CN110567492A - System-level calibration method for low-cost MEMS inertial sensors - Google Patents

Abstract

the invention discloses a low-cost MEMS inertial sensor system-level calibration method, which comprises the following steps: determining a sensor to-be-marked parameter based on the regression model; decoupling a gyroscope error coupling term in a speed error differential equation to obtain an inertial navigation error equation; establishing a Kalman filter according to the determined sensor to-be-marked parameter and the inertial navigation error equation; exciting each output shaft of the sensor according to a set motion track, and collecting the original data of the sensor; the original data is subjected to wavelet threshold denoising and then is led into the Kalman filter, so that the to-be-calibrated parameters are estimated; and substituting the estimated parameter to be calibrated into an inertial navigation error differential equation to verify whether the calibration result is accurate. The method has the advantage of improving the calibration accuracy of the low-cost MEMS inertial measurement.

Description

System-level calibration method for low-cost MEMS inertial sensors

technical field

The invention relates to the field of inertial guidance, in particular to a system-level calibration method for a low-cost MEMS inertial sensor.

Background technique

Inertial sensors based on Micro-Electro-Mechanical Systems (MEMS) have the characteristics of small size, low cost, and low power consumption, so they are widely used in various fields. Due to the immature research and analysis on the error characteristics of MEMS inertial sensors, the error items such as scale error, installation error, and zero bias of the sensor cannot be accurately modeled. There are differences, so the error parameters of the MEMS inertial sensor need to be calibrated again after a certain period of use. There are many error items in the MEMS inertial sensor, and the establishment and analysis of the error model is also relatively complicated. Experts and scholars at home and abroad have conducted a lot of analysis and research on this. One method uses the analytical method to establish a mathematical model of the dynamic error of the gyroscope, and analyzes the causes and effects of each error; the other method uses the radial basis function (RBF) neural network to adjust the scale factor of the gyroscope. The nonlinear coupling error is compensated to improve the average precision of the gyroscope output. For the selection of inertial sensor calibration schemes, Zhou Yanglin et al. introduced three-dimensional laser scanning measurement technology in the calibration process to achieve accurate estimation of high-precision placement parameters between sensors; another method is the hybrid inertial navigation system, and uses its own The rotating mechanism completes the calibration of the error parameters. In another method, the decoupling method of the mean value distribution is used to decouple the influence of the oblique symmetric error of the accelerometer and the gyroscope on the error equation to improve the calibration accuracy.

Common calibration methods are divided into two types: discrete calibration method and system calibration method. The discrete calibration method has high requirements on the precision of the turntable, resulting in high calibration costs. At present, the system calibration method is commonly used to complete the estimation of the parameters to be calibrated. The system calibration method takes the navigation error of the integrated navigation system as the observation quantity, expands the error parameters into the system state, and then uses the filtering algorithm to estimate and update. Compared with the discrete calibration method, the system calibration method needs to arrange the motion trajectory of the device to be marked reasonably to ensure that each error parameter to be marked is excited, so that the state quantity output by the integrated navigation system can be observed to complete the parameter estimation to be marked. One of the inertial measurement unit (Inertial Measurement Unit, IMU) 24-position continuous stop calibration scheme, through in-situ calibration to identify 15 error parameters of accelerometer and gyroscope zero bias, scale factor and accelerometer non-orthogonal error ; In another scheme, the six-position stationary and six-position rotating calibration test methods are used to directly solve and calibrate the theoretical value of the sensitive axis, and verify the effectiveness of this method for static error convergence; another scheme proposes A multi-position continuous rotation calibration scheme estimates all 21 error parameters by measuring the speed error of each position under static navigation state, and conducts multiple experiments to verify its effectiveness.

The above research results mostly focus on the calibration of inertial measurement units composed of high-precision laser gyroscopes or fiber optic gyroscopes and accelerations, and seldom involve low-precision low-cost MEMS inertial measurement, so there is a problem of poor calibration accuracy of low-cost MEMS inertial measurement .

Contents of the invention

The object of the present invention is to propose a system-level calibration method for low-cost MEMS inertial sensors to achieve the advantage of improving the calibration accuracy of low-cost MEMS inertial sensors.

In order to achieve the above object, the technical solution adopted in the embodiment of the present invention is:

A low-cost MEMS inertial sensor system-level calibration method, including:

Determine the sensor parameters to be marked based on the regression model;

The inertial navigation error equation is obtained by decoupling the gyroscope error coupling term in the velocity error differential equation;

Establishing a Kalman filter according to the determined parameters of the sensor to be marked and the inertial navigation error equation;

Excite each output shaft of the sensor according to the set motion trajectory, and collect the original data of the sensor;

Importing the original data into the Kalman filter after wavelet threshold noise reduction, so as to estimate the parameters to be marked;

Substitute the estimated parameters to be calibrated into the inertial navigation error differential equation to verify whether the calibration results are accurate.

As a specific implementation of the embodiment of the present invention, the determination of the parameters to be marked of the sensor based on the regression model includes:

Collect the output data when the sensor is in the set position;

A regression model is built based on the output data.

As a specific implementation manner of the embodiment of the present invention, the output data of the collecting sensor at a specific position includes: three-axis accelerometer and three-axis gyroscope data.

As a specific implementation of the embodiment of the present invention, the regression model includes an accelerometer regression model and a gyroscope regression model;

The accelerometer regression model is:

Among them, a, b, c are parameters to be estimated, Indicates the single-axis output value of the accelerometer, and f indicates the actual acceleration value sensed by the sensitive axis of the accelerometer;

The gyroscope regression model is:

Among them, m and n are parameters to be estimated, Indicates the single-axis output value of the gyroscope, and ω indicates the angular velocity value actually sensed by the sensitive axis of the gyroscope.

As a specific implementation of the embodiment of the present invention, the set position includes: the set position of the accelerometer and the set position of the gyroscope;

The set position of the accelerometer includes: the x-axis faces north, the y-axis faces west, and the z-axis faces the sky; or the x-axis faces north, the y-axis faces the ground, and the z-axis faces west; or the x-axis faces north, and the y-axis faces east , the z-axis is facing the ground; or the x-axis is facing north, the y-axis is facing the sky, and the z-axis is facing east; or the x-axis is facing north, the y-axis is facing west, and the z-axis is facing the sky;

Static collection of data at each position for 10 minutes;

The set position of the gyroscope includes: the position where the sensor rotates counterclockwise around the x-axis of the gyroscope every 10 minutes, and the rotation speed is 20°/s.

As a specific implementation of the embodiments of the present invention, the parameters to be marked include:

Three-axis accelerometer zero-offset matrix, three-axis accelerometer installation error matrix _{δL Sym/Sksym ,} three-axis accelerometer scale coefficient error matrix δL Scal , three-axis gyroscope zero-offset matrix ε ^b , three-axis gyroscope installation error matrix δK _{Sym /Sksym} and/or the three-axis gyroscope scale coefficient error matrix δK _Scal , wherein the three-axis accelerometer installation error matrix and the three-axis gyroscope installation error matrix both include a symmetric error matrix and a skew symmetric error matrix.

As a specific implementation of the embodiment of the present invention, the decoupling of the gyroscope error coupling term in the velocity error differential equation to obtain the inertial navigation error equation includes:

The definition of the carrier coordinate system coincides with I+δK _Sksym , so that the gyroscope's obliquely symmetric error matrix is a zero matrix, thereby eliminating the coupling term due to the gyroscope's obliquely symmetric error in the velocity error differential equation, and I represents the identity matrix.

As a specific implementation of the embodiment of the present invention, the inertial navigation error differential equation is:

The inertial navigation error differential equation includes a velocity error differential equation and a misalignment angle differential equation,

Among them, ψ represents the misalignment angle vector, δv is the velocity error vector, δf ^b is the accelerometer error, f is the accelerometer measurement value, is the gyroscope error, is the angular velocity of the earth's rotation, is the angular velocity involved from the inertial coordinate system to the navigation coordinate system, is the attitude transfer matrix from the carrier coordinate system to the inertial coordinate system.

As a specific implementation of the embodiment of the present invention, the establishment of the Kalman filter includes:

Set the state quantity;

Establishing a state transition matrix based on the state quantity;

A measurement matrix of the state transition matrix is determined.

As a specific implementation of the embodiment of the present invention, the wavelet threshold noise reduction includes:

parameter settings;

Perform wavelet decomposition based on the set parameters;

Screening the wavelet coefficients in the wavelet decomposition by a threshold function to complete signal denoising;

Perform wavelet reconstruction based on the denoised signal;

Wherein, the threshold function is:

In the formula, ω _{j, k} are decomposed wavelet coefficients, k and α are adjustment coefficients, and both are always positive, T is the set threshold, the formula is as follows:

In the formula, N is the signal length, σ is the noise standard deviation, and j is the number of decomposition layers.

Embodiments of the present invention have the following beneficial effects:

In the embodiment of the present invention, the regression model is used to determine the parameters to be marked in the sensor error model, and then through the coupling term in the velocity error differential equation, the identification of the sensor error parameters is realized by establishing a Kalman filter, and then the collected data is passed through the wavelet threshold After noise reduction, import the Kalman filter to obtain the estimated parameters to be calibrated, and then substitute the estimated parameters to be calibrated into the differential equation of inertial navigation error to verify the accuracy of the calibration results, so as to improve the accuracy of low-cost MEMS inertial measurement calibration degree purpose.

The technical solutions of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

Description of drawings

Fig. 1 is the flowchart of the low-cost MEMS inertial sensor system-level calibration method described in the embodiment of the present invention;

Fig. 2a to Fig. 2b are the comparison charts of accelerometer regression model verification described in the embodiment of the present invention;

Fig. 3 is the gyroscope regression model verification comparison diagram described in the embodiment of the present invention;

Fig. 4a to Fig. 4b are comparison diagrams of noise reduction effect of sensor raw data according to the embodiment of the present invention;

Fig. 5a to Fig. 5b are analysis diagrams of noise reduction performance of sensor raw data according to the embodiment of the present invention;

Figures 6a to 6c are comparison diagrams of posture verification of calibration results described in the embodiment of the present invention;

Fig. 7a to Fig. 7c are comparison diagrams of the speed verification of the calibration results described in the embodiment of the present invention.

Detailed ways

The preferred embodiments of the present invention will be described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described here are only used to illustrate and explain the present invention, and are not intended to limit the present invention.

As shown in Figure 1, a low-cost MEMS inertial sensor system-level calibration method includes:

S101: Determine the parameters to be marked of the sensor based on the regression model;

S102: Decoupling the gyroscope error coupling term in the velocity error differential equation to obtain an inertial navigation error equation;

S103: Establish a Kalman filter according to the determined parameters of the sensor to be marked and the inertial navigation error equation;

S104: excite each output shaft of the sensor according to the set motion trajectory, and collect the original data of the sensor;

S105: Import the original data into the Kalman filter after wavelet threshold denoising, so as to estimate the parameters to be marked;

S106: Substitute the estimated parameters to be calibrated into the inertial navigation error differential equation to verify whether the calibration result is accurate.

In an optional embodiment, the parameters to be marked of the sensor are determined based on the regression model, including:

Collect the output data when the sensor is in the set position;

A regression model is built based on the output data.

In an optional embodiment, in the output data collected when the sensor is at a specific position, the output data includes: data of a three-axis accelerometer and a three-axis gyroscope.

In an optional embodiment, the regression model includes an accelerometer regression model and a gyroscope regression model;

The accelerometer regression model is:

Among them, a, b, c are parameters to be estimated, Indicates the single-axis output value of the accelerometer, and f indicates the actual acceleration value sensed by the sensitive axis of the accelerometer;

The gyroscope regression model is:

Among them, m and n are parameters to be estimated, Indicates the single-axis output value of the gyroscope, and ω indicates the angular velocity value actually sensed by the sensitive axis of the gyroscope.

In an optional embodiment, the set position includes: an accelerometer set position and a gyroscope set position;

The set position of the accelerometer includes: the x-axis faces north, the y-axis faces west, and the z-axis faces the sky; or the x-axis faces north, the y-axis faces the ground, and the z-axis faces west; or the x-axis faces north, and the y-axis faces east , the z-axis is facing the ground; or the x-axis is facing north, the y-axis is facing the sky, and the z-axis is facing east; or the x-axis is facing north, the y-axis is facing west, and the z-axis is facing the sky;

Static collection of data at each position for 10 minutes;

The set position of the gyroscope includes: the position where the sensor rotates counterclockwise around the x-axis of the gyroscope every 10 minutes, and the rotation speed is 20°/s.

In a specific application scenario, the sensor rotates counterclockwise around the x-axis of the gyroscope to collect data every 10 minutes.

Among them, both the accelerometer sensor and the gyroscope sensor complete the parameter estimation by the least square method

In an optional embodiment, the parameters to be marked include:

Three-axis accelerometer zero-offset matrix, three-axis accelerometer installation error matrix _{δL Sym/Sksym ,} three-axis accelerometer scale coefficient error matrix δL Scal , three-axis gyroscope zero-offset matrix ε ^b , three-axis gyroscope installation error matrix δK _{Sym /Sksym} and/or the three-axis gyroscope scale coefficient error matrix δK _Scal , wherein the three-axis accelerometer installation error matrix and the three-axis gyroscope installation error matrix both include a symmetric error matrix and a skew symmetric error matrix.

The installation error matrix of the three-axis accelerometer and the installation error matrix of the three-axis gyroscope are as follows:

In the formula, the matrix with the subscript Scal represents the scale coefficient error matrix, the matrix with the subscript Sym represents the symmetric error matrix in the installation error, and the matrix with the subscript Sksym represents the skew symmetric error matrix.

In an optional embodiment, the decoupling of the gyroscope error coupling term in the speed error differential equation to obtain the inertial navigation error equation includes:

The definition of the carrier coordinate system coincides with I+δK _Sksym , so that the gyroscope's obliquely symmetric error matrix is a zero matrix, thereby eliminating the coupling term due to the gyroscope's obliquely symmetric error in the velocity error differential equation, and I represents the identity matrix.

In an optional embodiment, the differential equation of the inertial navigation error is:

The inertial navigation error differential equation includes a velocity error differential equation and a misalignment angle differential equation,

is the velocity error differential equation, is the misalignment angle differential equation.

Among them, ψ represents the misalignment angle vector, δv is the velocity error vector, δf ^b is the accelerometer error, f is the accelerometer measurement value, is the gyroscope error, is the angular velocity of the earth's rotation, is the angular velocity involved from the inertial coordinate system to the navigation coordinate system, is the attitude transfer matrix from the carrier coordinate system to the inertial coordinate system.

In an optional embodiment, the establishment of a Kalman filter includes:

Set the state quantity;

Establishing a state transition matrix based on the state quantity;

A measurement matrix of the state transition matrix is determined.

In specific application scenarios,

state quantity

Among them, ψ represents the misalignment angle vector, δv is the velocity error vector, ε ^T and ▽ ^T are the zero bias of the gyroscope and accelerometer, respectively, s ^ω and s ^a are the scale coefficient errors of the sensor, and k ^Sym is the installation error of the gyroscope Vector, l ^Sym/Sksym is the accelerometer installation error vector.

State transition matrix:

Measurement matrix:

H=[0 _3×3 I 0 _3×21 ],

where the quantity measure is velocity error.

In a specific application scenario, the output shafts of the sensor are excited according to the set motion trajectory, and the raw data of the sensor is collected,

The set motion trajectory is shown in Table 1, which uses the 19-position method proposed in the multi-position calibration method of the laser gyro strapdown inertial navigation system in the literature.

Table 1: Motion track layout table.

In an optional embodiment, the wavelet threshold noise reduction includes:

parameter settings;

Perform wavelet decomposition based on the set parameters;

Screening the wavelet coefficients in the wavelet decomposition by a threshold function to complete signal denoising;

Wavelet reconstruction is performed based on the denoised signal.

In specific application scenarios:

Parameter setting initialization:

The number of decomposition layers of wavelet threshold denoising is set to 4, the gyroscope wavelet base is selected as: sym4 (x-axis), coif4 (y-axis, z-axis); the accelerometer wavelet base is selected as: dmey (x-axis, y-axis), sym6(z axis); threshold function parameter k is set to 1;

⑵Wavelet decomposition, the decomposition equation is:

where ω _j,k is the decomposed wavelet coefficient, ψ _j,k is the wavelet system, f is the signal to be decomposed, a is the scale parameter, b is the translation parameter, and the discrete wavelet system function ψ _j,k can be written as :

The wavelet coefficients are screened through the threshold function to complete signal denoising. The threshold function is:

In the formula, k and α are the adjustment coefficients, and both are always positive, T is the set threshold, the formula is as follows:

In the formula, N is the signal length, σ is the noise standard deviation, and j is the number of decomposition layers.

Wavelet reconstruction restores useful signals from the wavelet coefficients processed by the threshold function. The formula for wavelet reconstruction is:

X is the signal after noise reduction processing, and C is a constant that has nothing to do with the signal.

In an optional embodiment, when verifying whether the calibration result is accurate, the data of the sensor in a static state for 10 minutes is collected for verification.

Figure 2a and Figure 2b are comparison diagrams of accelerometer regression model verification, and the comparison objects are the fitting model of the first term and the fitting model of the second term. The comparison results show that both models are valid for the accelerometer output. But in places where the rate of acceleration change is small, the fitting model of the quadratic term is better than the fitting model of the first term, and in places where the acceleration changes are more severe, the fitting accuracy of the fitting model of the first term is slightly better than that of the quadratic term Fit the model.

Figure 3 is a verification comparison chart of the gyroscope regression model, and the model used is a one-time item fitting model. The output mean value of the fitting data is basically consistent with the test data, and both are close to the set speed value of the turntable, so the output model can be considered to be correct for the micromachined gyroscope used in the test.

Figure 4a and Figure 4b are comparison diagrams of sensor raw data noise reduction effects. The noise reduction process is: ①parameter initialization; ②wavelet decomposition; ③calculation of threshold function; ④threshold screening; ⑤signal reconstruction. It can be seen from Figure 4a and Figure 4b that the raw data output by the low-cost MEMS sensor contains a lot of noise components.

Figure 5a and Figure 5b are the noise reduction performance analysis diagrams of the original data of the sensor. The power density of the output information of the axial gyroscope is distributed from 0 Hz to 2.75 Hz, and there is no obvious peak value, so the main component of the output information is the whole frequency band. White noise, after the wavelet noise reduction processing, there is an obvious peak in the power density, which proves that the wavelet noise reduction has basically filtered the white noise signal; the proportion of the white noise signal in the output information of the three-axis accelerometer is small, and after After noise processing, the power spectral density is further concentrated to the low frequency band, and the proportion of effective signals increases.

Figures 6a to 6c and Figures 7a to 7c are the comparison diagrams of velocity error and attitude error in three directions respectively, where the traditional calibration method without nonlinear factors is method 1, and the matrix decomposition method without second-order nonlinear factors is Method 2. The traditional calibration method containing second-order nonlinear factors is method 3. The matrix decoupling-based calibration method containing second-order nonlinear factors proposed by the present invention is method 4. Since the matrix decomposition process is mainly to eliminate the coupling item of acceleration and gyroscope in the speed error model, and the attitude angle error is only related to the gyroscope, method 1 and method 3 have limited influence on the attitude angle compared with method 2 and method 4. However, compared with before calibration, the four methods can effectively suppress the divergence of attitude angle errors, especially the suppression effect on the divergence of heading angle errors, which will play a positive role in the subsequent speed update and position update of inertial navigation. For the velocity error, the calibration method of matrix decoupling eliminates the coupling term related to the gyroscope in the velocity error differential equation, and the accelerometer error parameter plays a major role in the calibration process, so it significantly improves the solution accuracy. The influence of the second-order nonlinear factor on the velocity error is related to the output of the accelerometer. In the static measurement process, the output values of the x-axis and y-axis accelerometer in the horizontal direction mainly contain noise, so for the east and north velocity errors In other words, the second-order nonlinear factor has no obvious effect on error convergence; in the vertical direction, since the sensitive axis of the accelerometer is always affected by the acceleration of gravity, the error suppression effect of the second-order nonlinear factor in the z-axis direction is significantly better than that in the horizontal direction. , but for the accelerometer, it is still the coupling item related to the gyroscope that has a more obvious influence on the error.

The embodiment of the present invention performs regression analysis on the output models of the low-cost accelerometer and gyroscope, selects a model more suitable for its output characteristics, and provides a theoretical basis for the accuracy of subsequent calibration;

Perform wavelet noise reduction processing on the raw data of the target sensor to improve the signal-to-noise ratio and ensure the accuracy of subsequent parameter estimation;

By reasonably defining the carrier coordinate system, the decoupling of the accelerometer error and the gyroscope error in the velocity error differential equation is completed, reducing the number of parameters to be calibrated and increasing the calibration speed;

It has been proved by experiments that the embodiment of the present invention has a better effect of suppressing the divergence of the output speed and position error of the low-cost MEMS inertial sensor.

Finally, it should be noted that: the above is only a preferred embodiment of the present invention, and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, it still The technical solutions recorded in the foregoing embodiments may be modified, or some technical features thereof may be equivalently replaced. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (10)

1. A low-cost MEMS inertial sensor system level calibration method, is characterized in that, comprising: Determine the sensor parameters to be marked based on the regression model; The inertial navigation error equation is obtained by decoupling the gyroscope error coupling term in the velocity error differential equation; Establishing a Kalman filter according to the determined parameters of the sensor to be marked and the inertial navigation error equation; Excite each output shaft of the sensor according to the set motion trajectory, and collect the original data of the sensor; Importing the original data into the Kalman filter after wavelet threshold noise reduction, so as to estimate the parameters to be marked; Substitute the estimated parameters to be calibrated into the inertial navigation error differential equation to verify whether the calibration results are accurate. 2. the low-cost MEMS inertial sensor system-level calibration method according to claim 1, is characterized in that, the described regression model-based determination sensor to be marked parameter comprises: Collect the output data when the sensor is in the set position; A regression model is built based on the output data. 3. The low-cost MEMS inertial sensor system-level calibration method according to claim 2, wherein, in the output data when the acquisition sensor is in a specific position, the output data includes: a three-axis accelerometer and a three-axis accelerometer Gyroscope data. 4. low-cost MEMS inertial sensor system-level calibration method according to claim 2, is characterized in that, described regression model comprises accelerometer regression model and gyroscope regression model; The accelerometer regression model is: Among them, a, b, c are parameters to be estimated, Indicates the single-axis output value of the accelerometer, and f indicates the actual acceleration value sensed by the sensitive axis of the accelerometer; The gyroscope regression model is: Among them, m and n are parameters to be estimated, Indicates the single-axis output value of the gyroscope, and ω indicates the angular velocity value actually sensed by the sensitive axis of the gyroscope. 5. the low-cost MEMS inertial sensor system level calibration method according to claim 2, is characterized in that, described set position, comprises: accelerometer set position and gyroscope set position; The set position of the accelerometer includes: the x-axis faces north, the y-axis faces west, and the z-axis faces the sky; or the x-axis faces north, the y-axis faces the ground, and the z-axis faces west; or the x-axis faces north, and the y-axis faces east , the z-axis is facing the ground; or the x-axis is facing north, the y-axis is facing the sky, and the z-axis is facing east; or the x-axis is facing north, the y-axis is facing west, and the z-axis is facing the sky; Static collection of data at each position for 10 minutes; The set position of the gyroscope includes: the position where the sensor rotates counterclockwise around the x-axis of the gyroscope every 10 minutes, and the rotation speed is 20°/s. 6. the low-cost MEMS inertial sensor system level calibration method according to claim 1, is characterized in that, described parameter to be marked comprises: Three-axis accelerometer zero-offset matrix, three-axis accelerometer installation error matrix _{δL Sym/Sksym ,} three-axis accelerometer scale coefficient error matrix δL Scal , three-axis gyroscope zero-offset matrix ε ^b , three-axis gyroscope installation error matrix δK _{Sym /Sksym} and/or the three-axis gyroscope scale coefficient error matrix δK _Scal , wherein the three-axis accelerometer installation error matrix and the three-axis gyroscope installation error matrix both include a symmetric error matrix and a skew symmetric error matrix. 7. low-cost MEMS inertial sensor system-level calibration method according to claim 6, is characterized in that, The decoupling of the gyroscope error coupling term in the speed error differential equation is carried out to obtain the inertial navigation error equation, including: The definition of the carrier coordinate system coincides with I+δK _Sksym , so that the gyroscope's obliquely symmetric error matrix is a zero matrix, thereby eliminating the coupling term due to the gyroscope's obliquely symmetric error in the velocity error differential equation, and I represents the identity matrix. 8. low-cost MEMS inertial sensor system-level calibration method according to claim 1, is characterized in that, described inertial navigation error differential equation is: The inertial navigation error differential equation includes a velocity error differential equation and a misalignment angle differential equation, Among them, ψ represents the misalignment angle vector, δv is the velocity error vector, δf ^b is the accelerometer error, f is the accelerometer measurement value, is the gyroscope error, is the angular velocity of the earth's rotation, is the angular velocity involved from the inertial coordinate system to the navigation coordinate system, is the attitude transfer matrix from the carrier coordinate system to the inertial coordinate system. 9. the low-cost MEMS inertial sensor system level calibration method according to claim 1, is characterized in that, described establishment Kalman filter comprises: Set the state quantity; Establishing a state transition matrix based on the state quantity; A measurement matrix of the state transition matrix is determined. 10. the low-cost MEMS inertial sensor system level calibration method according to claim 1, is characterized in that, described wavelet threshold noise reduction comprises: parameter settings; Perform wavelet decomposition based on the set parameters; Screening the wavelet coefficients in the wavelet decomposition by a threshold function to complete signal denoising; Perform wavelet reconstruction based on the denoised signal; Wherein, the threshold function is: In the formula, ω _{j, k} are decomposed wavelet coefficients, k and α are adjustment coefficients, and both are always positive, T is the set threshold, the formula is as follows: In the formula, N is the signal length, σ is the noise standard deviation, and j is the number of decomposition layers.