Adaptive zero-speed detection method based on interval search under variable speed motion conditions

Abstract

The present invention relates to an adaptive zero-speed detection algorithm based on interval search under variable speed motion conditions, and belongs to the field of pedestrian navigation technology. It includes: 1) zero-speed interval proportion analysis; 2) gait cycle extraction; 3) preliminary screening of zero-speed intervals; 4) fine determination of zero-speed intervals. The present invention determines the search interval of the gait cycle by performing SWA filtering on the modulus of angular velocity, and preliminarily determines a zero-speed interval based on the statistical characteristics of the zero-speed interval in the gait cycle under different speeds, and sets a new threshold based on the 3σ criterion to achieve adaptive zero-speed detection. The present invention not only provides a new, efficient and accurate positioning solution for wearable devices and location services, but also provides a valuable reference for pedestrian navigation research under complex motion patterns.

Description

Technical Field

[0001] The invention relates to an adaptive zero-speed detection algorithm based on interval search under variable-speed motion conditions, and belongs to the technical field of pedestrian navigation. Background Art

[0002] In the current technological landscape, indoor pedestrian positioning and navigation technologies are primarily categorized into infrastructure-dependent and infrastructure-independent approaches. While infrastructure-dependent approaches, such as radio frequency identification (RFID), Wi-Fi, Bluetooth, and ultra-wideband (UWB) technologies, offer high positioning accuracy, their high installation and maintenance costs significantly limit their widespread adoption. In contrast, infrastructure-independent approaches, particularly the Pedestrian Inertial Navigation System (PINS), demonstrate significant potential for application in a variety of fields, including firefighting and disaster relief, due to their autonomy, flexibility, and resistance to weather and electromagnetic interference.

[0003] With the continuous advancement of micro-electromechanical system (MEMS) inertial measurement unit (IMU) technology, its low cost and high degree of autonomy have further promoted the research and development of PINS. However, pedestrian navigation systems based on miniature inertial measurement units (MIMUs) face a major problem: the accumulation of errors over time, which seriously affects positioning accuracy. To effectively address this problem and achieve continuous and accurate positioning in unknown environments, the Zero Velocity Update (ZUPT) technology based on foot MIMUs has emerged. ZUPT technology detects the zero velocity interval (ZVI) of foot movement and uses the velocity error as the observation value of the Kalman filter to estimate and compensate for the error state of the pedestrian inertial navigation system, thereby significantly improving positioning accuracy.

[0004] Traditional ZUPT algorithms primarily rely on threshold-based accelerometer or gyroscope measurements for zero-velocity detection, such as acceleration motion variance detectors, acceleration amplitude detectors, and angular rate energy detectors. While these methods can achieve zero-velocity detection to a certain extent, they struggle to adapt to diverse motion patterns and are highly sensitive to the threshold setting. To optimize and improve the ZUPT algorithm, researchers have proposed adaptive threshold-based methods, such as a weighted average detector that considers sensor prior performance, a method that models the zero-velocity threshold as a function of interval time and foot impact level, and a zero-velocity detector based on adaptive simulated energy consumption curves. While these methods improve the accuracy of zero-velocity detection to a certain extent, their precision is still limited by the accuracy of the threshold modeling. In recent years, with the development of artificial intelligence technology, deep learning methods have been applied to intelligent ZVI detection and have achieved remarkable results. However, deep learning methods require a large amount of training data to continuously train the model, which not only consumes a large amount of human and computing resources, but also has poor generalization ability. For new situations not covered in the training data, the stability and reliability of the classification are significantly reduced. Summary of the Invention

[0005] In response to the shortcomings of the above-mentioned prior art, the present invention provides an adaptive zero-speed detection algorithm based on interval search under variable speed motion conditions. The algorithm determines the search interval of the gait cycle by performing SWA filtering on the modulus of the angular velocity, and preliminarily determines a zero-speed interval based on the statistical characteristics of the zero-speed interval in the gait cycle under different speeds. Subsequently, a new threshold is set based on the 3σ criterion to achieve adaptive zero-speed detection. The present invention not only provides a new, efficient and accurate positioning solution for wearable devices and location services, but also provides a valuable reference for pedestrian navigation research under complex motion patterns.

[0006] The adaptive zero-speed detection algorithm based on interval search under variable speed motion conditions of the present invention is special in that it includes the following steps:

[0007] Step 1) Analysis of the proportion of zero-speed intervals

[0008] Through experimental data analysis, the proportion of zero-speed intervals at different walking speeds is determined, and the minimum proportion of walking zero-speed intervals is set accordingly. This provides a basis for the subsequent preliminary screening of zero-speed intervals and enhances the robustness of the algorithm.

[0009] Step 2) Gait cycle extraction

[0010] The angular velocity is synthesized and filtered, and the maximum points of the gait cycle are identified by taking the first-order and second-order derivatives. Based on the intervals and characteristics of the maximum points, representative maximum points are screened to determine the complete gait cycle.

[0011] Step 3) Preliminary screening of zero-speed intervals

[0012] A mathematical model for zero-speed detection was constructed, and the generalized likelihood ratio test method was used to calculate the detection statistic. Combining the duration of each gait cycle and the minimum proportion of the zero-speed interval set in step 1, the time range of some zero-speed intervals was preliminarily determined, providing a basis for subsequent fine-tuning.

[0013] Step 4) Fine determination of zero speed range

[0014] Within the preliminarily determined zero-speed interval, the mean and standard deviation of the detection data are calculated, and the threshold of the gait cycle is adaptively adjusted and set. The zero-speed interval and non-zero-speed interval are judged based on the new threshold to achieve precise determination of the zero-speed interval.

[0015] Preferably, the specific steps of step 1 are:

[0016] First, an experiment was conducted to analyze the zero-speed interval at different walking speeds. The analysis showed that the proportion of zero-speed intervals decreases with increasing speed. The daily walking speed of pedestrians does not exceed 8 km / h. Within this speed range, the minimum zero-speed interval proportion is as low as 13.58%, which is the lowest value among all tested speeds. A robust minimum proportion of zero-speed intervals was determined. , that is, the minimum proportion of the walking zero-speed interval is set to 10%.

[0017] Preferably, the specific steps of step 2 are:

[0018] To identify a complete gait cycle, the angular velocity is first synthesized to obtain :

[0019] ;

[0020] Where: 、 、 MIMU in The angular velocity components of the three axes are measured at all times. for The modulus of the angular velocity at the moment, SWA filtering is performed on the modulus of the angular velocity:

[0021] ;

[0022] Where: is the window size of SWA, which is set to half of the IMU output frequency, is the angular velocity modulus obtained after SWA filtering. Calculating its first-order and second-order derivatives, we can establish the following equation:

[0023] ;

[0024] Obtain n maximum points, recorded as ,in Indicates the At the moment corresponding to the maximum point, in order to screen out the representative maximum point, the following formula can be used to determine and eliminate the pseudo maximum point :

[0025] ;

[0026] in, and Used to represent the first two retained maximum points. If the formula is satisfied, Add to new collection If it is not satisfied, it means that the two maximum points are close to each other. and The corresponding values ​​are compared, if , then Value Assignment ,if ,but remain unchanged, and The time between 1 and 2 is a complete gait cycle;

[0027] Preferably, the specific steps of step 3 are:

[0028] The mathematical model of zero speed detection is regarded as a binary hypothesis testing problem, which can be expressed as:

[0029] ;

[0030] Where: express The hypothesis detection statistic of pedestrian gait at time t, represents the threshold value, The calculation methods used include acceleration amplitude detection, acceleration variance detection, angular velocity amplitude detection, angular velocity energy detection and generalized likelihood ratio detection. Based on the generalized likelihood ratio detection method, first, the first Subsampling , including 3-axis acceleration and 3-axis angular rate:

[0031] ;

[0032] is The data samples are a data set consisting of sliding windows:

[0033] ;

[0034] Always Using the generalized likelihood ratio test method, we get the test statistic :

[0035] ;

[0036] Where: express The generalized likelihood ratio test statistic at time t, Set the sliding window to 30, and is the variance of acceleration and angular rate noise, is the average acceleration of each axis in the current window, is the acceleration due to gravity In order to improve the accuracy of zero-speed detection, based on the generalized likelihood ratio detection algorithm and combined with the duration of each gait cycle, the accurate conservative zero-speed interval time range is preliminarily determined. :

[0037] ;

[0038] In this range Size is used as a sliding window, and the of Perform the calculation:

[0039] ;

[0040] Traverse to find The minimum value and record The value is ,but It is a preliminarily determined zero-speed interval.

[0041] Preferably, the specific steps of step 4) are:

[0042] Calculate the average value of GLR data within the initially determined zero-speed range and standard deviation :

[0043] ;

[0044] ;

[0045] By analyzing the characteristics of the zero-speed interval data initially determined in each gait cycle, the threshold of the gait cycle is adaptively adjusted. :

[0046] ;

[0047] Then, the zero speed interval and the non-zero speed interval are determined according to the formula:

[0048] ;

[0049] Where N=1 represents the zero speed interval, and N=0 represents the non-zero speed interval.

[0050] The adaptive zero-speed detection algorithm based on interval search under variable speed motion conditions of the present invention has the following beneficial effects:

[0051] (1) Traditional methods use fixed thresholds, which are difficult to adapt to changes in different walking speeds and complex motion patterns. This invention dynamically adjusts the threshold based on the 3σ criterion through interval search and statistical characteristic analysis, thereby adapting to the zero-speed detection requirements at various walking speeds.

[0052] (2) Compared with the zero-speed detection algorithm based on deep learning, this algorithm does not require a large amount of training data, which reduces the consumption of manpower and computing resources.

[0053] Experiments have shown that the interval search adaptive threshold zero-speed detection algorithm proposed in the present invention not only effectively solves the fixed threshold dependency problem in existing algorithms, but also achieves a significant improvement in navigation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a flow chart of the present invention;

[0055] Figure 2 This is a comparison chart of the proportion of zero-speed intervals at different walking speeds, where (a), (b), (c), and (d) are the proportion of zero-speed intervals at walking speeds of 8km / h, 6km / h, 4km / h, and 2km / h, respectively;

[0056] Figure 3 It is a comparison chart of the original and filtered angular velocity modulus values;

[0057] Figure 4 It is the angular velocity maximum point determination diagram after filtering;

[0058] Figure 5 It is a schematic diagram of the low threshold selection situation;

[0059] Figure 6 It is a low-threshold local magnification image;

[0060] Figure 7 This is a schematic diagram of the high threshold selection situation;

[0061] Figure 8 It is a high threshold local magnification image;

[0062] Figure 9 This is a comparison diagram of the present invention and the fixed threshold method for complex motion paths;

[0063] Figure 10 : is a motion solution comparison diagram, wherein (a) is a comparison of the position error solved by the present invention and the fixed threshold method, (b) is a comparison of the average position error solved by the present invention and the fixed threshold method, and (c) is a comparison of the standard deviation solved by the present invention and the fixed threshold method;

[0064] Figure 11 This is a comparison diagram of the present invention and the fixed threshold method in a long-distance experimental motion path. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0066] Example 1:

[0067] This embodiment is an adaptive zero speed detection algorithm based on interval search under variable speed motion conditions. The flowchart is shown in the attached Figure 1 , including the following steps:

[0068] Step 1) Analysis of the proportion of zero-speed intervals

[0069] Through experimental data analysis, such as Figure 2 As shown, (a), (b), (c), and (d) are the proportions of zero-speed intervals at walking speeds of 8 km / h, 6 km / h, 4 km / h, and 2 km / h, respectively, and the minimum proportion of the walking zero-speed interval is set to 10% accordingly.

[0070] Step 2) Gait cycle extraction

[0071] The angular velocity is synthesized and filtered, such as Figure 3 As shown in , the maximum points of the gait cycle are identified by finding the first and second order derivatives. Based on the intervals and characteristics of the maximum points, representative maximum points are screened out, such as Figure 4 As shown, the complete gait cycle is determined.

[0072] Step 3) Preliminary screening of zero-speed intervals

[0073] A mathematical model for zero-speed detection was constructed, and the generalized likelihood ratio test method was used to calculate the detection statistic. Combining the duration of each gait cycle and the set minimum proportion of 10% for walking zero-speed intervals, the time range of some zero-speed intervals was preliminarily determined, providing a basis for subsequent detailed determination.

[0074] Step 4) Fine determination of zero speed range

[0075] Within the initially determined zero-speed interval, the mean and standard deviation of the detection data are calculated, and the threshold for that gait cycle is adaptively adjusted. Based on the new threshold, the zero-speed interval and non-zero-speed interval are determined, achieving precise determination of the zero-speed interval.

[0076] The specific steps of step 1 are:

[0077] First, an experiment was conducted to analyze the zero-speed interval at different walking speeds. The analysis showed that the proportion of zero-speed intervals decreases with increasing speed. The daily walking speed of pedestrians does not exceed 8 km / h. Within this speed range, the minimum zero-speed interval proportion is as low as 13.58%, which is the lowest value among all tested speeds. A robust minimum proportion of zero-speed intervals was determined. , that is, the minimum proportion of the walking zero-speed interval is set to 10%.

[0078] The specific steps of step 2 are:

[0079] To identify a complete gait cycle, the angular velocity is first synthesized to obtain :

[0080] ;

[0081] Where: 、 、 MIMU in The angular velocity components of the three axes are measured at all times. for The modulus of the angular velocity at the moment, SWA filtering is performed on the modulus of the angular velocity:

[0082] ;

[0083] Where: is the window size of SWA, which is set to half of the IMU output frequency, is the angular velocity modulus obtained after SWA filtering. Calculating its first-order and second-order derivatives, we can establish the following equation:

[0084] ;

[0085] Obtain n maximum points, recorded as ,in Indicates the At the moment corresponding to the maximum point, in order to screen out the representative maximum point, the following formula can be used to determine and eliminate the pseudo maximum point :

[0086] ;

[0087] in, and Used to represent the first two retained maximum points. If the formula is satisfied, Add to new collection If it is not satisfied, it means that the two maximum points are close to each other. and The corresponding values ​​are compared, if , then Value Assignment ,if ,but remain unchanged, and The time between 1 and 2 is a complete gait cycle;

[0088] The specific steps of step 3 are:

[0089] The mathematical model of zero speed detection is regarded as a binary hypothesis testing problem, which can be expressed as:

[0090] ;

[0091] Where: express The hypothesis detection statistic of pedestrian gait at time t, represents the threshold value, The calculation methods used include acceleration amplitude detection, acceleration variance detection, angular velocity amplitude detection, angular velocity energy detection and generalized likelihood ratio detection. Based on the generalized likelihood ratio detection method, first, the first Subsampling , including 3-axis acceleration and 3-axis angular rate:

[0092] ;

[0093] is The data samples are a data set consisting of sliding windows:

[0094] ;

[0095] Always Using the generalized likelihood ratio test method, we get the test statistic :

[0096] ;

[0097] Where: express The generalized likelihood ratio test statistic at time t, Set the sliding window to 30, and is the variance of acceleration and angular rate noise, is the average acceleration of each axis in the current window, is the acceleration due to gravity In order to improve the accuracy of zero-speed detection, based on the generalized likelihood ratio detection algorithm and combined with the duration of each gait cycle, the accurate conservative zero-speed interval time range is preliminarily determined. :

[0098] ;

[0099] In this range Size is used as a sliding window, and the of Perform the calculation:

[0100] ;

[0101] Traverse to find The minimum value and record The value is ,but It is a preliminarily determined zero-speed interval.

[0102] The specific steps of step 4) are:

[0103] Calculate the average value of GLR data within the initially determined zero-speed range and standard deviation :

[0104] ;

[0105] ;

[0106] By analyzing the characteristics of the zero-speed interval data initially determined in each gait cycle, the threshold of the gait cycle is adaptively adjusted. :

[0107] ;

[0108] Then, the zero speed interval and the non-zero speed interval are determined according to the formula:

[0109] ;

[0110] Where N=1 represents the zero speed interval, and N=0 represents the non-zero speed interval.

[0111] Example 2:

[0112] (1) Verification of the effectiveness of the short-distance complex speed-varying motion algorithm.

[0113] This invention simulates walking on a short distance with complex road conditions. The experimental subjects walk along a 100-meter concave path, experience 6 turns and adjust the speed irregularly. Position error is used as a key indicator to measure walking accuracy and algorithm performance. In the process of calculating navigation error, the choice of reference point must be clarified first. and its corresponding point in the actual trajectory , through the position error To directly reflect the difference between the solved trajectory and the actual trajectory:

[0114] ;

[0115] To comprehensively evaluate the navigation performance of the entire trajectory, the average of these position errors was also calculated. . Provides an intuitive indicator to measure the overall deviation between the solved trajectory and the actual trajectory, that is, the average level of navigation error:

[0116] ;

[0117] in, Indicates the turning points and end points passed in sequence during the walking process. express The total number of this experiment is 7.

[0118] In addition, in order to reflect the fluctuation range of the position error, that is, the stability of the difference between the calculated trajectory and the actual trajectory, it is also necessary to calculate the standard deviation of the position error :

[0119] ;

[0120] By using and These two indicators can be used to quantitatively compare the trajectory deviations of different algorithms, and then evaluate the difference in navigation accuracy between Algorithm 1 (the algorithm of the present invention) and Algorithm 2 (the traditional fixed threshold algorithm).

[0121] Before the comparison, considering the randomness of walking speed, in order to make the algorithm 2 work best, the threshold traversal screening method is used, such as Figure 5-Figure 8 As shown in the figure, we ultimately determined a fixed threshold of 40,000 suitable for this experiment. This threshold can both stably identify the zero-speed interval and effectively reduce false positives. This threshold ensures the algorithm's stability when dealing with random walking speeds.

[0122] like Figure 9As shown in Table 1, when the experimental starting point is the same, Algorithm 1 is closer to the real trajectory than Algorithm 2. Specifically, as shown in Table 1, when comparing the performance of Algorithm 1 and Algorithm 2, an analysis is conducted based on the distance between each reference point and its corresponding point, the average position error, and the standard deviation. Figure 10 As shown in (a), the distance values ​​of Algorithm 1 are relatively stable, with most values ​​between 0.5 and 1.4, and the difference between the maximum and minimum distance error values ​​is small, indicating that the error distribution of the algorithm at each reference point is relatively uniform. In contrast, the distance error values ​​of Algorithm 2 fluctuate greatly, especially the distance values ​​of the last three reference points are significantly higher, indicating that the error at certain specific points is large and unstable. Further analysis of the average position error, such as Figure 10 As shown in (b), the error of Algorithm 1 is 0.93 meters, which is 0.21 meters lower than the 1.14 meters of Algorithm 2, and the reduction percentage is about 18.53%, which highlights the advantage of Algorithm 1 in positioning accuracy. Figure 10 As shown in (c), the standard deviation of Algorithm 1 is 0.37 meters, slightly lower than the 0.39 meters of Algorithm 2, a reduction of approximately 6.0%, indicating that Algorithm 1 performs more stably across different reference points. In summary, Algorithm 1 outperforms Algorithm 2 in both average position error and standard deviation, demonstrating its significant advantage in navigation accuracy.

[0123] Table 1 Position error statistics (unit: m)

[0124]

[0125] (2) Verification of the effectiveness of the long-distance variable speed motion algorithm.

[0126] To verify the effectiveness of the algorithm in long-distance movement, the experimental design uses a 400-meter standard runway, with the innermost line of the runway as the precise reference trajectory. The experimenter walks along the white inner edge line, with the speed randomly changing throughout the process, to test the performance of the algorithm under speed changes. In the algorithm application, three thresholds of Algorithm 2 were selected: 50,000, 60,000, and 70,000, and the inertial sensor data were processed using different thresholds of Algorithm 1 and Algorithm 2. As shown in Table 2 and Figure 11 As shown in Figure 2, Algorithm 1 is closer to the real trajectory than Algorithm 2.

[0127] Table 2 Relative position errors of the end points calculated by the two algorithms

[0128]

[0129] To address the problems of traditional zero-speed detection algorithms relying on fixed thresholds, the present invention proposes an interval search adaptive threshold algorithm. The algorithm determines the search interval through SWA filtering, preliminarily determines the zero-speed interval based on statistical characteristics, and then sets a new threshold based on the 3σ criterion to achieve adaptive detection. After verification by short-distance complex speed-varying motion experiments, by comparing Algorithm 1 with Algorithm 2, the results show that the average position error of Algorithm 1 is reduced by about 18.53%, reaching a low level of 0.93 meters, which is a significant improvement compared to Algorithm 2's 1.14 meters. In addition, Algorithm 1 also has a slightly lower standard deviation, indicating that its results are more consistent and reliable under different test conditions. In the 400-meter walking experiment, the relative error of Algorithm 1 is only 0.72%, which is 42.74% lower than the 1.24% of Algorithm 2 at the optimal threshold. This experimental result proves that the proposed algorithm significantly improves positioning accuracy. In summary, the interval search adaptive threshold zero-speed detection algorithm proposed in this paper not only effectively solves the problem of fixed threshold dependence in existing algorithms, but also achieves a significant improvement in navigation accuracy. This research result not only provides new technical ideas for wearable devices, location services and pedestrian navigation, but also lays a solid foundation for subsequent related research and practice.

[0130] The above examples are merely illustrative of the calculation process of the present invention and are not intended to limit the embodiments of the present invention. Persons skilled in the art will readily appreciate that other variations or modifications based on the above description are possible. This list of embodiments is not exhaustive; however, any obvious variations or modifications derived from the technical solution of the present invention remain within the scope of protection of the present invention.

Claims

1. An adaptive zero-speed detection method based on interval search under variable speed motion conditions, characterized in that The following steps are involved: Step 1) Analysis of the proportion of zero-speed intervals Through experimental data analysis, the proportion of zero-speed intervals at different walking speeds is determined, and the minimum proportion of walking zero-speed intervals is set accordingly. This provides a basis for the subsequent preliminary screening of zero-speed intervals and enhances the robustness of the algorithm. Step 2) Gait cycle extraction The angular velocity is synthesized and filtered, and the maximum points of the gait cycle are identified by taking the first-order and second-order derivatives. Based on the intervals and characteristics of the maximum points, representative maximum points are screened to determine the complete gait cycle. Step 3) Preliminary screening of zero-speed interval A mathematical model for zero-speed detection was constructed, and the generalized likelihood ratio test method was used to calculate the detection statistic. Combining the duration of each gait cycle and the minimum proportion of the zero-speed interval set in step 1, the time range of some zero-speed intervals was preliminarily determined, providing a basis for subsequent fine-tuning. Step 4) Fine determination of zero speed range Within the preliminarily determined zero-speed interval, the mean and standard deviation of the detection data are calculated, and the threshold of the gait cycle is adaptively adjusted and set. The zero-speed interval and non-zero-speed interval are judged based on the new threshold to achieve precise determination of the zero-speed interval.

2. The adaptive zero-speed detection method based on interval search under variable speed motion conditions according to claim 1 is characterized in that The specific steps of step 1 are: First, an experiment was conducted to analyze the zero-speed interval at different walking speeds. The analysis showed that the proportion of zero-speed intervals decreases with increasing speed. Pedestrians' daily walking speed does not exceed 8 km / h. Within this speed range, the minimum zero-speed interval proportion is as low as 13.58%, the lowest value among all tested speeds. A robust minimum proportion p of the walking zero-speed interval is determined, that is, the minimum proportion of the walking zero-speed interval is set to 10%.

3. The adaptive zero-speed detection method based on interval search under variable speed motion conditions according to claim 1 is characterized in that The specific steps of step 2 are: To identify a complete gait cycle, the angular velocity is first synthesized to obtain ||ω k ||: Where: ω x (k),ω y (k),ω z (k) are the angular velocity components of the three axes measured by MIMU at time k, ||ω k || is the modulus of the angular velocity at time k, and the modulus of the angular velocity is subjected to SWA filtering: Where: W1 is the window size of SWA, which is set to half of the MIMU output frequency, The angular velocity modulus is obtained after SWA filtering. The first and second order derivatives are calculated to establish the following equation: Obtain n maximum points, denoted as K = {k1, k2, k3......, k n }, where k i Indicates the time corresponding to the i-th maximum point. In order to screen out representative maximum points, the following formula can be used to determine and eliminate pseudo maximum points k i : Among them, index and index-1 are used to represent the first two retained maximum points. If the formula is satisfied, k index Add to the new set L = {l1,l2,l3......,l m }, if it is not satisfied, it means that the two maximum points are close to each other, and k i and k index The corresponding values ​​are compared, if k i >k index , then k i Assign value to k index , if k i ≤k index , then k index Keep unchanged, k index-1 and k index A complete gait cycle is the time between 4. The adaptive zero-speed detection method based on interval search under variable speed motion conditions according to claim 1 is characterized in that The specific steps of step 3 are: The mathematical model of zero speed detection is regarded as a binary hypothesis testing problem, which can be expressed as: Where: T(Z k ) represents the hypothesis detection statistic of pedestrian gait at time k, γ represents the threshold, T(Z k The calculation methods used include acceleration amplitude detection, acceleration variance detection, angular velocity amplitude detection, angular velocity energy detection and generalized likelihood ratio detection. Based on the generalized likelihood ratio detection method, first, define the k-th sampling y of MIMU k , including 3-axis acceleration and 3-axis angular rate: Z n It is a data set consisting of N data samples as sliding windows: Z n ={and n ,and n+1 ,...,and n+N-1 } k moment to Z n Using the generalized likelihood ratio test method, we get the test statistic T(Z k ): Where: T(Z k ) represents the generalized likelihood ratio test statistic at time k, and N is set as a sliding window of 30. and is the variance of acceleration and angular rate noise, is the average acceleration of each axis in the current window, g is the acceleration due to gravity 9.81m / s 2 In order to improve the accuracy of zero-speed detection, based on the generalized likelihood ratio detection algorithm and combined with the duration of each gait cycle, the accurate conservative zero-speed interval time range W2 is preliminarily determined: W2=(k index -k index-1 )p Take the size of W2 as the sliding window and calculate T(Z n ) of T k Perform the calculation: Traverse to find T k The minimum value of k and record the k value min , then [k min ,k min +W2-1] is a zero-speed interval preliminarily determined.

5. The method for adaptive zero-speed detection based on interval search under variable speed motion conditions according to claim 4, characterized in that The specific steps of step 4) are: In the initially determined zero-speed interval, calculate the mean μ and standard deviation σ of the GLR data: By analyzing the data characteristics of the zero-speed interval initially determined in each gait cycle, the threshold γ of the gait cycle is adaptively adjusted and set: Then, the zero speed interval and the non-zero speed interval are determined according to the formula: Wherein N=1 represents the zero speed interval, and N=0 represents the non-zero speed interval.

Images