CN107563255A - The filtering method and device of a kind of Inertial Measurement Unit - Google Patents
The filtering method and device of a kind of Inertial Measurement Unit Download PDFInfo
- Publication number
- CN107563255A CN107563255A CN201610509522.3A CN201610509522A CN107563255A CN 107563255 A CN107563255 A CN 107563255A CN 201610509522 A CN201610509522 A CN 201610509522A CN 107563255 A CN107563255 A CN 107563255A
- Authority
- CN
- China
- Prior art keywords
- sensor
- wavelet
- pending
- result
- noising
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Geophysics And Detection Of Objects (AREA)
- Gyroscopes (AREA)
Abstract
IMU provided by the invention filtering method and device, low order wavelet de-noising processing is carried out first to the output signal of pending sensor；It is then detected that the state of sensor, when pending sensor and relevant axle sensor are in static state, the low-frequency noise in wavelet de-noising result is filtered using average penalty method；When pending sensor, which is in static and relevant axle sensor, is in dynamic, Kalman filtering is carried out to wavelet de-noising result, so as to play inhibitory action to the low-frequency noise in wavelet de-noising result.Low order wavelet de-noising method is employed, therefore operand is small；Moreover, eliminating influence of the low-frequency noise to the sensor under static state by average penalty method, the static properties of filtering method is improved；By kalman filter method, suppress the low-frequency noise under pending transducer dwell, relevant axle sensor dynamic, improve the dynamic property of IMU filtering.
Description
Technical field
The invention belongs to automation field, more particularly to a kind of IMU (Inertial Measurement Unit,
Inertial Measurement Unit) filtering method and device.
Background technology
IMU includes three single-axis acceleration sensors and three single shaft angular-rate sensors.Three single shafts
The orthogonal installation of acceleration transducer, for measuring the acceleration signal on each direction of principal axis；Three single shafts
The orthogonal installation of angular-rate sensor, for measuring the angular velocity signal on each direction of principal axis.
MEMS (Micro-Electro-Mechanical Systems, MEMS) IMU is being led
The more and more extensive application that boat field obtains.By MEMS IMU manufacturing process and precision level
Limitation, the random noise of its output signal is larger, therefore, it is necessary to MEMS IMU are filtered,
To improve the precision of MEMS IMU output signals.
In the prior art, MEMS IMU filtering is realized using high-order wavelet de-noising mode.Wavelet de-noising
Method replaces Fourier transformation using wavelet analysis, overcome traditional filtering method to noise compose overlap ratio compared with
Not the defects of serious signal does not reach noise reduction.But the operand of wavelet de-noising method is in exponent number
Exponential increase, thus filtering take it is long.
The content of the invention
In view of this, it is an object of the invention to provide the filtering method of IMU a kind of and device, to solve
The filtering technical problem that time-consuming in the prior art.
In a first aspect, the present invention provides a kind of Inertial Measurement Unit IMU filtering method, applied to described
The filtering of each sensor in IMU, methods described include：
Obtain the output signal of pending sensor；
The high-frequency noise in the output signal is filtered using low order wavelet de-noising method, obtains institute
State wavelet de-noising result corresponding to pending sensor；
The state of the pending sensor is judged according to the wavelet de-noising result of the pending sensor,
And it is concerned with according to corresponding to the pending sensor described in the wavelet de-noising result judgement of axle sensor
The state of relevant axle sensor；The state includes static and dynamic, and the static state is the pending biography
Sensor does not detect the state of input, and the dynamic is that the pending sensor is able to detect that input
State；
If the pending sensor and the relevant axle sensor are in static state, after extending
Time interval in the obtained wavelet de-noising result, the wavelet de-noising is filtered using average penalty method
As a result the low-frequency noise in obtains the filter result of the pending sensor；
If the pending sensor is in static and described relevant axle sensor and is in dynamic, utilize
The wavelet de-noising result of the pending sensor and the small echo drop in the dynamic relevant axle sensor
Make an uproar result, carry out Kalman filtering and obtain the filter result of the pending sensor.
Alternatively, the wavelet de-noising result according to the pending sensor judges the pending biography
The state of sensor, including：
Obtain the energy threshold for judging the pending sensor states；
When the signal energy of the wavelet de-noising result of the pending sensor is more than the energy threshold,
Determine that the pending sensor is in dynamic；
When the signal energy of the wavelet de-noising result of the pending sensor is less than or equal to the energy cut-off
During value, determine that the pending sensor remains static.
Alternatively, the wavelet de-noising result in the time interval using after extending, using average
The low-frequency noise that penalty method is filtered in the wavelet de-noising result obtains the filtering knot of the pending sensor
Fruit, including：
Extend the time interval for obtaining wavelet de-noising result；
The average value of the wavelet de-noising result obtained in the time interval after extending is calculated, is filtered out described small
The low-frequency noise that ripple noise reduction result includes, obtain the filter result of the pending sensor.
Alternatively, it is described using the wavelet de-noising result of the pending sensor and in dynamic described
The wavelet de-noising result of relevant axle sensor, carry out Kalman filtering and obtain the filter of the pending sensor
Ripple result, including：
According to the wavelet de-noising result in dynamic relevant axle sensor and the pending sensor
With the misalignment between the relevant axle sensor, the relevant axle sensor is calculated and waits to locate to described
Manage the misalignment influence amount of sensor；
Using the misalignment influence amount as observational equation, by the wavelet de-noising knot of the pending sensor
Fruit establishes Kalman filter as state equation；
Included using the wavelet de-noising result of the Kalman filter filtering pending sensor low
Frequency noise, obtain the filter result of the pending sensor.
Alternatively, methods described also includes：
If the pending sensor is in dynamic, by the wavelet de-noising knot of the pending sensor
Fruit is as filter result.
Second aspect, the present invention also provides a kind of Inertial Measurement Unit IMU filters, applied to described
The filtering of each sensor in IMU, described device include：
Acquisition module, for obtaining the output signal of pending sensor；
Wavelet de-noising module, for being made an uproar using low order wavelet de-noising method to the high frequency in the output signal
Sound is filtered, and obtains wavelet de-noising result corresponding to the pending sensor；
Condition judgment module, for being treated according to the judgement of the wavelet de-noising result of the pending sensor
The state of sensor is handled, and, the small of axle sensor of being concerned with according to corresponding to the pending sensor
Ripple noise reduction result judges the state of the relevant axle sensor；The state includes static and dynamic, described
Static state does not detect the state of input for the pending sensor, and the dynamic is the pending biography
Sensor is able to detect that the state of input；
Average compensating module, for when the pending sensor and the relevant axle sensor be in it is quiet
During state, using the wavelet de-noising result obtained in the time interval after extension, using average penalty method
Filter the low-frequency noise in the wavelet de-noising result and obtain the filter result of the pending sensor；
Kalman filtering module, for being in static and described relevant axle sensing when the pending sensor
When device is in dynamic, using the wavelet de-noising result of the pending sensor and in the dynamic phase
The wavelet de-noising result of dry axle sensor, carry out Kalman filtering and obtain the filtering of the pending sensor
As a result.
Alternatively, the condition judgment module includes：
Acquisition submodule, for obtaining the energy threshold for being used for judging the pending sensor states；
First determination sub-module, the signal energy for the wavelet de-noising result when the pending sensor
During more than the energy threshold, determine that the pending sensor is in dynamic；
Second determination sub-module, the signal energy for the wavelet de-noising result when the pending sensor
During less than or equal to the energy threshold, determine that the pending sensor remains static.
Alternatively, the average compensating module includes：
Time domain extends submodule, and the time interval of wavelet de-noising result is obtained for extending；
Mean value computation submodule, for calculating the wavelet de-noising result obtained in the time interval after extending
Average value, the low-frequency noise that the wavelet de-noising result includes is filtered out, obtains the pending sensor
Filter result.
Alternatively, the Kalman filtering module includes：
Calculating sub module, for according to the wavelet de-noising result in dynamic relevant axle sensor and
Misalignment between the pending sensor and the relevant axle sensor, the relevant axle is calculated
Misalignment influence amount of the sensor to the pending sensor；
Wave filter setting up submodule, for using the misalignment influence amount as observational equation, being treated described
The wavelet de-noising result of processing sensor establishes Kalman filter as state equation；
Filtering process submodule, for filtering the pending sensor using the Kalman filter
The low-frequency noise that wavelet de-noising result includes, obtain the filter result of the pending sensor.
Alternatively, described device also includes：
Determining module, for when the pending sensor is in dynamic, by the pending sensor
Wavelet de-noising result as filter result.
Compared with prior art, above-mentioned technical proposal provided by the invention has the following advantages that：The present embodiment
The IMU of offer filtering method, low order wavelet de-noising is carried out first to the output signal of pending sensor
Processing；It is then detected that the state of sensor, when pending sensor and relevant axle sensor be in it is quiet
During state, the low-frequency noise in wavelet de-noising result is filtered using average penalty method；At pending sensor
When static and relevant axle sensor is in dynamic, Kalman filtering is carried out to wavelet de-noising result, from
And inhibitory action is played to the low-frequency noise in wavelet de-noising result.It follows that the process employs low
Rank wavelet de-noising method, operand are small；Moreover, low-frequency noise is eliminated under static state by average penalty method
Sensor influence, improve the static properties of filtering method；By kalman filter method, suppress
Low-frequency noise under pending transducer dwell, relevant axle sensor dynamic, improve IMU filtering side
The dynamic property of method.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to reality
The required accompanying drawing used in example or description of the prior art is applied to be briefly described, it should be apparent that, below
Accompanying drawing in description is some embodiments of the present invention, for those of ordinary skill in the art, not
On the premise of paying creative work, other accompanying drawings can also be obtained according to these accompanying drawings.
Fig. 1 shows a kind of IMU of embodiment of the present invention filtering method schematic flow sheet；
Fig. 2 shows a kind of schematic flow sheet for judging sensor states step of the present invention；
Fig. 3 shows a kind of schematic flow sheet of Kalman filtering step of the present invention；
Fig. 4 shows a kind of IMU filter block diagram of the present invention；
Fig. 5 shows a kind of block diagram of condition judgment module of the present invention；
Fig. 6 shows a kind of block diagram of average compensating module of the embodiment of the present invention；
Fig. 7 shows a kind of block diagram of Kalman filtering module of the embodiment of the present invention.
Embodiment
The filtering method for the IMU that the present embodiment provides, based on wavelet de-noising Theoretical Design, because small echo drops
The operand of method for de-noising uses low order wavelet de-noising method with the increase of exponent number exponentially in of the invention.But
Low order wavelet de-noising method is undesirable to the noise reduction of low-frequency noise, therefore, in sensor output signal
Signal to noise ratio it is less high when, using low order wavelet de-noising processing on the basis of, using other method mistakes
The low-frequency noise in low frequency signal is filtered, IMU filter result is all reached under different motion states
It is optimal.
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with this hair
Accompanying drawing in bright embodiment, the technical scheme in the embodiment of the present invention is clearly and completely described,
Obviously, described embodiment is part of the embodiment of the present invention, rather than whole embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained under the premise of creative work is not made
The every other embodiment obtained, belongs to the scope of protection of the invention.
Fig. 1 is referred to, shows a kind of schematic flow sheet of IMU filtering method of the embodiment of the present invention,
All comprise the following steps for the filtering method of any one sensor in IMU：
S110, obtain the output signal of pending sensor.
Pending sensor can be any one sensor in IMU.
Output signal is the testing result of pending sensor, and for acceleration transducer, it is exported
Signal is the acceleration signal of carrier where the IMU detected；For angular-rate sensor, its is defeated
Go out the angular velocity signal that signal is carrier where the INU detected.
S120, the high-frequency noise in output signal is filtered using low order wavelet de-noising method, obtained
Wavelet de-noising result corresponding to pending sensor.
For the output signal of pending sensor, first with wavelet de-noising method to pending sensor
Output signal carry out noise reduction process, filter out the high-frequency noise in output signal, obtain the pending biography
The wavelet de-noising result of sensor.
Wavelet de-noising principle is according to possessed by the wavelet coefficient of useful signal and noise on different scale
Heterogeneity, corresponding rule is constructed, the small echo in wavelet field using other mathematical methods to signals and associated noises
Coefficient is handled, and its essence is to reduce even to reject the wavelet coefficient as caused by noise completely, meanwhile,
Retain the wavelet coefficient of useful signal to greatest extent.Wherein, the wavelet de-noising based on threshold value is to small echo
It is less than the wavelet coefficient zero setting of threshold value in each layer coefficients after decomposition, inverse transformation is then carried out again, so as to reach
To the purpose for removing noise.
In some embodiments of the invention, wavelet de-noising processing procedure may comprise steps of：
1) wavelet decomposition is carried out to the output signal of pending sensor.It is small using 4 ranks in the present embodiment
Wave Decomposition, will the output signal resolve into 4 stratum.
2) the wavelet coefficient estimation noise variance of the first stratum is utilized.Normally most of noise is included in the
In one stratum, therefore, noise variance is estimated using the wavelet coefficient of the first stratum.
3) hard -threshold of each stratum of Noise Variance Estimation is utilized.
According to the noise variance estimated in 2), calculated in each rank wavelet coefficient and made an uproar by ssystem transfer function
The estimation of sound varianceAnd then the estimation of the noise variance in each rank wavelet coefficientIt is small to calculate each rank
The hard -threshold of wave system numberWherein, δiPoor for the noise criteria in i rank wavelet coefficients, N is
The length of the output signal collected.
If hard -threshold selection is too small, wavelet coefficient will can not be zeroed out corresponding to the noise of some.
So as to remain partial noise in the signal after denoising, the effect of denoising is not ideal.If hard -threshold
Select bigger than normal, then a part of useful signal can be filtered so that occur the phenomenon of distortion after denoising.
4) for the wavelet coefficient of each stratum, compared with the hard -threshold of the stratum, if wavelet systems
Number is less than or equal to hard -threshold corresponding to the stratum, then by the wavelet coefficient zero setting.
5) wavelet coefficient of each stratum is reconstructed, obtains wavelet de-noising result.
Restructuring procedure is to carry out the conversion opposite with wavelet decomposition to the wavelet coefficient of non-zero in each stratum,
The process of (that is, inverse transformation).
S130, the state of pending sensor is judged according to the wavelet de-noising result of pending sensor, such as
The pending sensor of fruit is in static state, then performs S140；If pending sensor is in dynamic, hold
Row S170.
The state includes static and dynamic, and static state refers to state of the sensor without input, for example, right
For X-axis angular-rate sensor, static state refers to that X-axis angular-rate sensor does not detect in X-direction
To angular speed, i.e. X-axis angular-rate sensor does not input.Corresponding, it is defeated dynamically to refer to that sensor has
The state entered, for example, during IMU inactive states, the accelerometer in acceleration of gravity direction can be examined
Measure acceleration of gravity, i.e. the accelerometer is in dynamic.
When sensor is in dynamic, the energy of its output signal is larger；And when sensor is in static,
The energy of its output signal is smaller；In turn, biography be can interpolate that out according to the energy of sensor output signal
Sensor is in static or dynamic.
S140, the wavelet de-noising result for the axle sensor that is concerned with according to corresponding to pending sensor, judges phase
The state of dry axle sensor, if relevant axle sensor remains static, perform S150；If phase
Dry axle sensor is in dynamic, then performs S160.
For some angular-rate sensor or acceleration transducer in IMU, if installation when by
There is misalignment between the sensor on the sensor and other two axles that error causes the axle (that is, should
The angle between sensor on the sensor of axle and other two axles is not 90 °), then on other two axles
Sensor be properly termed as corresponding to the axle sensor be concerned with axle sensor.
The condition adjudgement process of relevant axle sensor refers to the associated description in S130, and here is omitted.
S150, using the wavelet de-noising result obtained in the time interval after extension, using equal value complement
Repay the low-frequency noise that method is filtered in the wavelet de-noising result and obtain the filtering knot of the pending sensor
Fruit.
When pending sensor and relevant axle sensor are in static state, are limited and treated by wavelet de-noising exponent number
Handle in the wavelet de-noising result of the output signal of sensor and low-frequency noise be present.Low-frequency noise can influence to filter
The performance of the inactive state of ripple algorithm.
Low-frequency noise belongs to the noise of low-frequency range, and the ability in longer time section is needed in time-domain
Detect low-frequency noise.Therefore, the time interval for obtaining wavelet de-noising result is extended, that is, when obtaining longer
Between wavelet de-noising result in section, then, calculate wavelet de-noising result in the time interval after extending
Average value, to filter the low-frequency noise in wavelet de-noising result, to eliminate the influence of static lower low-frequency noise.
Finally give the filter result of pending sensor.
S160, sensed using the wavelet de-noising result of pending sensor and in the dynamic relevant axle
The wavelet de-noising result of device, carry out Kalman filtering and obtain the filter result of the pending sensor.
When pending sensor is in static state, and relevant axle sensor is in dynamic, due to pending sensing
The presence of misalignment between device and relevant axle sensor, the output of relevant axle sensor some decompose and treat
Handle on sensor, i.e., the dynamic introducing due to the axle sensor that is concerned with pending transducer dwell is micro-
Small-signal, the tiny signal are limited by wavelet decomposition exponent number and coexisted in low-frequency noise in wavelet de-noising result.
Inhibitory action is played to the low-frequency noise in wavelet de-noising result using kalman filter method, finally
Obtain the filter result of pending sensor.
S170, the filter result using the wavelet de-noising result of pending sensor as pending sensor.
When pending sensor is in dynamic, the signal to noise ratio of pending sensor output signal is higher, low
Rank wavelet de-noising effect is relatively good, therefore, can be directly using wavelet de-noising result as pending sensor
Filter result.
The filtering method for the IMU that the present embodiment provides, is carried out first to the output signal of pending sensor
The processing of low order wavelet de-noising；It is then detected that the state of sensor, when pending sensor and relevant axle pass
When sensor is in static state, the low-frequency noise in wavelet de-noising result is filtered using average penalty method；When treating
When processing sensor is in static and relevant axle sensor and is in dynamic, wavelet de-noising result is blocked
Kalman Filtering, so as to play inhibitory action to the low-frequency noise in wavelet de-noising result.It follows that should
Method employs low order wavelet de-noising method, and operand is small；Moreover, low frequency is eliminated by average penalty method
Influence of the noise to the sensor under static state, improve the static properties of filtering method；Filtered by Kalman
Wave method, suppress the low-frequency noise under pending transducer dwell, relevant axle sensor dynamic, improve
The dynamic property of IMU filtering method.
The step of below in conjunction with Fig. 2 to judging sensor states in above-described embodiment (S120 in Fig. 1)
It is described in detail.
Fig. 2 is referred to, a kind of schematic flow sheet for judging sensor states step of the present invention is shown, such as schemes
Shown in 2, judge that the process of sensor states may comprise steps of：
S210, obtain the energy threshold for judging pending sensor states.
In some embodiments of the invention, the noise side for estimating to obtain in wavelet de-noising processing procedure is utilized
The parameter such as poor (referring to the related content in S120), the transmission function of each stratum and probability coefficent,
Estimation obtains energy threshold.Wherein, probability coefficent can be obtained ahead of time by experiment.
S220, the signal energy of wavelet de-noising result and the size of energy threshold of more pending sensor
Relation；If the signal energy of wavelet de-noising result is more than energy threshold, S230 is performed；If small echo
The signal energy of noise reduction result is less than or equal to energy threshold, then performs S240.
If sensor is in dynamic, the energy of its output signal is higher, moreover, must be included in small
In ripple noise reduction result；If sensor is in static state, the energy ratio of its output signal is relatively low, moreover,
Useful signal included in wavelet de-noising result is also than relatively low.Therefore, by wavelet de-noising result
Signal energy can determine that the state of sensor.
S230, determine that pending sensor is in dynamic.
When the energy for detecting the signal in wavelet de-noising result is more than energy threshold, determine that this is pending
Sensor is in dynamic.
S240, determine that pending sensor is in static state.
When the energy for detecting signal in wavelet de-noising result is less than or equal to energy threshold, it is determined that this is treated
Handle sensor and be in static state.
The process for judging sensor states that the present embodiment provides, according to signal energy with being calculated in real time
Noise energy judge the state of sensor, environment and sensor type are not strict with, are applicable model
Enclose wide.
The step of below in conjunction with Fig. 3 to Kalman filtering, (S160 in Fig. 1) was described in detail.
Fig. 3 is referred to, a kind of schematic flow sheet of Kalman filtering step of the present invention is shown, such as Fig. 3 institutes
Show, Kalman filtering process may comprise steps of：
S310, according to the wavelet de-noising result in the relevant axle sensor of dynamic and pending sensor
With the misalignment between the relevant axle sensor, relevant axle sensor is calculated to the pending biography
The misalignment influence amount of sensor.
For example, pending sensor is X-axis sensor, Y-axis sensor is in dynamic, then according to Y-axis
The wavelet de-noising result of sensor, and, the misalignment between X-axis sensor and Y-axis sensor, meter
Calculation obtains component of the Y-axis sensor in X-axis sensor orientation, i.e. misalignment influence amount.
S320, using misalignment influence amount as observational equation, by the wavelet de-noising result of pending sensor
Kalman filter is established as state equation.
Kalman filtering algorithm is according to the state equation and observational equation of foundation to needing signal to be processed to make
Meet the estimation of least mean-square error.Its basic thought is：Using least mean-square error as best estimate criterion,
Using signal and the state-space model of noise, the estimate of previous moment and the observation at current time are utilized
Value updates the estimation to state variable, obtains the estimate at current time.
In the present embodiment, using misalignment influence amount as observational equation, dropped with the small echo of pending sensor
Result of making an uproar utilizes the estimate of the wavelet de-noising signal of the pending sensor of previous moment as state equation
The filter result that pending device sensor is updated with current time misalignment influence amount is estimated, obtains
The estimate at pending sensor current time at current time.
S330, included using the wavelet de-noising result of the pending sensor of Kalman filter filtering low
Frequency noise, obtain the filter result of pending sensor.
Kalman filter can make the filter result of pending sensor more level off to pending sensor pair
The misalignment influence amount answered, eliminates the influence of low-frequency noise, finally gives the filtering knot of pending sensor
Fruit.
Corresponding to above-mentioned IMU filtering method embodiment, present invention also offers IMU filter
Embodiment.
Fig. 4 is referred to, a kind of IMU filter block diagram of the present invention is shown, as shown in figure 4, IMU
The filter of each interior sensor includes：Acquisition module 110, wavelet de-noising module 120, shape
State judge module 130, average compensating module 140 and Kalman filtering module 150.
Acquisition module 110, for obtaining the output signal of pending sensor.
Wavelet de-noising module 120, for utilizing low order wavelet de-noising method to the high frequency in the output signal
Noise is filtered, and obtains wavelet de-noising result corresponding to the pending sensor.
Condition judgment module 130, described in being judged according to the wavelet de-noising result of the pending sensor
The state of pending sensor, and, be concerned with axle sensor according to corresponding to the pending sensor
Wavelet de-noising result judges the state of the relevant axle sensor.
The state includes static and dynamic, and static state does not detect the state of input for pending sensor,
The dynamic is the state that pending sensor is able to detect that input.
Average compensating module 140, for being in when the pending sensor and the relevant axle sensor
When static, using the wavelet de-noising result obtained in the time interval after extension, compensated using average
The low-frequency noise that method is filtered in the wavelet de-noising result obtains the filter result of the pending sensor.
Kalman filtering module 150, for being passed when the pending sensor is in static and described relevant axle
When sensor is in dynamic, using the wavelet de-noising result of the pending sensor and in dynamic described
The wavelet de-noising result of relevant axle sensor, carry out Kalman filtering and obtain the filter of the pending sensor
Ripple result.
In addition, the filter for the IMU that the present embodiment provides also includes determining module 210, for working as shape
When state judge module judges that pending sensor is in dynamic, directly using wavelet de-noising result as filtering
As a result.
The filter for the IMU that the present embodiment provides, is carried out first to the output signal of pending sensor
The processing of low order wavelet de-noising；It is then detected that the state of sensor, when pending sensor and relevant axle pass
When sensor is in static state, the low-frequency noise in wavelet de-noising result is filtered using average penalty method；When treating
When processing sensor is in static and relevant axle sensor and is in dynamic, wavelet de-noising result is blocked
Kalman Filtering, so as to play inhibitory action to the low-frequency noise in wavelet de-noising result.It follows that should
Device employs low order wavelet de-noising method, and operand is small；Moreover, low frequency is eliminated by average penalty method
Influence of the noise to the sensor under static state, improve the static properties of filtering method；Filtered by Kalman
Wave method, suppress the low-frequency noise under pending transducer dwell, relevant axle sensor dynamic, improve
The dynamic property of IMU filtering method.
The condition judgment module 130 in above-described embodiment is described in detail below in conjunction with Fig. 5.
Fig. 5 is referred to, a kind of block diagram of condition judgment module of the present invention is shown, as shown in figure 5, the shape
State judge module includes：Acquisition submodule 131, the first determination sub-module 132 and the second determination sub-module
133。
Acquisition submodule 131, for obtaining the energy threshold for being used for judging the pending sensor states.
First determination sub-module 132, the signal energy for the wavelet de-noising result when pending sensor are big
When the energy threshold, determine that pending sensor is in dynamic.
Second determination sub-module 133, the signal energy for the wavelet de-noising result when pending sensor are small
When the energy threshold, determine that pending sensor remains static.
The noise energy that the condition judgment module that the present embodiment provides is calculated according to signal energy and in real time
Judge the state of sensor, environment and sensor type are not strict with, it is applied widely.
The average compensating module 140 in above-described embodiment is described in detail below in conjunction with Fig. 6.
Fig. 6 is referred to, shows a kind of block diagram of average compensating module of the embodiment of the present invention, average compensation
Module includes：Time domain extends submodule 141 and mean value computation submodule 142.
Time domain extends submodule 141, and the time interval of wavelet de-noising result is obtained for extending.
Mean value computation submodule 142, for calculating the wavelet de-noising result obtained in the time interval after extending
Average value, the low-frequency noise that the wavelet de-noising result includes is filtered out, obtains the pending sensor
Filter result.
The average compensating module that the present embodiment provides, dropped using the small echo obtained in the time interval after extension
Result of making an uproar averaged descends the influence of low-frequency noise, average compensation method to eliminate sensor in static state
Calculating process is simple, operand is small.
The Kalman filtering module 150 in above-described embodiment is described in detail below in conjunction with Fig. 7.
Fig. 7 is referred to, a kind of block diagram of Kalman filtering module of the embodiment of the present invention is shown, such as Fig. 7 institutes
Show, the Kalman filtering module includes：Calculating sub module 151, wave filter setting up submodule 152 and filtering
Handle submodule 153.
Calculating sub module 151, for according to the wavelet de-noising result in dynamic relevant axle sensor, with
And the misalignment between the pending sensor and the relevant axle sensor, it is calculated described relevant
Misalignment influence amount of the axle sensor to the pending sensor.
Wave filter setting up submodule 152, for using the misalignment influence amount as observational equation, will described in
The wavelet de-noising result of pending sensor establishes Kalman filter as state equation.
Filtering process submodule 153, for filtering the pending sensor using the Kalman filter
The low-frequency noise that includes of wavelet de-noising result, obtain the filter result of the pending sensor.
It should be noted that each embodiment in this specification is described by the way of progressive, each
What embodiment stressed is all the difference with other embodiment, identical similar between each embodiment
Part mutually referring to.For device class embodiment, due to itself and the basic phase of embodiment of the method
Seemingly, so what is described is fairly simple, the relevent part can refer to the partial explaination of embodiments of method.
Finally, it is to be noted that, herein, such as first and second or the like relational terms
It is used merely to make a distinction an entity or operation with another entity or operation, and not necessarily requires
Either imply between these entities or operation any this actual relation or order be present.Moreover, art
Language " comprising ", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that bag
Including process, method, article or the equipment of a series of elements not only includes those key elements, but also including
The other element being not expressly set out, or it is this process, method, article or equipment also to include
Intrinsic key element.In the absence of more restrictions, the key element limited by sentence "including a ...",
It is not precluded from addition identical in the process including the key element, method, article or equipment being also present
Key element.
The foregoing description of the disclosed embodiments, those skilled in the art are enable to realize or use this hair
It is bright.A variety of modifications to these embodiments will be apparent for a person skilled in the art, herein
Defined in General Principle can without departing from the spirit or scope of the present invention, in other realities
Apply in example and realize.Therefore, the present invention is not intended to be limited to the embodiments shown herein, but will
Meet the most wide scope consistent with principles disclosed herein and features of novelty.
Described above is only the preferred embodiment of the present invention, it is noted that for the general of the art
For logical technical staff, under the premise without departing from the principles of the invention, some improvement and profit can also be made
Decorations, these improvements and modifications also should be regarded as protection scope of the present invention.
Claims (10)
- A kind of 1. Inertial Measurement Unit IMU filtering method, applied to each biography in the IMU The filtering of sensor, it is characterised in that methods described includes：Obtain the output signal of pending sensor；The high-frequency noise in the output signal is filtered using low order wavelet de-noising method, obtains institute State wavelet de-noising result corresponding to pending sensor；The state of the pending sensor is judged according to the wavelet de-noising result of the pending sensor, And it is concerned with according to corresponding to the pending sensor described in the wavelet de-noising result judgement of axle sensor The state of relevant axle sensor；The state includes static and dynamic, and the static state is the pending biography Sensor does not detect the state of input, and the dynamic is that the pending sensor is able to detect that input State；If the pending sensor and the relevant axle sensor are in static state, after extending Time interval in the obtained wavelet de-noising result, the wavelet de-noising is filtered using average penalty method As a result the low-frequency noise in obtains the filter result of the pending sensor；If the pending sensor is in static and described relevant axle sensor and is in dynamic, utilize The wavelet de-noising result of the pending sensor and the small echo drop in the dynamic relevant axle sensor Make an uproar result, carry out Kalman filtering and obtain the filter result of the pending sensor.
- 2. according to the method for claim 1, it is characterised in that described according to the pending sensing The wavelet de-noising result of device judges the state of the pending sensor, including：Obtain the energy threshold for judging the pending sensor states；When the signal energy of the wavelet de-noising result of the pending sensor is more than the energy threshold, Determine that the pending sensor is in dynamic；When the signal energy of the wavelet de-noising result of the pending sensor is less than or equal to the energy cut-off During value, determine that the pending sensor remains static.
- 3. according to the method for claim 1, it is characterised in that the time zone using after extending The interior wavelet de-noising result, the low frequency in the wavelet de-noising result is filtered using average penalty method Noise obtains the filter result of the pending sensor, including：Extend the time interval for obtaining wavelet de-noising result；The average value of the wavelet de-noising result obtained in the time interval after extending is calculated, is filtered out described small The low-frequency noise that ripple noise reduction result includes, obtain the filter result of the pending sensor.
- 4. according to the method for claim 1, it is characterised in that described to utilize the pending sensing The wavelet de-noising result of device and the wavelet de-noising result in the dynamic relevant axle sensor, are blocked Kalman Filtering obtains the filter result of the pending sensor, including：According to the wavelet de-noising result in dynamic relevant axle sensor and the pending sensor With the misalignment between the relevant axle sensor, the relevant axle sensor is calculated and waits to locate to described Manage the misalignment influence amount of sensor；Using the misalignment influence amount as observational equation, by the wavelet de-noising knot of the pending sensor Fruit establishes Kalman filter as state equation；Included using the wavelet de-noising result of the Kalman filter filtering pending sensor low Frequency noise, obtain the filter result of the pending sensor.
- 5. according to the method described in claim any one of 1-4, it is characterised in that methods described also includes：If the pending sensor is in dynamic, by the wavelet de-noising knot of the pending sensor Fruit is as filter result.
- A kind of 6. Inertial Measurement Unit IMU filters, applied to each sensing in the IMU The filtering of device, it is characterised in that described device includes：Acquisition module, for obtaining the output signal of pending sensor；Wavelet de-noising module, for being made an uproar using low order wavelet de-noising method to the high frequency in the output signal Sound is filtered, and obtains wavelet de-noising result corresponding to the pending sensor；Condition judgment module, for being treated according to the judgement of the wavelet de-noising result of the pending sensor The state of sensor is handled, and, the small of axle sensor of being concerned with according to corresponding to the pending sensor Ripple noise reduction result judges the state of the relevant axle sensor；The state includes static and dynamic, described Static state does not detect the state of input for the pending sensor, and the dynamic is the pending biography Sensor is able to detect that the state of input；Average compensating module, for when the pending sensor and the relevant axle sensor be in it is quiet During state, using the wavelet de-noising result obtained in the time interval after extension, using average penalty method Filter the low-frequency noise in the wavelet de-noising result and obtain the filter result of the pending sensor；Kalman filtering module, for being in static and described relevant axle sensing when the pending sensor When device is in dynamic, using the wavelet de-noising result of the pending sensor and in the dynamic phase The wavelet de-noising result of dry axle sensor, carry out Kalman filtering and obtain the filtering of the pending sensor As a result.
- 7. device according to claim 6, it is characterised in that the condition judgment module includes：Acquisition submodule, for obtaining the energy threshold for being used for judging the pending sensor states；First determination sub-module, the signal energy for the wavelet de-noising result when the pending sensor During more than the energy threshold, determine that the pending sensor is in dynamic；Second determination sub-module, the signal energy for the wavelet de-noising result when the pending sensor During less than or equal to the energy threshold, determine that the pending sensor remains static.
- 8. device according to claim 6, it is characterised in that the average compensating module includes：Time domain extends submodule, and the time interval of wavelet de-noising result is obtained for extending；Mean value computation submodule, for calculating the wavelet de-noising result obtained in the time interval after extending Average value, the low-frequency noise that the wavelet de-noising result includes is filtered out, obtains the pending sensor Filter result.
- 9. device according to claim 6, it is characterised in that the Kalman filtering module includes：Calculating sub module, for according to the wavelet de-noising result in dynamic relevant axle sensor and Misalignment between the pending sensor and the relevant axle sensor, the relevant axle is calculated Misalignment influence amount of the sensor to the pending sensor；Wave filter setting up submodule, for using the misalignment influence amount as observational equation, being treated described The wavelet de-noising result of processing sensor establishes Kalman filter as state equation；Filtering process submodule, for filtering the pending sensor using the Kalman filter The low-frequency noise that wavelet de-noising result includes, obtain the filter result of the pending sensor.
- 10. according to the device described in claim any one of 6-9, it is characterised in that described device also includes：Determining module, for when the pending sensor is in dynamic, by the pending sensor Wavelet de-noising result as filter result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610509522.3A CN107563255B (en) | 2016-06-30 | 2016-06-30 | Filtering method and device for inertial measurement unit |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610509522.3A CN107563255B (en) | 2016-06-30 | 2016-06-30 | Filtering method and device for inertial measurement unit |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107563255A true CN107563255A (en) | 2018-01-09 |
| CN107563255B CN107563255B (en) | 2020-09-29 |
Family
ID=60970016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610509522.3A Active CN107563255B (en) | 2016-06-30 | 2016-06-30 | Filtering method and device for inertial measurement unit |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107563255B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109358745A (en) * | 2018-08-31 | 2019-02-19 | 青岛小鸟看看科技有限公司 | The position filtering method, apparatus and computer storage medium of interactive handle |
| CN109613299A (en) * | 2018-11-14 | 2019-04-12 | 深圳绿米联创科技有限公司 | Viberation detector, method, intelligent door lock and system |
| CN112931393A (en) * | 2021-01-18 | 2021-06-11 | 湖南文理学院 | Cleaning device, cleaning system and cleaning method for aquaculture |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103335650A (en) * | 2013-05-29 | 2013-10-02 | 哈尔滨工程大学 | Coordinate mismatch measuring method based on inertial measurement unit |
| CN104699881A (en) * | 2013-12-10 | 2015-06-10 | 中国航空工业第六一八研究所 | Design method of IMU digital filter for inertial navigation system |
| CN105318876A (en) * | 2014-07-09 | 2016-02-10 | 北京自动化控制设备研究所 | Inertia and mileometer combination high-precision attitude measurement method |
| CN105628026A (en) * | 2016-03-04 | 2016-06-01 | 深圳大学 | Positioning and posture determining method and system of mobile object |
-
2016
- 2016-06-30 CN CN201610509522.3A patent/CN107563255B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103335650A (en) * | 2013-05-29 | 2013-10-02 | 哈尔滨工程大学 | Coordinate mismatch measuring method based on inertial measurement unit |
| CN104699881A (en) * | 2013-12-10 | 2015-06-10 | 中国航空工业第六一八研究所 | Design method of IMU digital filter for inertial navigation system |
| CN105318876A (en) * | 2014-07-09 | 2016-02-10 | 北京自动化控制设备研究所 | Inertia and mileometer combination high-precision attitude measurement method |
| CN105628026A (en) * | 2016-03-04 | 2016-06-01 | 深圳大学 | Positioning and posture determining method and system of mobile object |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109358745A (en) * | 2018-08-31 | 2019-02-19 | 青岛小鸟看看科技有限公司 | The position filtering method, apparatus and computer storage medium of interactive handle |
| CN109358745B (en) * | 2018-08-31 | 2022-07-19 | 青岛小鸟看看科技有限公司 | Position filtering method and device of interactive handle and computer storage medium |
| CN109613299A (en) * | 2018-11-14 | 2019-04-12 | 深圳绿米联创科技有限公司 | Viberation detector, method, intelligent door lock and system |
| CN112931393A (en) * | 2021-01-18 | 2021-06-11 | 湖南文理学院 | Cleaning device, cleaning system and cleaning method for aquaculture |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107563255B (en) | 2020-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Miao et al. | Sparse maximum harmonics-to-noise-ratio deconvolution for weak fault signature detection in bearings | |
| CN107563255A (en) | The filtering method and device of a kind of Inertial Measurement Unit | |
| Narasimhappa et al. | A modified Sage-Husa adaptive Kalman filter for denoising Fiber Optic Gyroscope signal | |
| CN104596514B (en) | The Real-time Noisy Reducer and method of accelerometer and gyroscope | |
| Dong et al. | Structural damage detection using empirical-mode decomposition and vector autoregressive moving average model | |
| CN109883392B (en) | Strapdown inertial navigation heave measurement method based on phase compensation | |
| US20090282906A1 (en) | Apparatus and method for detecting decrease in tire air pressure and program for detecting decrease in tire air pressure | |
| CN105701456B (en) | A kind of angular accelerometer signal adaptive denoising method based on wavelet analysis | |
| KR101438628B1 (en) | Method and system for controlling anti-jerk of vehicle | |
| CN106199743A (en) | Magnetic anomaly signal detecting method | |
| CN105698789A (en) | Ship heaving measurement method and system | |
| CN104280047A (en) | Gyroscope shift filtering system and method integrating multiple sensors | |
| CN104215833B (en) | power system frequency measuring method and device | |
| Cho et al. | Analysis of dynamic model of a top-loading laundry machine with a hydraulic balancer | |
| CN103940447B (en) | Mooring state initial aligning method based on self-adaptive digital filter | |
| KR101444506B1 (en) | Method for monitoring status of a bearing apparatus for monitoring status of a bearing | |
| CN106679659A (en) | Signal denoising method based on parameter-adjustable nonlinear track differentiator | |
| Yongpan et al. | Real-time zero phase filtering for heave measurement | |
| JP6475954B2 (en) | Tire pressure drop detection device, method and program | |
| CN110596425B (en) | Noise elimination method for MEMS acceleration sensor of unmanned aerial vehicle | |
| Guo et al. | Butterworth low-pass filter for processing inertial navigation system raw data | |
| CN107664499B (en) | On-line noise reduction method for accelerometer of ship strapdown inertial navigation system | |
| CN113625214B (en) | Magnetometer electromagnetic interference diagnosis filtering method and system | |
| Mizuno et al. | An adaptive filtering technique for driver's heart rate monitoring through vibration signal by seat-embedded piezoelectric sensors | |
| CN112924154A (en) | Method and device for extracting damping ratio of structural response signal in high-noise flowing environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |