical problems mentioned above, this application employs a zero-velocity interval detector based on the generalized likelihood ratio test to detect zero-velocity intervals. Gait intervals for each step are obtained through interval constraints and periodic division. The moment with the minimum zero-velocity detection value within each gait interval is extracted as the maximum probability zero-velocity point for each step and used for navigation error correction. Simultaneously, to address the divergence of inertial navigation height error during stair climbing, an online stair climbing height estimation method based on inertial / barometric sensors is proposed. A stair climbing energy characteristic function is constructed within the gait cycle, and then the number of steps and step height are estimated.

[0006] To achieve the above objectives, this application provides an indoor pedestrian navigation method based on behavioral probability analysis, the steps of which include:

[0007] Based on the measurement results of the IMU sensor, Kalman filtering is used to predict the current state of pedestrians indoors, and the prediction result is obtained.

[0008] Using gait detection methods, the highest probability zero-velocity point of indoor pedestrians is extracted;

[0009] The probability detection method for the number of stair steps is used to estimate the height a pedestrian climbs, and the estimation results are obtained.

[0010] Using the maximum probability zero velocity point and the estimation result, the prediction result is corrected to complete indoor pedestrian navigation.

[0011] Preferably, the method for obtaining the prediction result includes: estimating the 9-dimensional state error of the measurement result using a Kalman filter algorithm, wherein the state variables include:

[0012] X k =[P k V k ψ k ] T

[0013] In the formula, X k P represents the IMU sensor state at time k; k V k and ψ k These represent the three-axis position, velocity, and attitude angles of the IMU sensor, respectively; [·] T Represents the transpose matrix;

[0014] Take systematic measurements to observe speed and altitude:

[0015] Z = [Z v Z p ] T

[0016] In the formula, Z vZ represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor. p This represents the difference between the theoretical and actual altitude values.

[0017] According to the state-space model of Kalman filtering, the state variable X and the quantity Z satisfy the following relationship:

[0018]

[0019] In the formula, W represents the system noise matrix; R represents the measurement noise matrix;

[0020] F is the system matrix:

[0021]

[0022] In the formula, S t The skew-symmetric matrix representing the projection of the accelerometer in the navigation frame; O i×j I represents the zero matrix of i×j; i×j Represents the identity matrix of i×j;

[0023] H is the measurement matrix:

[0024]

[0025] In the formula, h 21 =[0 0 1].

[0026] Preferably, the method for performing the gait detection includes:

[0027] A method based on generalized likelihood ratio detection was used to perform zero-speed detection of pedestrians indoors, and the detection results were obtained.

[0028] The detection is divided based on the interval analysis method to obtain the gait interval of each step of the indoor pedestrian;

[0029] The maximum probability zero velocity point is searched within the gait interval of each step.

[0030] Preferably, the method for obtaining the estimation result includes:

[0031] At the moment of zero velocity with the highest probability for each gait of an indoor pedestrian going up and down stairs, extract the inertial calculated height data and barometric altimeter data within that gait interval;

[0032] Based on the altitude data and the barometric altimeter data, the window size and step size are set, and the gait data is traversed using a sliding window to obtain the estimation result.

[0033] Preferably, the method for correcting the prediction result using the maximum probability zero velocity point includes:

[0034] Error correction is performed at the maximum probability zero-velocity point for each gait of an indoor pedestrian; at this point, the theoretical velocity value at time Z is zero. v This represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor, including:

[0035] Z v =[-V x -V y -V z ] T

[0036] In the formula, V x V y V z These represent the three axial components of velocity; [·] T This represents the transpose of the matrix.

[0037] Preferably, the method for correcting the prediction result using the estimation result includes: a height correction scheme based on indoor pedestrian gait type constraints, determining the current gait movement type, and a height correction value Z. p satisfy:

[0038]

[0039] In the formula, h i，k h represents the solution height of the zero-velocity system at time k with the highest probability; p，k h represents the barometric altimeter altitude at the point of maximum probability zero velocity at time k; s n represents the average height of the steps; k This indicates the number of steps traversed during the current uphill / downhill gait; s k Indicators of current gait (going up or down stairs):

[0040]

[0041] This application also provides an indoor pedestrian navigation system based on behavioral probability analysis, including: a prediction module, an extraction module, an estimation module, and a correction module;

[0042] The prediction module is used to predict the current state of indoor pedestrians based on the measurement results of the IMU sensor and using Kalman filtering to obtain the prediction result;

[0043] The extraction module is used to extract the highest probability zero-velocity point of indoor pedestrians using gait detection methods.

[0044] The estimation module is used to estimate the height of a pedestrian climbing stairs using a staircase step probability detection method, and obtain the estimation result;

[0045] The correction module uses the maximum probability zero velocity point and the estimation result to correct the prediction result, thereby completing indoor pedestrian navigation.

[0046] Preferably, the workflow of the prediction module includes: estimating the 9-dimensional state error of the measurement results using a Kalman filter algorithm, wherein the state variables include:

[0047] X k =[P k V k ψ k ] T

[0048] In the formula, X k P represents the IMU sensor state at time k; k V k and ψ k These represent the three-axis position, velocity, and attitude angles of the IMU sensor, respectively; [·] T Represents the transpose matrix;

[0049] Take systematic measurements to observe speed and altitude:

[0050] Z = [Z v Z p ] T

[0051] In the formula, Z v Z represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor. p This represents the difference between the theoretical and actual altitude values.

[0052] According to the state-space model of Kalman filtering, the state variable X and the quantity Z satisfy the following relationship:

[0053]

[0054] In the formula, W represents the system noise matrix; R represents the measurement noise matrix;

[0055] F is the system matrix:

[0056]

[0057] In the formula, S t The skew-symmetric matrix representing the projection of the accelerometer in the navigation frame; O i×j I represents the zero matrix of i×j; i×j Represents the identity matrix of i×j;

[0058] H is the measurement matrix:

[0059]

[0060] In the formula, h 21 =[0 0 1].

[0061] Compared with the prior art, the beneficial effects of this application are as follows:

[0062] This application extracts the zero-velocity point with maximum probability, exhibiting strong adaptability to different time states and reducing the impact of unstable barometer output on altitude correction. Experimental verification demonstrates that this application can achieve high positioning accuracy in indoor scenarios. Attached Figure Description

[0063] To more clearly illustrate the technical solutions of this application, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 This is a schematic diagram of the method flow of an embodiment of this application;

[0065] Figure 2 This is a schematic diagram of the staircase detection algorithm according to an embodiment of this application;

[0066] Figure 3 This is a schematic diagram of the system structure according to an embodiment of this application. Detailed Implementation

[0067] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0068] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0069] Example 1

[0070] like Figure 1 The diagram shown is a schematic representation of the method flow in this embodiment, and the steps include:

[0071] S1. Based on the measurement results of the IMU sensor, Kalman filtering is used to predict the current state of pedestrians indoors, and the prediction results are obtained.

[0072] This embodiment uses a ForSense IMU614 nine-axis MEMS-IMU sensor, which is fixed to the heel of the right insole of the experimenter's foot to obtain measurement results. Based on the prediction results, a Kalman filter is used to predict the current state of the pedestrian indoors, obtaining the prediction result. Specific steps include:

[0073] The 9-dimensional state error of the measurement results is estimated using the Kalman filter algorithm, where the state variables include:

[0074] X k =[P k V k ψ k ] T

[0075] In the formula, X k P represents the IMU sensor state at time k; k V k and ψ k These represent the three-axis position, velocity, and attitude angles of the IMU sensor, respectively; [·] T Represents the transpose matrix;

[0076] Take systematic measurements to observe speed and altitude:

[0077] Z = [Z v Z p ] T

[0078] In the formula, Z v Z represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor. p This represents the difference between the theoretical and actual altitude values.

[0079] According to the state-space model of Kalman filtering, the state variable X and the quantity Z satisfy the following relationship:

[0080]

[0081] In the formula, W represents the system noise matrix; R represents the measurement noise matrix;

[0082] F is the system matrix:

[0083]

[0084] In the formula, S t The skew-symmetric matrix representing the projection of the accelerometer in the navigation frame; O i×j I represents the zero matrix of i×j; i×j Represents the identity matrix of i×j;

[0085] H is the measurement matrix:

[0086]

[0087] In the formula, h 21 =[0 0 1].

[0088] S2. Using gait detection methods, extract the highest probability zero-velocity point of indoor pedestrians.

[0089] Zero-velocity correction, as a primary error correction method for inertial pedestrian navigation, typically involves detecting the magnitude and variance of acceleration and angular velocity within a window and comparing them to a set threshold. To improve the robustness of the zero-velocity correction algorithm, this embodiment proposes a maximum probability zero-velocity point extraction method based on gait detection. The steps include: performing zero-velocity detection using a generalized likelihood ratio-based method to obtain the detection results; dividing the detection area using an interval analysis method to obtain the gait interval for each step; and searching for the maximum probability zero-velocity point within each step's gait interval.

[0090] The steps for the generalized likelihood ratio detection described above include:

[0091] Taking into account both acceleration and angular velocity information, the detection value within the window at time k is calculated using the formula. The detection value is then compared with a threshold to determine whether the current sample is in the zero-velocity interval.

[0092]

[0093] In the formula, T k Represents the zero-velocity detection value at time k; W represents the window size; ω represents the noise variance of the gyroscope and accelerometer, respectively; k f k represents the output values ​​of the gyroscope and accelerometer at time k, respectively; g represents the local gravitational acceleration; This represents the average acceleration across the three axes within the window. When the detected value is less than the zero-velocity detector threshold, the sample data at that time can be considered to be in the zero-velocity interval, and the zero-velocity judgment value is true, as shown in the formula. When pedestrian motion involves multiple motion states, an appropriate detector threshold should be set to detect the possible intervals where zero-velocity points may exist in each state. In order to extract the intervals where zero-velocity points exist in each motion state, considering the large dynamic range of running, the zero-velocity detector threshold T should be set relatively large here.

[0094]

[0095] In the formula, z represents the judgment value of zero-speed detection.

[0096] To reduce the error introduced by inaccurate measurement information, this embodiment performs a systematic error correction only once at the maximum probability zero-velocity point within each gait cycle. Based on generalized likelihood ratio detection, zero-velocity intervals are divided, and interval analysis-based methods are used to constrain these intervals, thus completing the gait cycle division and enabling the extraction of the maximum probability zero-velocity point within each gait cycle. Specific steps include:

[0097] The sample analysis comprehensively considers walking, running, and climbing stairs, firstly constraining the gait interval length. Based on the gait segmentation method based on interval analysis, the length of the non-zero speed interval in each gait is considered to be no less than 0.12 seconds, and the interval length of a gait is no less than 0.6 seconds. The non-zero speed interval length is calculated; if the length is less than 0.12 seconds, a zero-speed point is considered misjudged, and the subsequent interval is designated as a zero-speed interval.

[0098] Based on the periodic pattern of foot movement during pedestrian movement, the zero-speed interval and the non-zero-speed interval alternate, which can be used to divide the complete gait interval. Taking the beginning time of each zero-speed interval as the starting point and the end time of the next non-zero-speed interval as the ending point, it is regarded as a complete gait interval.

[0099] Finally, the identification of the zero-velocity interval was achieved based on the generalized likelihood ratio test. Within a gait cycle, the point with the highest probability of zero foot velocity is near the minimum value detected by the zero-velocity detector. Therefore, within the zero-velocity interval of each gait, the point with the highest probability of zero velocity can be further extracted. This point can be considered as the moment when the heel is fully in contact with the ground and begins to leave the ground in the swing phase at the next moment. Based on the above steps, the detection value T... k The smaller the value, the greater the probability that the velocity is zero at time k. Therefore, the point with the highest probability of zero velocity can be extracted by finding the minimum detection value within the obtained zero velocity interval.

[0100] S3. Using the staircase step probability detection method, the height of pedestrians climbing stairs is estimated, and the estimation results are obtained.

[0101] Currently, the main method for height correction in indoor navigation environments for pedestrians is to use barometric altimeters as height reference information. However, barometric altimeters are easily affected by environmental factors such as temperature and humidity. In addition, as the amplitude of pedestrian movement changes, the barometric altimeter will be subjected to varying degrees of vibration, resulting in significant fluctuations in the output value of the barometric altimeter when the pedestrian is moving.

[0102] To address this issue, this embodiment does not refer to barometric altimeter information when a pedestrian is moving on the same floor; instead, it uses the height of the previous step as a reference for height correction. Since the height of steps within a building is usually consistent, this embodiment assumes that the height of steps within the building remains constant when estimating the height of pedestrians going up and down stairs. To determine the number of steps crossed by a pedestrian's gait, this embodiment combines inertial calculation height information and barometric altimeter information for analysis. The algorithm flowchart is shown below. Figure 2 As shown. At the maximum probability zero-velocity point of each stair-climbing gait, inertial calculated altitude data and barometric altimeter data within that gait interval are extracted. An appropriate window size and step length are set, and a sliding window is used to traverse the gait data.

[0103] Differential feature values ​​are extracted from the inertial altitude and barometric altitude data within each sliding window, and then the energy functions of the inertial altitude difference and barometric altitude difference are calculated. The energy function calculation method includes:

[0104]

[0105] In the formula, z k D represents the maximum probability zero velocity point of the k-th gait; i This represents the difference value. The two energy functions above reflect the degree of change in inertial solution altitude and barometric altitude within the current data window, respectively.

[0106] Secondly, the energy functions are judged based on preset thresholds. If the energy functions of both the inertial solution altitude difference and the barometric altitude difference are below the preset thresholds, it is determined that the current gait segment has not crossed a step. Conversely, if either energy value exceeds the preset threshold, the current gait segment has crossed a step.

[0107] Finally, the number of steps crossed in all windows of the current gait is statistically analyzed to obtain the number of steps crossed in the pedestrian's gait when going up and down stairs.

[0108] To minimize the impact of inaccurate barometric altimeter output on altitude, this embodiment, based on motion state recognition results, only references the barometric altimeter's altitude information and calculates the average step height during the first segment of up and down stairs. Based on the assumption of a constant step height, subsequent up and down stairs are corrected by referencing the average step height information.

[0109]

[0110] In the formula, h s Δh represents the average height of the steps. p1 This represents the height difference measured by the barometric altimeter during the first segment of the up-and-down movement; n irepresents the number of steps crossed in the i-th step; p and q represent the starting and ending gaits of the first segment of going up and down the stairs, respectively.

[0111] Subsequent changes in height during gait while going up and down stairs include:

[0112] Δh k =h s ·n k

[0113] In summary, this embodiment proposes a gait prediction method based on energy function to identify the number of steps crossed by the gait when going up and down stairs and obtain the estimation result.

[0114] S4. Using the maximum probability zero velocity point and the estimation results, the prediction results are corrected to complete indoor pedestrian navigation.

[0115] First, velocity correction is performed using the maximum probability zero-velocity point. In this embodiment, error correction is performed at the maximum probability zero-velocity point for each gait; at this time, the theoretical velocity value is zero, Z. v This represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor, including:

[0116] Z v =[-V x -V y -V z ] T

[0117] In the formula, V x V y V z These represent the three axial components of velocity; [·] T This represents the transpose of the matrix.

[0118] Finally, the height is corrected using the estimation results. In this embodiment, a height correction scheme based on gait type constraints is used to determine the current gait movement type, and the height correction value Z is determined. p satisfy:

[0119]

[0120] In the formula, h i，k h represents the solution height of the zero-velocity system at time k with the highest probability; p，k h represents the barometric altimeter altitude at the point of maximum probability zero velocity at time k; s n represents the average height of the steps; k This indicates the number of steps traversed during the current uphill / downhill gait; s k Indicators of current gait (going up or down stairs):

[0121]

[0122] Thus, indoor pedestrian navigation is completed through the above steps.

[0123] Example 2

[0124] like Figure 3 The diagram shown illustrates the system structure of this embodiment, including a prediction module, an extraction module, an estimation module, and a correction module. The prediction module uses Kalman filtering based on IMU sensor measurements to predict the current state of an indoor pedestrian, obtaining a prediction result. The extraction module uses gait detection to extract the maximum probability zero-velocity point of the indoor pedestrian. The estimation module uses a staircase step probability detection method to estimate the pedestrian's climbing height, obtaining an estimation result. The correction module uses the maximum probability zero-velocity point and the estimation result to correct the prediction result, completing indoor pedestrian navigation.

[0125] The following will, in conjunction with this embodiment, explain in detail how this application solves technical problems in real life.

[0126] First, using the prediction module, based on the measurement results of the IMU sensor, Kalman filtering is used to predict the current state of pedestrians indoors, and the prediction result is obtained.

[0127] This embodiment uses a ForSense IMU614 nine-axis MEMS-IMU sensor, which is fixed to the heel of the right insole of the experimenter's foot to obtain measurement results. Based on the prediction results, a Kalman filter is used to predict the current state of the pedestrian indoors, obtaining the prediction result. The specific process includes:

[0128] The 9-dimensional state error of the measurement results is estimated using the Kalman filter algorithm, where the state variables include:

[0129] X k =[P k V k ψ k ] T

[0130] In the formula, X k P represents the IMU sensor state at time k; k V k and ψ k These represent the three-axis position, velocity, and attitude angles of the IMU sensor, respectively; [·] T Represents the transpose matrix;

[0131] Take systematic measurements to observe speed and altitude:

[0132] Z = [Z v Z p ] T

[0133] In the formula, Z v Z represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor. p This represents the difference between the theoretical and actual altitude values.

[0134] According to the state-space model of Kalman filtering, the state variable X and the quantity Z satisfy the following relationship:

[0135]

[0136] In the formula, W represents the system noise matrix; R represents the measurement noise matrix;

[0137] F is the system matrix:

[0138]

[0139] In the formula, S t The skew-symmetric matrix representing the projection of the accelerometer in the navigation frame; O i×j I represents the zero matrix of i×j; i×j Represents the identity matrix of i×j;

[0140] H is the measurement matrix:

[0141]

[0142] In the formula, h 21 =[0 0 1].

[0143] Then, the extraction module uses gait detection methods to extract the highest probability zero-velocity point of indoor pedestrians.

[0144] Zero-velocity correction, as a primary error correction method for inertial pedestrian navigation, typically involves detecting the magnitude and variance of acceleration and angular velocity within a window and comparing them to a set threshold. To improve the robustness of the zero-velocity correction algorithm, this embodiment proposes a maximum probability zero-velocity point extraction method based on gait detection. The process includes: performing zero-velocity detection using a generalized likelihood ratio-based method to obtain the detection results; dividing the detection area using an interval analysis method to obtain the gait interval for each step; and searching for the maximum probability zero-velocity point within each step's gait interval.

[0145] The above generalized likelihood ratio detection process includes:

[0146] Taking into account both acceleration and angular velocity information, the detection value within the window at time k is calculated using the formula. The detection value is then compared with a threshold to determine whether the current sample is in the zero-velocity interval.

[0147]

[0148] In the formula, Tk Represents the zero-velocity detection value at time k; W represents the window size; ω represents the noise variance of the gyroscope and accelerometer, respectively; k f k represents the output values ​​of the gyroscope and accelerometer at time k, respectively; g represents the local gravitational acceleration; This represents the average acceleration across the three axes within the window. When the detected value is less than the zero-velocity detector threshold, the sample data at that time can be considered to be in the zero-velocity interval, and the zero-velocity judgment value is true, as shown in the formula. When pedestrian motion involves multiple motion states, an appropriate detector threshold should be set to detect the possible intervals where zero-velocity points may exist in each state. In order to extract the intervals where zero-velocity points exist in each motion state, considering the large dynamic range of running, the zero-velocity detector threshold T should be set relatively large here.

[0149]

[0150] In the formula, z represents the judgment value of zero-speed detection.

[0151] To reduce the error introduced by inaccurate measurement information, this embodiment performs a systematic error correction only once at the maximum probability zero-velocity point within each gait cycle. Based on generalized likelihood ratio detection, zero-velocity intervals are divided, and interval analysis-based methods are used to constrain these intervals, thus completing the gait cycle division and enabling the extraction of the maximum probability zero-velocity point within each gait cycle. The specific process includes:

[0152] The sample analysis comprehensively considers walking, running, and climbing stairs, firstly constraining the gait interval length. Based on the gait segmentation method based on interval analysis, the length of the non-zero speed interval in each gait is considered to be no less than 0.12 seconds, and the interval length of a gait is no less than 0.6 seconds. The non-zero speed interval length is calculated; if the length is less than 0.12 seconds, a zero-speed point is considered misjudged, and the subsequent interval is designated as a zero-speed interval.

[0153] Based on the periodic pattern of foot movement during pedestrian movement, the zero-speed interval and the non-zero-speed interval alternate, which can be used to divide the complete gait interval. Taking the beginning time of each zero-speed interval as the starting point and the end time of the next non-zero-speed interval as the ending point, it is regarded as a complete gait interval.

[0154] Finally, the identification of the zero-velocity interval was achieved based on the generalized likelihood ratio test. Within a gait cycle, the point with the highest probability of zero foot velocity is near the minimum value detected by the zero-velocity detector. Therefore, within the zero-velocity interval of each gait, the point with the highest probability of zero velocity can be further extracted; this can be considered as the moment when the heel is fully in contact with the ground and begins to leave the ground in the swing phase at the next moment. Based on the above process, the detection value T... k The smaller the value, the greater the probability that the velocity is zero at time k. Therefore, the point with the highest probability of zero velocity can be extracted by finding the minimum detection value within the obtained zero velocity interval.

[0155] The estimation module uses a staircase step probability detection method to estimate the height a pedestrian climbs, and obtains the estimation result.

[0156] Currently, the main method for height correction in indoor navigation environments for pedestrians is to use barometric altimeters as height reference information. However, barometric altimeters are easily affected by environmental factors such as temperature and humidity. In addition, as the amplitude of pedestrian movement changes, the barometric altimeter will be subjected to varying degrees of vibration, resulting in significant fluctuations in the output value of the barometric altimeter when the pedestrian is moving.

[0157] To address this issue, this embodiment does not refer to barometric altimeter information when a pedestrian is moving on the same floor; instead, it uses the height of the previous step as a reference for height correction. Since the height of steps within a building is usually consistent, this embodiment assumes that the height of steps within the building remains constant when estimating the height of pedestrians going up and down stairs. To determine the number of steps crossed by a pedestrian's gait, this embodiment combines inertial calculation height information and barometric altimeter information for analysis. The algorithm flowchart is shown below. Figure 2 As shown. At the maximum probability zero-velocity point of each stair-climbing gait, inertial calculated altitude data and barometric altimeter data within that gait interval are extracted. An appropriate window size and step length are set, and a sliding window is used to traverse the gait data.

[0158] Differential feature values ​​are extracted from the inertial altitude and barometric altitude data within each sliding window, and then the energy functions of the inertial altitude difference and barometric altitude difference are calculated. The energy function calculation method includes:

[0159]

[0160] In the formula, z k D represents the maximum probability zero velocity point of the k-th gait; i This represents the difference value. The two energy functions above reflect the degree of change in inertial solution altitude and barometric altitude within the current data window, respectively.

[0161] Secondly, the energy functions are judged based on preset thresholds. If the energy functions of both the inertial solution altitude difference and the barometric altitude difference are below the preset thresholds, it is determined that the current gait segment has not crossed a step. Conversely, if either energy value exceeds the preset threshold, the current gait segment has crossed a step.

[0162] Finally, the number of steps crossed in all windows of the current gait is statistically analyzed to obtain the number of steps crossed in the pedestrian's gait when going up and down stairs.

[0163] To minimize the impact of inaccurate barometric altimeter output on altitude, this embodiment, based on motion state recognition results, only references the barometric altimeter's altitude information and calculates the average step height during the first segment of up and down stairs. Based on the assumption of a constant step height, subsequent up and down stairs are corrected by referencing the average step height information.

[0164]

[0165] In the formula, h s Δh represents the average height of the steps. p1 This represents the height difference measured by the barometric altimeter during the first segment of the up-and-down movement; n i represents the number of steps crossed in the i-th step; p and q represent the starting and ending gaits of the first segment of going up and down the stairs, respectively.

[0166] Subsequent changes in height during gait while going up and down stairs include:

[0167] Δh k =h s ·n k

[0168] In summary, this embodiment proposes a gait prediction method based on energy function to identify the number of steps crossed by the gait when going up and down stairs and obtain the estimation result.

[0169] Finally, the correction module uses the maximum probability zero velocity point and the estimation results to correct the prediction results and complete indoor pedestrian navigation.

[0170] First, velocity correction is performed using the maximum probability zero-velocity point. In this embodiment, error correction is performed at the maximum probability zero-velocity point for each gait; at this time, the theoretical velocity value is zero, Z. v This represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor, including:

[0171] Z v =[-V x -V y -V z ] T

[0172] In the formula, Vx V y V z These represent the three axial components of velocity; [·] T This represents the transpose of the matrix.

[0173] Finally, the height is corrected using the estimation results. In this embodiment, a height correction scheme based on gait type constraints is used to determine the current gait movement type, and the height correction value Z is determined. p satisfy:

[0174]

[0175] In the formula, h i，k h represents the solution height of the zero-velocity system at time k with the highest probability; p，k h represents the barometric altimeter altitude at the point of maximum probability zero velocity at time k; s n represents the average height of the steps; k This indicates the number of steps traversed during the current uphill / downhill gait; s k Indicators of current gait (going up or down stairs):

[0176]

[0177] Thus, indoor pedestrian navigation is completed through the above process.

[0178] The embodiments described above are merely preferred embodiments of this application and are not intended to limit the scope of this application. Any modifications and improvements made to the technical solutions of this application by those skilled in the art without departing from the spirit of this application shall fall within the protection scope defined by the claims of this application.

Claims

1. A method for indoor pedestrian navigation based on behavioral probability analysis, characterized by the steps of include: Based on the measurement results of the IMU sensor, Kalman filtering is used to predict the current state of pedestrians indoors, and the prediction result is obtained. Using gait detection methods, the highest probability zero-velocity point of indoor pedestrians is extracted; A staircase step probability detection method is used to estimate the height of pedestrians climbing stairs, and the estimation results are obtained. The steps include: when pedestrians are moving on the same floor, the height of the previous step is used as a reference for height correction without referring to barometric altimeter information; the inertial calculated height information and barometric altimeter information are analyzed, and at the maximum probability zero velocity point of each stair-climbing gait, the inertial calculated height data and barometric altimeter data within that gait interval are extracted, with the start time of each zero velocity interval as the starting point and the end time of the next non-zero velocity interval as the ending point, which is regarded as a complete gait interval; a sliding window is used to traverse the gait data, and the difference feature values ​​of the inertial calculated height and barometric altimeter data are extracted in each sliding window, and then the energy functions of the inertial calculated height difference and the barometric altimeter difference are calculated. In the formula, Indicates the first k The maximum probability zero velocity point of each gait; Indicates the difference value; The energy function is judged according to the preset threshold. If the energy functions of the inertial solution altitude difference and the barometric altitude difference are both lower than the preset threshold, it is determined that the current window gait segment has not crossed a step; if either of the energy values ​​exceeds the preset threshold, the current window gait segment has crossed a step. The number of steps crossed in all windows of the current gait is statistically analyzed to obtain the number of steps crossed in the pedestrian's gait when going up and down stairs; During the first phase of going up and down the stairs, the barometric altimeter's height information was used to calculate the average step height. Based on the assumption that the step height remained constant, the height was corrected during subsequent phases of going up and down the stairs by referring to the average step height information. In the formula, Indicates the average height of the steps; This indicates the height difference measured by the barometric altimeter during the first segment of the up-and-down movement. Indicates the first i The number of steps you cross; p , q These represent the beginning and end of the first segment of going up and down the stairs, respectively; Subsequent changes in height during gait while going up and down stairs include: ； Using the maximum probability zero velocity point and the estimation result, the prediction result is corrected to complete indoor pedestrian navigation.

2. The indoor pedestrian navigation method based on behavioral probability analysis according to claim 1, characterized in that, The method for obtaining the prediction result includes: estimating the 9-dimensional state error of the measurement result using a Kalman filter algorithm, wherein the state variables include: In the formula, X k express k IMU sensor state at time P; k V k and ψ k These represent the three-axis position, velocity, and attitude angles of the IMU sensor, respectively. Represents the transpose matrix; Take systematic measurements to observe speed and altitude: In the formula, Z v Z represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor. p This represents the difference between the theoretical and actual altitude values. According to the state-space model of Kalman filtering, the state variable X and the quantity Z satisfy the following relationship: In the formula, W represents the system noise matrix; R represents the measurement noise matrix; F is the system matrix: In the formula, S t The skew-symmetric matrix representing the projection of the accelerometer in the navigation frame; O i × j express i × j The zero matrix; I i × j express i × j The identity matrix; H is the measurement matrix: In the formula, .

3. The indoor pedestrian navigation method based on behavioral probability analysis according to claim 2, characterized in that, The method for performing the gait detection includes: A method based on generalized likelihood ratio detection was used to perform zero-speed detection of pedestrians indoors, and the detection results were obtained. The detection is divided based on the interval analysis method to obtain the gait interval of each step of the indoor pedestrian; The maximum probability zero velocity point is searched within the gait interval of each step.

4. The indoor pedestrian navigation method based on behavioral probability analysis according to claim 3, characterized in that, The method for correcting the prediction result using the maximum probability zero velocity point includes: Error correction is performed at the maximum probability zero-velocity point for each gait of an indoor pedestrian; at this moment, the theoretical velocity value is zero, Z. v This represents the difference between the theoretical and actual values ​​of the three-axis velocity of the IMU sensor, including: In the formula, Vx , Vy , Vz These represent the three axial components of velocity; This represents the transpose of the matrix.

5. The indoor pedestrian navigation method based on behavioral probability analysis according to claim 1, characterized in that, The method for correcting the prediction result using the estimation result includes: a height correction scheme based on indoor pedestrian gait type constraints, determining the current gait movement type, and a height correction value Z. p satisfy: In the formula, hi , k express k The system height calculated at the highest probability zero velocity point at any given moment; hp , k express k The barometric altimeter height at the point of maximum probability of zero velocity; Indicates the average height of the steps; nk This indicates the number of steps crossed during the current up or down gait; Indicators of current gait (going up or down stairs): 。 6. An indoor pedestrian navigation system based on behavioral probability analysis, the system being used to implement the method according to any one of claims 1-5, characterized in that, include: Prediction module, extraction module, estimation module, and correction module; The prediction module is used to predict the current state of indoor pedestrians based on the measurement results of the IMU sensor and using Kalman filtering to obtain the prediction result; The extraction module is used to extract the highest probability zero-velocity point of indoor pedestrians using gait detection methods. The estimation module is used to estimate the height of a pedestrian climbing stairs using a staircase step probability detection method, and obtain the estimation result; The correction module uses the maximum probability zero velocity point and the estimation result to correct the prediction result, thereby completing indoor pedestrian navigation.

Images