CN106908060A - A kind of high accuracy indoor orientation method based on MEMS inertial sensor - Google Patents

Abstract

The invention discloses a high-precision indoor positioning method based on a MEMS inertial sensor, which includes the following steps: (1) Fixing the MEMS inertial sensor on the pedestrian's foot, so that the MEMS inertial sensor senses the motion state of the foot, and acquires it in real time. Foot navigation information, and transmit it through Bluetooth; (2) Pedestrians hold Android phones, and receive and save the data provided by MEMS inertial sensors in real time on the Android client; (3) Denoise the data; (4) First use zero The speed detection algorithm is used to obtain the zero speed interval, and then the zero speed correction algorithm is combined with the state estimation algorithm to correct the error; (5) The Android client displays the data after error correction and compensation in real time through the user interface. The invention does not require additional auxiliary basic settings; it can maintain good positioning accuracy for various complex motion states in indoor positioning; it uses the mobile terminal to correct and compensate errors in real time, and displays the walking track in real time through the user interface.

Description

A high-precision indoor positioning method based on MEMS inertial sensors

technical field

The invention relates to an indoor positioning method, in particular to a MEMS inertial sensor-based high-precision indoor positioning method without installing additional infrastructure.

Background technique

Due to the development of the Internet and the popularity of mobile devices and personal devices, LBS (Location Based Services) has become more and more important. Users obtain location information and use it for navigation, tracking, monitoring, information push and other services. GPS can conveniently provide outdoor personal positioning information, but because GPS needs to receive at least 4 satellites to achieve positioning, its positioning effect is greatly affected by the received satellite signals, and in indoor environments, satellite signals cannot be received due to occlusion , seriously affecting its positioning effect.

The inertial navigation system has the advantages of strong autonomy, high output frequency, and high short-term accuracy. Especially in recent years, the rapid development of MEMSIMU has made it small in size and low in cost. Using MEMS IMU indoor positioning to adopt pedestrian dead reckoning (Pedestrian Dead Reckoning, PDR) method can give full play to the advantages of its autonomous navigation. Judging from the current research status of indoor positioning, MEMS indoor positioning has become a major research hotspot in foreign countries. Through the information output by MEMS inertial sensors, first judge the motion state of the carrier, and then perform zero-speed interval detection, and then perform zero-speed correction ( ZeroVelocity Update), which can well suppress the accumulation of MEMS inertial sensor errors. At the same time, the low-cost MEMS IMU gyroscope has low precision and cannot perform initial alignment according to the high-precision inertial navigation system initial alignment method. Because the magnetometer More stable than a gyro, therefore, a magnetometer is introduced to aid in heading alignment and provide heading compensation.

In the process of indoor positioning of MEMS inertial sensors, zero-speed correction is used to correct the errors accumulated in a walking cycle. Therefore, this process is of great significance in improving the positioning accuracy of MEMS inertial sensors. The detection directly determines the accuracy of the zero-speed correction. Most of the detection methods for the zero-speed interval of the data found at home and abroad are to set a threshold and compare the corresponding detection quantity with the threshold. However, the selection of this threshold is often one-sided. It can only target a specific motion state, but indoor footsteps may have various complex motion states, such as walking, running, going upstairs, going downstairs, etc., so the detection accuracy needs to be improved urgently.

Contents of the invention

Purpose of the invention: In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a high-precision indoor positioning method based on MEMS inertial sensors that is suitable for autonomous positioning in indoor environments and does not require additional installation of infrastructure.

Technical scheme: In order to realize the purpose of the present invention, adopt following technical scheme: A kind of based on MEMS inertial sensor

The high-precision indoor positioning method comprises the following steps:

(1) The MEMS inertial sensor is fixedly connected to the pedestrian's foot, so that the MEMS inertial sensor senses the motion state of the foot, and obtains the navigation information of the foot in real time, and the navigation information is transmitted through the Bluetooth transmission module;

(2) Pedestrians hold Android mobile phones, receive the data provided by the MEMS inertial sensor for navigating the foot movement state in real time on the Android client, and save the data;

(3) Denoise the data;

(4) To solve the problem of low positioning accuracy of MEMS inertial sensors, first use the zero-speed detection algorithm to obtain the zero-speed interval, and then use the zero-speed correction algorithm combined with the state estimation algorithm to correct the error and improve the positioning accuracy; among them, the zero-speed interval detection is through Android The client obtains the data transmitted by the MEMS inertial sensor, and uses the sliding window method to detect the zero-speed interval. The specific steps are as follows:

(a) According to the noise characteristics of the accelerometer and gyroscope of the MEMS inertial sensor, repeated experiments are carried out under different motion states to obtain the threshold value of the zero-speed interval section under the corresponding motion state, and the different motion states include normal walking, running, going up or down stairs;

(b) According to the acceleration information of the accelerometer obtained by the Android client, firstly, by analyzing the characteristics of the acceleration in the sliding window, comparing the changes in the acceleration at adjacent moments, and the maximum and minimum values of the acceleration in the window, determine the current position of the footsteps. It is in the state of motion, and automatically selects the threshold value of zero speed detection corresponding to the state of motion;

(c) Determine whether the specific force modulus, the specific force variance, and the angular velocity modulus are within the threshold range in the sliding window interval. If the three meet the threshold conditions at the same time, it can be concluded that the interval is a zero velocity interval. After obtaining the zero speed range, this range is the range where the foot is at zero speed;

At this time, the zero-speed correction combined with the adaptive Kalman filter algorithm is used to eliminate the accumulated error in a walk interval.

(5) The Android client displays the data after error correction and compensation in real time through the user interface display module, and presents the motion state to the user in real time.

The data provided by the MEMS inertial sensor includes data measured by a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer. The Android client has the function of calibrating the above data, and can record data and process data in real time.

The Android client has the function of displaying the user's walking track in real time.

In step (3), aiming at the noise problem generated by the MEMS inertial sensor when the foot is shaken violently, the denoising processing method is wavelet denoising, which can remove noise and interference signals contained in high-frequency signals.

In step (4), for the problem that the error of the low-cost inertial sensor is easy to diverge with time, the state estimation algorithm is an adaptive Kalman filter algorithm, and the steps of error correction using the zero-speed correction algorithm in combination with the state estimation algorithm are as follows:

A more accurate zero-speed interval can be obtained through the aforementioned zero-speed detection algorithm. In the process of the zero-speed interval, the speed is zero, so the speed output by the navigation solution is the speed error. The relationship between speed error and gyro drift, acceleration zero bias and attitude angle error can be described by error differential equation. On the basis of the error equation, the Kalman filter equation is established, and the velocity error is used as an observation, and the attitude angle error, position error and other error quantities are estimated and fed back to the navigation solution module, so as to achieve the purpose of correcting the system error.

Attitude Angle Error Equation

speed error equation

position error equation

On the basis of the error equation, the Kalman filter equation is established, and the velocity error is used as an observation, and the attitude angle error, position error and other differences are estimated and fed back to the navigation solution module.

The state quantity of the selected Kalman filter is:

in Indicates the attitude angle error, δv ⁿ indicates the velocity error; δp ⁿ indicates the position error; ε indicates the gyro constant value drift, Indicates the constant zero bias of the accelerometer, and the Kalman system equation is expressed as

X(t)=FX(t)+W(t)

The velocity error of the measurement equation, that is,

Z _k ＝[δv ⁿ ]＝H _k X _k

The quantity measurement can only be obtained in the zero-speed interval, and the quantity measurement is related to the specific movement and interference. After the zero-speed interval is detected by the above-mentioned zero-speed detection algorithm, the Kalman filter performs time update and measurement update, and estimates The error is fed back to the system for error compensation.

The MEMS inertial sensor and the bluetooth module adopt a patch-type hardware design method, which has the advantage of small size.

Beneficial effects: the present invention uses a low-cost MEMS inertial sensor based on the human walking dynamics model, which has the advantages of small size and light weight. Compared with the wireless positioning method, the system structure can be realized without additional auxiliary basic settings Autonomous positioning; the present invention adopts an adaptive threshold matching method, which can maintain good positioning accuracy for various complex footsteps in the indoor positioning process, and has strong popularity; for the current popular mobile terminals, the design The mobile terminal receives the measurement data of the inertial sensor transmitted by Bluetooth in real time, and has the function of saving and processing the data, correcting and compensating the data in real time, and designing a user interface to display the user's walking track in real time.

Description of drawings

Figure 1 is a structural diagram of the MEMS inertial sensor indoor positioning system;

Fig. 2 is the curve before denoising of MEMS inertial sensor and after denoising and smoothing;

Fig. 3 is a flow chart of a zero-speed interval detection method;

Figure 4 is a detection effect diagram of the zero-speed interval;

Figure 5 is the Android client interface;

Fig. 6 is a two-dimensional plane walking trajectory diagram of the present invention;

Fig. 7 is a three-dimensional walking trajectory diagram of going up and down stairs according to the present invention.

detailed description

The technical solution of the present invention will be further described in detail below in conjunction with the embodiments and the accompanying drawings.

As shown in Figure 1, the high-precision indoor positioning equipment used in the present invention includes: low-cost MEMS inertial sensors such as MPU9260, MPU6050, etc., Bluetooth transmission module, android client; wherein, the android client includes a data calibration module, a data storage module, A data processing module and a user interface display module. The MEMS inertial sensor is firmly connected to the foot, sensitive to the change of the footstep state, and transmits the measured parameters through the Bluetooth transmission module, and receives the data transmitted by the Bluetooth transmission module through the Android client. The Android client first calibrates the three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer through the data calibration module, and displays the data on the user interface in real time while saving and processing the data in real time. Both the MEMS inertial sensor and the Bluetooth module mentioned above adopt a patch-type hardware design method, which has the advantages of compact structure and small volume.

The specific implementation steps are as follows:

(1) Fix the low-cost inertial sensor on the foot to make it sensitive to the motion state of the foot. The MEMS inertial sensor can normally output the information of the three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer. Obtain the navigation information of the foot in real time, and transmit the navigation information through the Bluetooth transmission module.

(2) Pedestrians hold Android mobile phones and receive data from foot inertial sensors in real time on the Android client. First, open the data calibration module of the Android client to measure the three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer respectively. The main purpose of calibration is to remove the zero bias of the accelerometer, gyroscope, and magnetometer. The calibration method of the accelerometer and gyroscope is to set the MEMS inertial sensor to the level, and the zero bias displayed by the module to be calibrated will no longer change. At this time, the zero offset at this time is saved, that is, the calibration of the accelerometer and gyroscope is completed. For the calibration of the magnetic field, the MEMS inertial sensor is rotated repeatedly around different directions, and the movement is repeated until the zero offset of the magnetic field is displayed. Change again to complete the calibration of the magnetic field. After the calibration of the data is completed, check whether the accelerometer and gyroscope are zero when the MEMS inertial sensor is in a static state, and save the data.

(3) Perform wavelet denoising processing on the obtained data, as shown in FIG. 2 , which is a comparison chart of the data measured by the MEMS inertial sensor and the data after wavelet denoising.

(4) The method of sliding window is used for the data of MEMS inertial sensor to detect the zero-speed interval, and the detection method of the zero-speed interval is shown in Fig. 3 .

The zero-speed detection algorithm plays an important role in the whole MEMS positioning process and directly determines the positioning accuracy. The present invention adopts a detection algorithm that first identifies and then judges. The identification method is as follows:

Calculate the triaxial acceleration modulus at each sampling moment i:

Select a sliding window with a size of 2N+1, and N represents the window width, then the acceleration variance in the window is:

in,

The acceleration data is scanned at regular intervals, and the difference between the maximum and minimum values of the scanned acceleration can be used to identify whether the pace during this period is normal walking, slow, fast, or going up and down stairs. The test determines that the acceleration thresholds of normal walking, slow speed, fast or going up and down stairs are respectively th _a1 , th _a2 , th _a3 , th _a4 ; the angular velocity thresholds are respectively th _w1 , th _w2 , th _w3 , th _w4 ; the covariance of acceleration th _σ1 , th _σ2 , th _σ3 , th _σ4 .

The above motion state identification method mainly based on the acceleration modulus value can also be identified based on the angular velocity or acceleration covariance method. After the motion state is identified, the zero-speed interval detection is performed. In order to improve the detection accuracy, according to the motion state identified above , relying on a single element for detection will inevitably have errors. This embodiment uses a combination of the three methods for detection:

Condition 1: proportional modulus When the footsteps are at zero speed, the acceleration output measured by the MEMS inertial sensor fixed on the feet is theoretically only the local gravity acceleration value, and the acceleration conforms to the chi-square distribution. You only need to set an appropriate confidence interval to make a judgment. suppose are the minimum and maximum values of the confidence interval, respectively;

Among them, 1 indicates that the detection is in a static state, and 0 indicates a motion state.

Condition two: the variance of the ratio. Observing the specific force variance with a sliding window can effectively detect the abrupt phase of the signal in the gait cycle. In the zero-speed range, the variance of the ratio is relatively weak, with almost no fluctuations, while in the non-zero-speed range, the footstep changes are highly volatile, so the zero-speed range can be judged by comparing the variance of the force with less than a given threshold. Calculated as follows:

in, is the mean value of the ratio; m is the size of the sliding window, which is related to the output frequency;

Among them, 1 indicates that the detection is in a static state, and 0 indicates a motion state.

Condition 3: Angular velocity modulus. Since the bottom of the foot is in full contact with the ground in the zero-speed zone, the angular velocity and its change value should tend to zero in the zero-speed zone, while the angular velocity will show strong fluctuations outside the zero-speed zone, so The zero-speed interval can be determined by determining the corresponding angular velocity threshold th _w through the identified walking state, where the angular velocity modulus is

Among them, 1 indicates that the detection is in a static state, and 0 indicates a motion state.

When condition 1, condition 2, and condition 3 are met at the same time, record the state at this time as 1, which is the zero-speed interval that needs to be detected, and if it is 0, it indicates the motion state.

FIG. 4 shows the zero-speed interval detected by the above-mentioned zero-speed interval detection method.

If the zero-speed interval is detected by the above method, the error accumulated by the MEMS inertial sensor within a walking cycle must be corrected in the zero-speed interval to compensate the entire positioning system.

A more accurate zero-speed interval can be obtained through the above-mentioned zero-speed detection algorithm. In the process of the zero-speed interval, the speed is zero, so the speed output by the navigation solution is the speed error. The relationship between speed error and gyro drift, acceleration zero bias and attitude angle error can be described by error differential equation. On the basis of the error equation, the Kalman filter equation can be established, and the velocity error is used as the observation quantity to estimate the attitude angle error, position error and other error quantities to be fed back to the navigation calculation module, so as to achieve the purpose of correcting the system error.

(a) Attitude error equation

(b) Speed error

(c) Position error equation

The state quantity of the selected Kalman filter is:

in Indicates attitude error angle, δv ⁿ indicates velocity error; δp ⁿ indicates position error; ε indicates gyro constant value drift, Indicates the constant zero bias of the accelerometer. In theory, as many factors as possible should be considered. The more factors considered, the higher the accuracy of the navigation system estimation, but this also increases the order of the system model ^[11] , Kalman The system equation of can be expressed as

X(t)=FX(t)+W(t)

The velocity error of the measurement equation, that is,

Z _k ＝[δv ⁿ ]＝H _k X _k

The quantity measurement δv _k can only be obtained in the zero-speed interval, and the quantity measurement is related to the specific movement and interference. After the zero-speed interval is detected by the above-mentioned zero-speed detection algorithm, the Kalman filter performs time update and measurement update, and Feedback the estimated error to the system for error compensation

(5) Display the corrected data in real time on the user interface display module, so that users can clearly see their own exercise status.

The present invention is an independent high-precision indoor positioning method, and the beneficial effects of the present invention will be verified through some experiments below.

In order to verify the effectiveness of this method, the normal two-dimensional plane walking experiment and the three-dimensional up-and-down stairs walking experiment were carried out with the low-cost MEMS MPU9250 respectively. The data transmitted by the MEMS sensor through Bluetooth is received in real time by a handheld Android mobile phone. The Android client interface of the mobile phone is as follows: Figure 5 shows.

Experiment 1: Two-dimensional plane normal walking experiment

The experimental environment is selected in the corridor of the foreign student building of Southeast University. The corridor is in the shape of a rectangle with a length of 5m and a width of 1m.

Fig. 6 is the walking trajectory diagram after adopting the algorithm of the present invention to the data received by the Android client, under the ideal situation without error, the walking trajectory is a closed rectangular curve, and the starting point and the end point are coincident, but due to the existence of the error , the starting point and the ending point are often not coincident. The positioning accuracy is determined by the degree of misalignment between the two, and the positioning accuracy of the two-dimensional walking experiment is within 0.5m.

Experiment 2: Three-dimensional walking up and down stairs

The experimental environment is selected in the center building of Southeast University. The starting point is the 2nd floor of the center building. The walking route is from the 2nd floor of the center building to the right staircase, passing through the 3rd floor to the 4th floor of the center building, and then turning left from the corridor on the 4th floor to the left side of the center building. The stairs go down through the 3rd floor to the 2nd floor and return to the starting point of the 2nd floor, also walking a closed route. During this walking process, going up stairs, normal walking, going down stairs, and normal walking have passed.

Fig. 7 is the walking track figure after adopting the algorithm of the present invention to process the data obtained by the Android client, and its positioning error is calculated within 2m.

Through experimental analysis, the present invention can maintain high positioning accuracy for normal two-dimensional plane walking, and still has high positioning accuracy for complex movement states of going up and down stairs. The corrected data is presented to the user through the user interface display module of the Android client, which can realize the autonomous navigation of the user in various unknown environments.

Claims (6)

1. A high-precision indoor positioning method based on an MEMS inertial sensor is characterized by comprising the following steps:

(1) the MEMS inertial sensor is fixedly connected to the foot of the pedestrian, so that the MEMS inertial sensor senses the motion state of the foot and acquires navigation information of the foot in real time, and the navigation information is transmitted through the Bluetooth transmission module;

(2) the pedestrian holds the android mobile phone, receives data provided by the MEMS inertial sensor for navigating the foot motion state in real time at an android client, and stores the data;

(3) denoising the data;

(4) firstly, obtaining a zero-speed interval by adopting a zero-speed detection algorithm, and secondly, carrying out error correction by adopting a zero-speed correction algorithm in combination with a state estimation algorithm; the zero-speed interval detection method comprises the following steps of acquiring data transmitted by the MEMS inertial sensor through an android client, and detecting the zero-speed interval by adopting a sliding window method:

(a) according to the noise characteristics of an accelerometer and a gyroscope of the MEMS inertial sensor, repeated experiments are carried out in different motion states to obtain zero-speed interval thresholds in the corresponding motion states, wherein the different motion states comprise normal walking, running, climbing stairs or descending stairs;

(b) according to the acceleration information of the accelerometer acquired by the android client, firstly, the motion state of the footstep at the moment is determined by analyzing the characteristics of the acceleration in the sliding window and comparing the change of the acceleration at the adjacent moments and the conditions of the maximum value and the minimum value of the acceleration in the window, and the threshold value of the zero-speed detection corresponding to the motion state is automatically selected;

(c) respectively judging whether the specific force module value, the specific force variance and the angular velocity module value are within a threshold range in the sliding window interval, if the specific force module value, the specific force variance and the angular velocity module value simultaneously meet a threshold condition, judging that the interval is a zero-velocity interval, and after obtaining the zero-velocity interval, determining that the interval is an interval in which the foot is at zero velocity;

(5) and the android client displays the data subjected to error correction and compensation in real time through the user interface display module.

2. The high-precision indoor positioning method based on the MEMS inertial sensor of claim 1, wherein: the data provided by the MEMS inertial sensor comprises data measured by a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer.

3. The high-precision indoor positioning method based on the MEMS inertial sensor of claim 1, wherein: the android client has the function of displaying the walking track of the user in real time.

4. The high-precision indoor positioning method based on the MEMS inertial sensor of claim 1, wherein: in the step (3), the denoising processing method is wavelet denoising.

5. The high-precision indoor positioning method based on the MEMS inertial sensor of claim 1, wherein: in the step (4), the state estimation algorithm is an adaptive kalman filtering algorithm, and the step of performing error correction by adopting a zero-speed correction algorithm and combining the adaptive kalman filtering algorithm is as follows:

and (3) describing the relationship between the speed error and the gyro drift, the acceleration zero offset and the attitude angle error by using an error differential equation:

error equation of attitude angle

Equation of speed error

Equation of position error

$\mathrm{\δ}{\stackrel{\·}{p}}^n={\mathrm{\δv}}^n$

Establishing a Kalman filtering equation on the basis of an error equation, taking a speed error as an observed quantity, estimating an attitude angle error, a position error, a gyro drift and an acceleration zero offset, feeding back to a navigation resolving module,

selecting the state quantity of the Kalman filter as follows:

wherein the attitude angle error, v, is expressedⁿIndicating a speed error; p is a radical ofⁿIndicating a position error; the gyro is represented by a constant drift of the gyro,the system equation representing the accelerometer constant zero offset, Kalman is expressed as

X(t)＝FX(t)+W(t)

Measuring velocity errors of equations, i.e.

Z_k＝[vⁿ]＝H_kX_k

And after the zero-speed interval is detected, the Kalman filtering updates time and measurement, and feeds the estimated error back to the system for error compensation.

6. The high-precision indoor positioning method based on the MEMS inertial sensor of claim 1, wherein: the MEMS inertial sensor and the Bluetooth module adopt a patch type hardware design method.