Pedestrian dead reckoning method based on bimodal odometer

Abstract

The invention provides a pedestrian dead reckoning method based on a bimodal odometer. The method comprises the following steps: firstly, constructing a frequency discrimination magnetic field odometer f-MOR by using double magnetometers and an IMU; fFT conversion is carried out on forward and backward magnetic signals output by the double magnetometers, the time difference of the double magnetometers is determined by using the maximum frequency amplitude f and the phase, and then an f-MOR speed measurement value is calculated based on the time difference; the time difference is determined in the more stable frequency domain of the signal, direct measurement and brute force matching of DTW on geometric correlation of the signal are avoided, and the speed measurement precision is higher. And then, constructing a state transition model about a step size scale factor error and an f-MOR speed error, constructing observation by using a difference between a speed integral of the f-MOR in a single-step interval and a BOR step size measurement result, and estimating an error state through a Kalman filter, so that the step size error is compensated with relatively high precision, and the positioning precision of autonomous navigation is effectively improved.

Description

Technical Field

[0001] The present invention relates to the fields of navigation, guidance and control, and in particular to a pedestrian dead reckoning method based on a dual-mode magnetic field odometer. Background Art

[0002] Pedestrian Dead Reckoning (PDR) is a pedestrian autonomous navigation method that relies on the body odometer (BOR) and heading angle. BOR is a statistical method that measures human stride length by fitting pedestrian motion inertial data. Compared with inertial navigation, satellite navigation, and other methods, BOR is a black box model that uses human inertial characteristics (such as peak-to-peak acceleration and cadence) as input and stride length as output. This lacks clear physical constraints, making the method non-analytical at the physical level. Consequently, it performs poorly in long-duration pedestrian navigation. Therefore, it is necessary to introduce reliable physical information sources to assist in correcting BOR measurement results to improve the positioning accuracy of the navigation system in long-duration positioning.

[0003] Magnetic Odometer (MOR) is a speed measurement method with strong autonomy and simple sensor layout and installation. This method can be used in navigation and positioning tasks for pedestrians, vehicles and other vehicles. Figure 1 As shown, compared to traditional wheel odometers that rely on encoder pulse counting for speed measurement, MOR deploys dual magnetometers along the vehicle's central axis, located on the front and rear sides. When the vehicle passes through an area with rich indoor magnetic field texture, the signals measured by the dual magnetometers exhibit a certain phase difference. Dynamic Time Wrapping (DTW) is used to measure the geometric similarity of the signals, allowing this phase difference to be estimated. The vehicle's forward velocity can be measured based on the phase difference and the pre-calibrated relative distance between the dual sensors. Combined with the vehicle's heading information output by the inertial navigation system, the vehicle's position can be inferred. Existing DTW methods, however, directly measure the geometric correlation of the signals and use brute-force matching, resulting in inaccurate and inefficient phase difference calculations. This, in turn, limits the positioning accuracy of autonomous navigation and the scope of MOR's applications. Summary of the Invention

[0004] In view of this, the present invention provides a pedestrian dead reckoning method based on a dual-mode magnetic field odometer, which can effectively improve the positioning accuracy of autonomous navigation.

[0005] In order to solve the above technical problems, the present invention is implemented as follows.

[0006] A pedestrian dead reckoning method based on a dual-modal odometry, comprising:

[0007] Step 1: The forward magnetometer in the dual magnetometer and the IMU form a forward sensor, which is strapped down to the front of the pedestrian's waist. The rearward magnetometer and the IMU form a rearward sensor, which is strapped down to the back of the pedestrian's waist. The two sensors form the frequency-determined magnetic field odometer (f-MOR).

[0008] Step 2: f-MOR velocity measurement: Perform FFT transformation on the forward and backward magnetic signals output by the dual magnetometers to obtain two frequency spectra. Determine the maximum frequency amplitude f and its corresponding phase based on the two frequency spectra. Use the maximum frequency amplitude f and phase to determine the time difference Δt between the dual magnetometers, and then calculate the f-MOR velocity measurement value based on the time difference Δt.

[0009] Step 3: Establish a combined navigation system model combining f-MOR and BOR:

[0010] Using the BOR method to obtain step length measurement right Perform single-step cycle integration to obtain mileage increment l k ; The observation vector for constructing the integrated navigation system model is Choose the state vector to be x k =[δk k δv f-MOR,k ] T ,δk k is the step size proportional error factor, δv f-MOR,k is the velocity measurement error of f-MOR; based on the state vector and observation vector, the state equation and observation equation of the integrated navigation system model are constructed;

[0011] Step 4: Design a Kalman filter based on the integrated navigation system model, estimate the step-size proportional error factor, and compensate it to the step-size measurement value Get the corrected step size The heading vector measurement value obtained by inertial navigation solution and correction step size Update pedestrian positions.

[0012] Preferably, step 2 further includes: truncating the two spectra obtained by FFT transformation according to a frequency amplitude threshold; and finding the maximum frequency amplitude f and the corresponding phase for the truncated spectrum.

[0013] Preferably, the time difference Δt of the dual magnetometers determined by using the maximum frequency amplitude f and phase is:

[0014] The phases corresponding to the maximum frequency amplitudes f of the two spectra are the forward phases and backward phase

[0015] Construct an auxiliary vector:

[0016]

[0017] The equivalent relationship of phase difference is constructed using the inner product of auxiliary vectors:

[0018]

[0019] Then the calculation using the time difference Δt is:

[0020]

[0021] Preferably, δv is calculated according to a first-order Markov model. f-MOR,k Modeling, and δk k The random walk process is modeled as follows:

[0022] δk k =δk k-Δ +w δk,k-Δ

[0023]

[0024] In the formula, Δ represents the single-step cycle, k-Δ represents the end time of the previous single-step cycle; δk k-Δ is the step length proportional error factor of the previous single-step cycle; δv f-MOR,k-Δ is the speed measurement error of the previous single-step cycle; w δk,k-Δ and is zero-mean Gaussian white noise; τ is the correlation time coefficient;

[0025] Based on δv f-MOR,k and δk k The model, combined with the state vector and observation vector, constructs the state equation and observation equation of the integrated navigation system model as follows:

[0026]

[0027] Among them, Φ k-Δ is the state transfer matrix of the previous single-step cycle, x k-Δ is the state of the previous single-step cycle, w y,k-Δ is the additive measurement noise; H k is the measurement matrix at time k.

[0028] Preferably, the method further comprises: constructing a Kalman filter of an adaptive observation covariance matrix using the observation residuals:

[0029]

[0030] Where R kis the measurement noise covariance matrix at time k, and α is the fusion threshold. Therefore, the overall process of the Kalman filter is designed as follows:

[0031]

[0032] Where, is the state estimate at time k, is the one-step prediction covariance matrix at time k, P k-Δ is the covariance matrix at k-Δ time, Q k-Δ is the process noise at time k-Δ, K k is the gain matrix at time k, and I is the identity matrix.

[0033] Preferably, α=0.2.

[0034] The present invention also provides an electronic device comprising: a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of any one of the above-mentioned pedestrian dead reckoning methods based on a dual-modal odometer are implemented.

[0035] The present invention further provides a readable storage medium having a program or instruction stored thereon, which, when executed by a processor, implements the steps of any one of the above-mentioned pedestrian dead reckoning methods based on a dual-modal odometer.

[0036] Beneficial effects:

[0037] (1) The frequency-modulated magnetic field odometry (f-MOR) scheme constructed in this invention determines the delay time Δt based on the amplitude and phase spectra of the signal in the more stable frequency domain, and then estimates the pedestrian's forward speed based on the pre-calibrated relative positions of the two sensors. This scheme avoids the direct measurement and brute-force matching of the geometric correlation of the signal used in the existing dynamic time warping (DTW) method, and the f-MOR speed measurement accuracy is higher.

[0038] (2) When calculating Δt, we further consider that the range of phase angle estimated by FFT is limited. Directly calculating the phase difference according to formula (2) may result in a problem where the zero-crossing difference is too large. Therefore, we construct an auxiliary vector to map the phase to the unit circle. The inner product of two vectors on the unit circle is used to calculate the time, thereby improving the stability of the calculation results.

[0039] (3) Compared with the MOR method based on DTW, the computational efficiency of f-MOR is faster: the DTW method first needs to expand the dual MMS into a distance matrix, and search for the shortest path in the matrix according to the minimum distance principle. Assuming that the length of the matching sequence is w, running DTW once requires about w calculations. If the length of the template sequence is n, the matching sequence needs to be slid in time and matched with the template sequence n times, then a total of nw calculations are required to calculate the speed value. The f-MOR method does not need to slide the matching sequence, but only needs to save the dual sensor MMS window of length w in real time, and calculate the speed through two FFTs and a small amount of calculations. Therefore, the proposed method has higher computational efficiency than the DTW method. Figure 8 It can be seen from the verification that the computational efficiency of the f-MOR proposed in this invention is improved by about 240 times compared with the EDTW method.

[0040] (4) The present invention constructs a state transition model for the step-size scale factor error and the f-MOR velocity error, and uses the difference between the velocity integral of the f-MOR within a single-step interval and the BOR step-size measurement result to construct an observation. The error state is estimated by the Kalman filter, thereby compensating the step-size error with high precision, effectively improving the positioning accuracy of autonomous navigation. The reasons are as follows: First, the step-size model used by the PDR is statistical rather than physically analytical, which makes the original PDR accumulate a large mileage error. This is beyond doubt. The present invention introduces a reliable step-size measurement method at the physical level to the statistical step-size model, and realizes the fusion of the BOR statistical characteristics and the f-MOR physical characteristics by constructing a Kalman filter, thereby improving the step-size estimation accuracy; second, the use of the existing EDTW method and GDTW method is based on the premise of ensuring the relative position stability of the sensors. Under the premise of stability, the MMS of the two sensors are more consistent in amplitude, which is conducive to phase matching according to the DTW method. Because the present invention addresses the requirement for immediate sensor use and avoids stringent installation requirements, the long-term geometry of the MMS is generally consistent. This means that the signal frequency beneficial for navigation in the MMS is stable, while the data details vary slightly over short periods of time. EDTW and GDTW measure phase differences based on geometric similarity. These methods consider data characteristics across the entire frequency band, and variations in data details over short periods of time lead to phase errors. The f-MOR proposed in this invention only considers stable signal characteristics at key frequencies. Therefore, the estimation accuracy of the EDTW and GDTW methods is lower than that of the proposed f-MOR method. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Schematic diagram of the speed measurement principle of MOR in the prior art.

[0042] Figure 2Schematic diagram of the principle of the pedestrian dead reckoning method based on the dual-modal odometer of the present invention.

[0043] Figure 3 Schematic diagram of magnetic intensity measurement signal simulation.

[0044] Figure 4 Schematic diagram of the amplitude spectrum and phase spectrum.

[0045] Figure 5 A comparison chart of the odometer step length estimation results for the 185-meter walking experiment.

[0046] Figure 6 A comparison chart of position estimation results from a 185-meter walking experiment.

[0047] Figure 7 This is a comparison chart of the mileage estimation results of the 185-meter walking experiment.

[0048] Figure 8 Comparison chart of loads calculated for the 185-meter walking experiment.

[0049] Figure 9 Schematic diagram of an electronic device for executing the pedestrian dead reckoning method of the present invention. DETAILED DESCRIPTION

[0050] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0051] To address the problem of inaccurate BOR estimation results in conventional PDR methods, the present invention proposes a pedestrian dead reckoning method based on a combination of frequency discrimination magnetic field odometry (f-MOR) and BOR. The f-MOR constructed by the present invention places dual magnetometers and an inertial measurement unit on the front and back sides of the waist. When a pedestrian passes through an indoor magnetic field distortion area, a certain phase difference exists in the magnetic intensity measurement signals of the dual sensors. f-MOR avoids the direct measurement and brute force matching of the geometric correlation of the signal by the existing dynamic time wrapping (DTW) method, and determines the delay time based on the amplitude spectrum and phase spectrum of the signal in the more stable frequency domain. Based on the pre-calibrated relative position of the dual sensors, the pedestrian's forward movement speed can be approximately estimated.

[0052] Furthermore, the present invention constructs a state transition model for the step size scale factor error and the f-MOR velocity error, and uses the difference between the velocity integral of the f-MOR in a single-step interval and the BOR step size measurement result to construct an observation. The error state is estimated by the Kalman filter to compensate for the step size error with high precision, thereby effectively improving the positioning accuracy of autonomous navigation.

[0053] The execution process of the pedestrian dead reckoning method based on the frequency discrimination magnetic field odometer of the present invention is as follows: Figure 2 As shown, it includes f-MOR velocity measurement method, f-MOR / BOR integrated navigation system design, application method and effect, etc. The specific implementation plan is as follows.

[0054] Step S1: Data acquisition.

[0055] The sensor module consists of dual magnetometers and an inertial navigation unit (IMU). The forward-facing magnetometer and IMU are integrated together to form the forward sensor, while the rearward-facing magnetometer and IMU are integrated together to form the rearward sensor. The two sensor modules are strapped down to the front and rear of the pedestrian's waist, respectively, and output a magnetic measurement signal (MMS) and inertial measurement information at a constant frequency.

[0056] Step S2: f-MOR velocity measurement.

[0057] The present invention selects the northeast celestial coordinate system as the navigation coordinate system (n system), selects the right front upper coordinate system as the carrier coordinate system (b system), and assumes that the carrier coordinate system coincides with the sensor coordinate system. When a pedestrian walks forward, the forward sensor and the backward sensor will measure the magnetic field intensity along the pedestrian's motion trajectory. Due to the consistency of the sensor motion trajectory, ideally, the MMS geometric features of the two should remain consistent but the phases should be different. The time difference Δt between the two can be calculated from the phase difference, and the speed can be approximately solved based on the pre-calibrated relative distance ΔL of the dual magnetometers.

[0058]

[0059] Due to the influence of factors such as sensor device errors, installation errors and pedestrian gait asymmetry, the MMS produces certain geometric deformations such as scaling and offset, making it impossible to directly calculate using the original signal features. Figure 3 The simulated MMS in this paper takes into account scaling, geomagnetic field and bias stacking offsets, a low-intensity 5Hz sinusoidal signal, and white noise errors. The forward true MMS leads the backward true MMS by pi / 4. Existing methods typically use correlation analysis based on minimum distance information (i.e., DTW) to measure the similarity of magnetic signals, which is susceptible to signal strength.

[0060] Compared with the time domain correlation measurement of signal strength, the frequency domain characteristics of the signal are more stable. Figure 3It is not difficult to conclude from the long-term geometric shape of the medium signal that the long-term trends of the two MMS are consistent, which shows that the effective low-frequency signal used for speed measurement is hidden in the MMS. Considering the consistency of the measurement trajectories of the two sensors, it can be considered that the frequencies of the effective signals of the two measurement sequences are close to the same, and the signal power at this specific frequency is larger.

[0061] Since the system analyzes the phase of the MMS based on measurements within the window, which satisfies the Dirichlet condition of the Fourier transform, it is feasible to analyze the signal characteristics from the frequency domain based on the Fast Fourier Transformation (FFT), and avoids direct measurement of the geometric correlation of the signal.

[0062] Therefore, if the MMS sequences of the forward sensor and the backward sensor from time kw to time k are m front,k-w:k and m back,k-w:k , perform FFT transformation on these two signals and obtain two spectra, whose amplitude spectrum (Magnitude) and phase spectrum (Phase) are as follows Figure 4 shown.

[0063] Depend on Figure 4 As can be seen, the MMS frequency is primarily concentrated around 0Hz, 1Hz, and 5Hz. The 0Hz signal is the DC component of the MMS. When a pedestrian walks in a straight line, this signal is primarily caused by the magnetometer bias and the relatively stable geomagnetic field. When a pedestrian walks along magnetic flux lines, the magnetic field gradient is assumed to be unchanged, making it impossible to effectively measure the signal phase difference. Therefore, the ultra-low-frequency signal around 0Hz can be ignored. The 1Hz and 5Hz signals are more likely to cause confusion. The position in three-dimensional space corresponds to the MMS intensity in a one-to-one relationship. Assuming that the set of position points of the pedestrian's motion trajectory is the horizontal axis of the coordinate system, the MMS intensity is the vertical axis of the coordinate system, and the frequency of the position-intensity signal curve is fixed, the slower the carrier moves along the motion trajectory, the lower the frequency of the time-intensity curve collected by the sensor. Since the sampling frequency of the sensor used in the present invention is relatively low, the walking speed range of the pedestrian is set to approximately 0.6m / s-1m / s. Therefore, the effective signal in the space often exhibits characteristics such as continuity, narrow bandwidth, and low frequency. Therefore, the cutoff frequency γHz can be set for the spectrum, that is, only the signal amplitude and phase less than γHz (γ=2Hz is selected in this embodiment) are retained.

[0064] The valid signal can be uniquely determined based on the maximum signal amplitude within the cutoff range. Figure 4 As shown in the figure, when the signal frequency is 1 Hz, the forward MMS leads the backward MMS by approximately 44 degrees, close to the reference setting of 45 degrees. Based on trigonometric relationships, the calculation formula for Δt is as follows.

[0065]

[0066] Where, For m front,k-w:k The phase of the forward MMS obtained by FFT, For m back,k-w:k The phase of the backward MMS obtained by FFT; f is the main frequency of the signal, that is, the frequency corresponding to the maximum amplitude in the amplitude-frequency curve.

[0067] Therefore, after obtaining the two spectra of FFT transformation in the above steps, the maximum frequency amplitude f and its corresponding phase are determined based on the two spectra, which are the forward phase and the phase. and backward phase Then use the maximum frequency amplitude f, and The time difference Δt between the two magnetometers is determined using the above formula (2), and the f-MOR velocity measurement value is calculated based on the time difference Δt.

[0068] In a preferred solution, considering that the range of phase angles estimated by FFT is limited, directly calculating the phase difference according to formula (2) may result in a problem where the zero-crossing difference is too large. In order to improve stability, the present invention constructs an auxiliary vector as follows:

[0069]

[0070] The inner product expression of the two is as follows:

[0071]

[0072] According to formula (4), we can get:

[0073]

[0074] Therefore, the improved calculation formula of Δt is as follows:

[0075]

[0076] Then, according to the pre-calibrated relative distance ΔL of the dual magnetometers, the velocity can be approximately solved using formula (2): As a measure of f-MOR speed.

[0077] The present invention calculates the delay Δt by constructing an auxiliary vector. This is because: the estimation range of FFT is -180 degrees to 180 degrees. For example, at time k-1, the phase of the forward MMS is 160 degrees, and the phase of the backward MMS is 140 degrees. The forward MMS leads the forward and backward MMS by 20 degrees. If the lead angle remains unchanged at time k, the phase of the forward MMS may be -170 degrees and the backward MMS becomes 170 degrees. The two signals have a large difference in numerical value, but the relative position on the unit circle is still 20 degrees. Therefore, the present invention maps the phase to the unit circle through an auxiliary vector and calculates the time by the inner product of the two vectors on the unit circle.

[0078] Step S3: f-MOR / BOR integrated navigation system design.

[0079] In order to improve the positioning accuracy of the pedestrian navigation system, the present invention constructs a combined navigation model of f-MOR and BOR.

[0080] The BOR method of the PDR position update model is used to measure the single step length. The output of f-MOR is the velocity measurement Define the pedestrian's single-step cycle as Δ. If it is assumed that the pedestrian moves at a uniform speed during the single-step cycle, it can be calculated by The pedestrian's speed can be calculated, thus enabling a simple fusion of BOR and f-MOR. However, the speed of the pedestrian's waist fluctuates periodically within a single step, and the fusion method mentioned above will lose accuracy.

[0081] Therefore, the present invention converts the speed measurement value output by f-MOR into Convert points into mileage increments k , and combined with BOR step measurement Construct the observation equation of the integrated navigation system:

[0082]

[0083] This step specifically includes the following sub-steps:

[0084] Step S3.1: Determine the heading vector using the PDR position update model and step length measurement

[0085] step length and heading It is a key element in the update of pedestrian dead reckoning system.

[0086] If the angular rate and acceleration of the waist backward sensor in frame b at time k are and Select the quaternion at time k To represent the waist backward sensor posture, the quaternion can be updated according to the following equation.

[0087]

[0088] Where Δt is the sampling period, [·]× is the fourth-order antisymmetric operator, I is the unit matrix, The angular rate of the b frame output at time k-1. The heading at time k The calculation formula is as follows:

[0089]

[0090] The BOR part of the PDR of the present invention uses the Weinberg step length model to calculate the pedestrian's single step length, and the single step length at time k is The calculation method is as follows:

[0091]

[0092] The specific calculation process is as follows:

[0093] 1) Obtain acceleration data from inertial measurement data 1) The modulus length of the pedestrian is determined and low-pass filtered; 2) The pedestrian's single-step interval is divided according to the peak detection method or the zero-crossing axis method; 3) The maximum value of the acceleration modulus f in the single-step interval is searched max and the minimum value f min 4) Set the step ratio parameter K and calculate the single step length according to formula (9); 5) At the update point, convert the INS quaternion into heading according to formula (8), and update the position at time k according to the following formula

[0094]

[0095] In the formula, the subscript Δ represents the single-step cycle, and k-Δ represents the end time of the previous single-step cycle. This formula only shows how the position is updated. The solution of the present invention will be used for optimization.

[0096] Step S3.2: f-MOR / BOR integrated navigation system model construction.

[0097] According to the basic principles of kinematics, the relationship between mileage p and velocity v in the carrier coordinate system is as follows:

[0098]

[0099] From this, the mileage calculation formula at time k and time k-Δ is:

[0100]

[0101] Where p k is the mileage at time k. Subtracting the two equations in (12) yields:

[0102]

[0103] Definition k =p k -p k-Δ , so we can get the ideal condition from time k-Δ to time k about l k and v f-MOR The equation is as follows:

[0104]

[0105] If there is and Where, and l k and v f-MOR,k The measured value, δk k is the step size proportional error factor, δv f-MOR,k is the velocity measurement error of f-MOR. Substituting the above equation into (14), the error equation is as follows:

[0106]

[0107] Assume δv f-MOR It is a constant from time k-Δ to time k, so we can simplify Equation (15) to get the value of δv f-MOR The observation equations for δk and δk are as follows:

[0108]

[0109] Considering that f-MOR may cause speed misestimation, it is necessary to reduce δv f-MOR Therefore, the present invention uses the first-order Markov model to calculate δv f-MOR Modeling is performed. Since there is a certain relationship between adjacent gaits of pedestrians, the present invention models δk as a random walk process. The state transition model is as follows:

[0110] δk k =δk k-Δ +w δk,k-Δ

[0111]

[0112] Where τ is the relevant time coefficient, which needs to be selected according to the specific user parameters.

[0113] Equations (17) and (16) are the state equation and observation equation of the integrated navigation system. Define the state vector at time k as x k =[δk k δv f-MOR,k ] T , then the f-MOR / BOR integrated navigation system model can be obtained as follows:

[0114]

[0115] Among them, Φ k-Δ is the state transfer matrix of the previous single-step cycle, x k-Δ is the state of the previous single-step cycle; w y,k-Δ is the additive measurement noise, H k is the measurement matrix at time k.

[0116] In order to further reduce the impact of f-MOR mismeasurement, the present invention uses the observation residual to construct a Kalman filter with an adaptive observation covariance matrix, namely:

[0117]

[0118] Where R k is the measurement noise covariance matrix at time k, α is the fusion threshold, and in this embodiment, α=0.2. Therefore, the overall process of the Kalman filter is as follows:

[0119]

[0120] Where, is the state estimate at time k, is the one-step prediction covariance matrix at time k, P k-Δ is the covariance matrix at k-Δ time, Q k-Δ is the process noise at time k-Δ, K k is the gain matrix at time k, and I is the identity matrix.

[0121] After the error state is estimated based on equations (17), (18) and (19) and combined with Kalman filtering, it is compensated to Then the step length after compensation is The heading vector measurement value obtained by inertial navigation solution and the compensated step length Update the pedestrian position. Equation (10) can be modified as follows:

[0122]

[0123] The application method and effects of the present invention are described below.

[0124] To verify the proposed algorithm and considering the portability and reliability of the sensor, the present invention selected Xsens DOT as the MEMS MIMU of the system. Xsens DOT integrates a three-axis gyroscope, a three-axis accelerometer, a three-axis magnetometer, and Bluetooth. The gyroscope has a range and bias stability of ±2000° / s and 10° / h, respectively; the accelerometer has a range and bias stability of ±16g and 0.03mg, respectively; and the magnetometer has a range and root mean square noise level of ±8Gauss and 0.5Gauss, respectively. All three sensors can upload data to a mobile phone via Bluetooth at a 60Hz output frequency and store it there. The sensor network can be automatically synchronized via control signals from the mobile phone.

[0125] The proposed method is compared with the DTW method based on empirical mode decomposition (EDTW) and the dual-window weighted DTW method based on gradient extraction (GDTW). Since the overall accuracy of the step length model output results is acceptable under normal gait, the speed output by the three odometry methods is converted into step length according to Equation (14) and compared with the BOR output results. The closer the odometry step length estimate is to the step length model, the higher the odometry estimation accuracy can be considered. Subsequently, the three odometry methods are applied to PDR, and the original PDR results are combined to compare the odometry accuracy and terminal positioning accuracy of the four methods to evaluate the effectiveness of each method.

[0126] The present invention designs an indoor 185-meter walking experiment to comprehensively evaluate the performance of each method. Figure 5 This is a comparison of the step-size estimation results (Step-Size) in this experiment, where SSM stands for the Weinberg step-size model. Figure 5 As can be seen from the curve on the left, the step length estimation results of the three methods are basically around the SSM estimation results, and there are no abnormal points with particularly large values. This seems to indicate the feasibility of the three methods, so they need to be further demonstrated. Compared with the proposed f-MOR, EDTW and GDTW have more large outliers, which can be confirmed by the histogram on the right. The histogram on the right represents the difference between the step length estimation results of the odometer and the SSM results (Step-Size Difference). It can be seen from the figure that the distribution range of f-MOR (about -0.5m~1m) is narrower than that of EDTW (-0.7m~2m) and GDTW (-0.7m~1.6m). And from the perspective of the mean difference, the overall distribution of f-MOR is more uniform than that of the comparison method, and its mean is about -0.0297m, which is much lower than the mean of EDTW (-0.2829m) and the mean of GDTW (-0.2076m). Therefore, Figure 5It shows that the method proposed in this paper is closer to the output value of the step size model. From the perspective of multi-source information fusion based on Kalman filtering, compared with the EDTW method and the GDTW method, the proposed method can achieve higher accuracy when fused with the SSM output. This can be seen from Figure 6 and 7 verify.

[0127] Figure 6 The following is a comparison of the position estimation results of the four methods. Due to the lack of dense indoor absolute reference points due to experimental conditions, a portable laser rangefinder was used to measure the total walking distance (Step Counts). Figure 6 In the figure, the pedestrian initially faces approximately east and walks approximately 81 meters east before walking in the opposite direction, for a total walking distance of approximately 185 meters. The terminal errors further validate the effectiveness of the proposed method. The terminal errors for the original PDR method, EDTW-aided PDR method, GDTW-aided PDR method, and the proposed method are 2.2409m, 1.4170m, 0.5548m, and 0.1419m, respectively. Figure 7 This also helps validate the accuracy of the proposed method. Compared to the original PDR method, the EDTW-aided PDR method, and the GDTW-aided PDR method, the proposed method improves by approximately 93.67%, 89.99%, and 74.42%, respectively. This is primarily due to the following reasons: First, the step-size model used in PDR is statistical rather than physically analytical, which undoubtedly leads to a large accumulated odometry error in the original PDR. Second, the EDTW and GDTW methods require the relative position of the sensors to be stable. Under such conditions, the amplitudes of the MMS of the two sensors are more consistent, thus facilitating phase matching based on the DTW method. Since this paper aims to ensure the sensors are ready for immediate use, rigorous installation is avoided. As a result, the long-term geometry of the MMS is generally consistent, while the data details may differ slightly over short periods of time, resulting in reduced matching accuracy for the EDTW and GDTW methods.

[0128] In addition to method accuracy, calculated load is also one of the important indicators for measuring method performance. Since both the EDTW method and the GDTW method are based on DTW, and in order to better clearly show the operating efficiency of the method proposed in this article, the present invention compares the computational efficiency of the proposed method with that of the EDTW method, runs the two programs in the MATLAB R2021b environment equipped with an Intel i9-14900HX notebook, and records the running time of a single processing. The results are recorded as follows: Figure 8 shown.

[0129] Depend on Figure 8 It can be seen that the computational efficiency of the proposed f-MOR method is approximately 240 times higher than that of the EDTW method. This is because the DTW method first needs to expand the dual MMS into a distance matrix and search for the shortest path within the matrix based on the principle of minimum distance. Assuming the matching sequence length is w, a DTW run requires approximately w calculations. If the template sequence length is n, the matching sequence must be slid in time and matched with the template sequence n times, resulting in a total of nw calculations required to calculate the velocity value. The f-MOR method, on the other hand, does not require sliding the matching sequence. It only needs to save the dual-sensor MMS window of length w in real time and calculate the velocity through two FFTs (the time complexity of FFT is O(nlogn)) and a small number of operations. Therefore, the proposed method has higher computational efficiency than the DTW method.

[0130] In addition, the pedestrian dead reckoning method based on the dual-modal odometer in the embodiments of the present application can be implemented by a computer device. Figure 9 Schematic diagram of the hardware structure of the computer device of the embodiment of the present application. Figure 9 As shown, the device may include a processor 201 and a memory 202 storing computer program instructions.

[0131] Specifically, the processor 201 may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0132] Among them, the memory 202 may include a large-capacity memory for data or instructions. By way of example and not limitation, the memory 202 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 202 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 202 may be inside or outside the data processing device. In a specific embodiment, the memory 202 is a non-volatile memory. In a specific embodiment, the memory 202 includes a read-only memory (ROM) and a random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically alterable ROM (EAROM) or a flash memory (FLASH), or a combination of two or more of these. Under appropriate circumstances, the RAM can be a static random access memory (SRAM) or a dynamic random access memory (DRAM), where the DRAM can be a fast page mode dynamic random access memory (FPMDRAM), an extended data output dynamic random access memory (EDODRAM), a synchronous dynamic random access memory (SDRAM), etc.

[0133] The memory 202 may be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 201 .

[0134] The processor 201 reads and executes computer program instructions stored in the memory 202 to implement any one of the pedestrian dead reckoning methods based on the dual-modal odometry in the above embodiments.

[0135] In some embodiments, the computer device may further include a communication interface 203 and a bus 200. Figure 9 As shown, the processor 201 , the memory 202 , and the communication interface 203 are connected via a bus 200 and communicate with each other.

[0136] The communication interface 203 is used to implement communication between the various modules, devices, units, and / or devices in the embodiments of the present application. The communication interface 203 can also implement data communication with other components such as: external devices, image / data acquisition equipment, databases, external storage, and image / data processing workstations.

[0137] The bus 200 includes hardware, software, or both, and couples components of a computer device to each other. The bus 200 includes, but is not limited to, at least one of the following: a data bus, an address bus, a control bus, an expansion bus, and a local bus. By way of example, and not limitation, bus 200 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Bus 200 may include one or more buses, where appropriate. Although embodiments herein describe and illustrate a particular bus, this application contemplates any suitable bus or interconnect.

[0138] The computer device can execute the pedestrian dead reckoning method based on the dual-modal odometry in the embodiment of the present application, thereby realizing the pedestrian dead reckoning described in the present application.

[0139] In addition, in conjunction with the pedestrian dead reckoning method based on a dual-modal odometry in the above-mentioned embodiments, embodiments of the present application may provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when executed by a processor, the computer program instructions implement any of the pedestrian dead reckoning methods based on a dual-modal odometry in the above-mentioned embodiments.

[0140] It should be noted that the various technical features of the above-described embodiments can be combined in any manner. To simplify the description, not all possible combinations of the various technical features in the above-described embodiments are described. However, as long as there are no contradictions in the combination of these technical features, they should be considered to be within the scope of this specification. In addition, according to the needs of implementation, the various steps / components described in this application can be split into more steps / components, and two or more steps / components or partial operations of steps / components can be combined into new steps / components to achieve the purpose of the present invention.

[0141] Those skilled in the art will readily understand that the above-described embodiments merely represent several implementation methods of the present application, and their descriptions are relatively specific and detailed, but should not be construed as limiting the scope of the invention. It should be noted that a person of ordinary skill in the art may make several variations and improvements without departing from the concept of the present application, and these variations and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be based on the appended claims.

Claims

1. A pedestrian dead reckoning method based on a dual-modal odometry, characterized in that: include: Step 1: The forward magnetometer in the dual magnetometer and the IMU form a forward sensor, which is strapped down to the front side of the pedestrian's waist. The rearward magnetometer and IMU form a rearward sensor, strapped to the back of the pedestrian's waist. The two sensors form the frequency-determined magnetic field odometer (f-MOR). Step 2: f-MOR velocity measurement: Perform FFT transformation on the forward and backward magnetic signals output by the dual magnetometers to obtain two spectra. Based on the two spectra, determine the maximum frequency amplitude f and its corresponding phase. The maximum frequency amplitude f and phase are used to determine the time difference Δt between the two magnetometers, and then the f-MOR velocity measurement value is calculated based on the time difference Δt. Step 3: Establish a combined navigation system model combining f-MOR and BOR: Using the BOR method to obtain step length measurement right Perform single-step cycle integration to obtain mileage increment l k ; The observation vector for constructing the integrated navigation system model is Choose the state vector to be x k =[δk k δv f-MOR,k ] T ,δk k is the step size proportional error factor, δv f-MOR,k is the velocity measurement error of f-MOR; based on the state vector and observation vector, the state equation and observation equation of the integrated navigation system model are constructed; Step 4: Design a Kalman filter based on the integrated navigation system model, estimate the step-size proportional error factor, and compensate it to the step-size measurement value Get the corrected step size The heading vector measurement value obtained by inertial navigation solution and correction step size Update pedestrian positions.

2. The pedestrian dead reckoning method based on a dual-modal odometer according to claim 1, wherein: The step 2 further includes: truncating the two spectra obtained by FFT transformation according to the frequency amplitude threshold; and finding the maximum frequency amplitude f and the corresponding phase for the truncated spectrum.

3. The pedestrian dead reckoning method based on a dual-modal odometer according to claim 1 or 2, wherein: The time difference Δt of the dual magnetometers determined by using the maximum frequency amplitude f and phase is: The phases corresponding to the maximum frequency amplitudes f of the two spectra are the forward phases and backward phase Construct an auxiliary vector: The equivalent relationship of phase difference is constructed using the inner product of auxiliary vectors: Then the calculation using the time difference Δt is:

4. The pedestrian dead reckoning method based on a dual-modal odometer according to claim 1, wherein: According to the first-order Markov model, δv f-MOR,k Modeling, and δk k The random walk process is modeled as follows: δk k =δk k-Δ +w δk,k-Δ In the formula, Δ represents the single-step cycle, k-Δ represents the end time of the previous single-step cycle; δk k-Δ is the step length proportional error factor of the previous single-step cycle; δv f-MOR,k-Δ is the speed measurement error of the previous single-step cycle; w δk,k-Δ and is zero-mean Gaussian white noise; τ is the correlation time coefficient; Based on δv f-MOR,k and δk k The model, combined with the state vector and observation vector, constructs the state equation and observation equation of the integrated navigation system model as follows: Among them, Φ k-Δ is the state transfer matrix of the previous single-step cycle, x k-Δ is the state of the previous single-step cycle, w y,k-Δ is the additive measurement noise; H k is the measurement matrix at time k.

5. The pedestrian dead reckoning method based on a dual-modal odometer according to claim 4, wherein: The method further includes constructing a Kalman filter with an adaptive observation covariance matrix using the observation residuals: Where R k is the measurement noise covariance matrix at time k, and α is the fusion threshold. Therefore, the overall process of the Kalman filter is designed as follows: Where, is the state estimate at time k, is the one-step prediction covariance matrix at time k, P k-Δ is the covariance matrix at k-Δ time, Q k-Δ is the process noise at time k-Δ, K k is the gain matrix at time k, and I is the identity matrix.

6. The pedestrian dead reckoning method based on a dual-modal odometer according to claim 5, wherein: Take α=0.

2.

7. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the pedestrian dead reckoning method based on a dual-modal odometer as described in any one of claims 1 to 6 are implemented.

8. A readable storage medium, characterized in that: A program or instruction is stored thereon, and when the program or instruction is executed by a processor, the steps of the pedestrian dead reckoning method based on a dual-modal odometer as described in any one of claims 1 to 6 are implemented.

Images