Base band time difference estimation method under the local oscillator conformity error of a kind of strange land
Technical field
The invention belongs to signal processing technology, the time difference high-acruracy survey particularly in positioning using TDOA system.
Background technology
TDOA estimation technology is widely used in radar, sonar, speech enhan-cement, automatic camera tracking etc., and particularly in radiation source location, TDOA estimation technology has important using value.For radiation source positioning using TDOA system, the electromagnetic signal that radiate source radiation goes out is intercepted by two receivers in strange land, and goes out the time of advent poor (TDOA) between two intercepted signals by TDOA estimation algorithm measurement.
TDOA estimation technology generally can realize at the intermediate frequency of receiver or base band, the major defect that intermediate frequency realizes TDOA estimation is that intermediate frequency data stream is too large, this all brings very large difficulty to the transmission of data and real-time process, consider this practical difficulty, realizing TDOA estimation in base band is comparatively ideal selection.
The most frequently used and effective time difference estimation method is general cross correlation, but when Practical Project problem, the precision that the classical delay time estimation method of direct utilization carries out time difference measurement will be limited to the time interval between base band data sample point, because baseband sampling rate is lower, the precision of the TDOA estimation result at the integral multiple sample point interval that traditional correlation method provides obviously is to meet actual requirement of engineering; Simultaneously, be down-converted in the process of base band in radiofrequency signal, the high consistency accomplishing local carrier frequency between carrier synchronization strict in receiver or receiver and phase place is all very difficult, in two remote reception machines, local oscillator (LO) there will be conformity error, this will cause the change of two receiver place baseband signal time delay relations, and these all can have influence on final TDOA estimation result.
Summary of the invention
Technical problem to be solved by this invention is, provides a kind of high-precision base band time difference estimation method that can overcome frequency invariance error and phase equalization error between remote reception machine local oscillator.
The present invention is that the base band time difference estimation method under a kind of local oscillator conformity error, comprises the following steps for solving the problems of the technologies described above adopted technical scheme:
1) radio frequency rf signal received the 1st receiver and the 2nd receiver carries out down-converted and obtains baseband digital signal, goes average and normalized to obtain base band observation data y required for time difference measurement to baseband digital signal
1, y
2;
2) the frequency invariance error between introducing two separation receiver mixing local oscillators and phase equalization error construct the target function L carrying out TDOA estimation
1,
wherein, τ
t, 2,1represent the signal arrival time difference value between the 2nd receiver and the 1st receiver; f
2,1represent the frequency invariance error between two remote reception machine local oscillators;
represent the phase equalization error between two receivers 1,2; Re represents and gets real part, y
1(τ
t, 2,1) represent that the signal that receives of receiver 1 is through time delay τ
t, 2,1after signal, H represents conjugate transpose; y
1(τ
t, 2,1)=y
1[0-τ
t, 2,1t
δ] ... y
1[n-τ
t, 2,1t
δ] ... y
1[N-1-τ
t, 2,1t
δ]
t, variable n is integer, and its scope is that 0 to N-1, N represent baseband sample point sum, T
δrepresent the base band adjacent sample point time interval, the observation data y at receiver 2 place
2=[y
2[0] y
2[1] ... y
2[n] ... y
2[N-1]]
t; Matrix
for:
3) peak value searching is carried out to target function,
by the time difference τ that the target function peak point searched is corresponding
t, 2,1obtain TDOA estimation value
∧ represents estimated value, and argmax represents the parameter found and have maximum scores; T representing matrix transposition.
The present invention considers frequency invariance error and initial phase conformity error between remote reception machine local oscillator when estimating the time difference, these errors always exist in Practical Project, if do not consider, these errors directly utilize conventional method to carry out time difference measurement, the accuracy of its time difference measurement is certain to be affected, these errors take into account in the data model of time difference measurement by the present invention, and estimate that the time difference not only can reduce the frequency of time difference measurement system to strange land local oscillator and the coherence request of phase place according to this model, and the performance of TDOA estimation can be improved.
Due to target function L
1comprise 3 unknown quantitys, need to carry out three-dimensional search, therefore computation complexity is higher, in order to be beneficial to Project Realization, needs to reduce operand.The invention provides the dimension-reduction treatment method reducing operand:
In step 2) in, to target function L
1carry out dimension-reduction treatment, structure binary target function L
3,
Wherein, matrix Φ
f(τ
t, 2,1, f
2,1) be:
In step 3), peak value searching is carried out to binary target function,
by the time difference τ that the target function peak point searched is corresponding
t, 2,1obtain TDOA estimation value
If also need to carry out a step compaction algorithms amount, improve estimating speed, the present invention also provides the method for two-dimensional search being down to linear search further: when calculating binary target function value, first estimate the frequency invariance error estimate between two strange land local oscillators under current time difference variable
by current time difference variable lower frequency conformity error estimated value
substitute into binary target function L
3after, then peak value searching is carried out to target function try to achieve TDOA estimation;
The estimated value of the frequency invariance error between two remote reception machine local oscillators under current time difference variable
computational methods be:
Wherein, intermediate quantity
Indirect observation amount
Angle represents and asks phase angle.
The invention has the beneficial effects as follows, the present invention is applied to the precision that not only can improve TDOA estimation in Practical Project can also reduce the requirement of remote reception station to local frequency phase equalization greatly, thus reaches the effect that not only can improve TDOA estimation precision but also can reduce system cost.
Accompanying drawing explanation
Fig. 1 is the two-dimensional search flow chart of embodiment 2.
Fig. 2 is embodiment 3 linear search flow chart.
Fig. 3 is the flow chart of embodiment 4.
Embodiment
Between the local oscillator of strange land frequency and the impact analysis of phase equalization error to time delay relation as follows:
The receiving station that suppose there is two diverse locations receives the signal of same emitter Signals after different delay and is respectively
with
here be with reference to receiving station with receiving station 1, estimate time difference τ
t, 2,1=τ
t, 2-τ
t, 1, the signal representation that so two receiving stations can be received is
with
here
f
cfor signal(-) carrier frequency; Conveniently express, order:
The emitter Signals received between the receiving station that obvious two spaces are separated has following time delay relation:
s
2(t)=s
1(t-τ
t,2,1)
Might not stringent synchronization between the carrier wave that the RF signal received in single receiver produces through down-conversion local oscillator and signal carrier, that is to say the mismatch that there is frequency and phase place.Assuming that there is carrier frequency mismatch amount in receiver 1 and receiver 2 is respectively f
1with f
2, local carrier initial phase is respectively
with
receiver 1 and receiver 2 receive emitter Signals and after down-conversion are
with
be expressed as being down-converted to the emitter Signals after base band:
Therefore the relation that can obtain between two-way baseband signal:
Here
f
2,1=f
2-f
1
Can find out that down-conversion signal does not meet x
2(t)=x
1(t-τ
t, 2,1) time delay relation; Under visual carrier frequency mismatch condition, between two receiver baseband signals, time delay relation there occurs change, and as can be seen from new time delay relation, works as f
2,1=0 and
time, that is to say f
1=f
2with
time, time delay relation is equivalent to desirable time delay relation, and causes the basic reason of desirable time delay relationship change to be the consistency of local carrier frequency and phase place between two receivers.And in Practical Project, accomplish that the high consistency of local carrier frequency between carrier synchronization strict in receiver or receiver and phase place is all very difficult.In the present invention, said baseband signal refers to only pass through the signal being down-converted to base band, not refers to the baseband signal through demodulation process.
Traditional time difference estimation method all adopts desirable time delay relational model to realize TDOA estimation usually.From the above analysis, the change of the time delay relation between two remote reception machine baseband signals makes these methods cannot estimate the time difference accurately.If new time delay relational model can be utilized.Estimate that the time difference will have greatly improved to the performance improving TDOA estimation, the frequency of receiver to local oscillator and the coherence request of phase place can also be reduced simultaneously, thus significant to radiation source location.
Embodiment 1
1. in radiation source location, to base band TDOA estimation, to analyze in down-conversion between the local oscillator of strange land frequency and phase equalization error to the impact of time difference relation, and obtaining the emitter Signals existed under the local oscillator conformity error of strange land, to be down-converted to the signal time difference relational model after base band as follows:
Be expressed as in digital baseband system:
Here x
i[n]=x
i(nT
δ), i=1,2, T
δfor the base band adjacent sample point time interval;
2., according to the baseband signal time difference relational model existed under local oscillator conformity error, give a kind of three-dimensional search maximum likelihood time difference estimation method based on maximum estimation principle of feeling relieved.
Three-dimensional search maximum likelihood time difference estimation method:
1) down-converted is carried out to the radio frequency rf signal of different receiving station and obtains baseband signal, and average and normalized are gone to baseband signal, obtain baseband digital signal as follows:
Receiving station 1 baseband signal: y
1[n]=x
1[n]+q
1[n]
Receiving station 2 baseband signal: y
2[n]=x
2[n]+q
2[n]
Here noise q
1[n] and q
2[n] is white Gaussian noise process.
2) according to the base band time difference relation that obtains above, the base band utilizing emitter Signals down-conversion to obtain represents without noise cancellation signal and is expressed as follows the baseband signal that receiving station 2 receives:
Due to signal x in real process
1[n] and x
1[n] is all unknown, and known is all noisy observation signal, therefore above formula observation signal can be expressed as:
Here
By as follows for each physical quantity vector expression:
y
1(τ
t,2,1)=y
1[0-τ
t,2,1T
δ]y
1[1-τ
t,2,1T
δ]…y
1[N-1-τ
t,2,1T
δ]
T
y
2=[y
2[0]y
2[1]y
2[2]…y
2[N-1]]
T
Definition matrix
as follows
The relation of vector expression two-way baseband signal is as follows:
Utilize maximum likelihood principle in this vector correlation formula and carry out abbreviation and obtain three-dimensional search likelihood estimator and be
Here
Namely to target function L
1carry out the accurate estimation that three-dimensional search just finally can realize the time difference, namely obtain TDOA estimation value by the time difference that the target function peak point searched is corresponding
three-dimensional search method is prior art, does not repeat at this.
Embodiment 2
Due to the maximum likelihood time difference estimation method provided, need the three-dimensional search that computation complexity is higher, and be unfavorable for Project Realization, so give the two-dimensional search after dimensionality reduction to realize TDOA estimation:
To matrix
carry out decomposition and obtain matrix Φ
f(f
2,1, τ
t, 2,1) and
I represents that diagonal entry is the unit matrix of 1 entirely.So be similar to correlation method time delay to estimate, we can obtain target function L
2:
So binary estimation objective function can be obtained
That is, the target function after dimensionality reduction is obtained
to target function L
3the time difference carrying out target function peak point that two-dimensional search obtains corresponding is TDOA estimation value
target function is when observation sample number is abundant simultaneously, has
at this moment also target function L can be derived
3(τ
t, 2,1, f
2,1) with
τ corresponding to peak value
t, 2,1with f
2,1identical.
Composition graphs 1, provides the specific implementation flow process of a two-dimensional search TDOA estimation, is divided into coarse search stage and precise search stage:
Step 201, starts two-dimensional search TDOA estimation algorithm;
Step 202, carries out TDOA estimation coarse search, and this search procedure is two-dimensional search process:
2021 arrange the time difference and frequency invariance error coarse search scope and precision, arrange time difference τ
t, 2,1hunting zone is [-τ
m, τ
m], frequency invariance error f
2,1hunting zone is [-f
m, f
m]; Arranging time difference search rough grade is T
i, frequency invariance error search rough grade is f
i; And with lower bound initialization time difference of hunting zone and frequency invariance error variance, i.e. τ
t, 2,1←-τ
m, f
2,1←-f
m, wherein ← represent assignment.
2022 judge the time difference whether in hunting zone, namely judge current τ
t, 2,1∈ [-τ
m, τ
m] whether set up, if set up, perform step 2023, otherwise perform step 2028.
2023 utilize the sinc function sample point that blocks as delay filter FIR weight coefficient and utilize the I/Q data of receiver 1 to be delayed signal y
1[n-τ
t, 2,1t
δ]:
Above as K → ∞, equation is just set up, and 2K+1 the sample point getting sinc function in Practical Project is similar to and realizes delay filter.
2024 judge ongoing frequency conformity error f
2,1whether in hunting zone, namely judge f
2,1∈ [-f
m, f
m] whether set up, if set up, perform step 2025, otherwise perform step 2027.
2025 calculating target functions
After enter next step.
2026 produce new frequency invariance error variance f according to the stepping of frequency invariance error search precision
2,1← f
2,1+ f
i, and jump to step 2024;
2027 produce new time difference variable τ according to time difference search precision stepping
t, 2,1← τ
t, 2,1+ T
i, and jump to step 2022.
Time difference τ corresponding to target function peak value is asked in 2028 search
t, 2,1with frequency invariance error f
2,1be designated as coarse search result
enter step 203.
Step 203, precise search process, this process is also two-dimensional search process, identical with the basic procedure in step 202:
2031 arrange the time difference and frequency invariance error precise search scope and precision, arrange time difference τ
t, 2,1hunting zone is
frequency invariance error f
2,1hunting zone is
arranging time difference search rough grade is T
j, frequency invariance error search rough grade is f
j; And with lower bound initialization time difference of hunting zone and frequency invariance error variance, namely
2032 judge the time difference whether in hunting zone, namely judge current
whether set up, if set up, perform step 2033, otherwise perform step 2038.
The delay filter of 2033 structures produces the inhibit signal y of the reception data of receiver 1
1[n-τ
t, 2,1t
δ].The present embodiment utilizes the sinc function sample point that blocks to be delayed signal to construct fractional delay filter FIR.
2034 judge ongoing frequency conformity error f
2,1whether in hunting zone, namely judge
whether set up, if set up, perform step 2035, otherwise perform step 2037.
2035 calculating target functions
After enter next step.
2036 produce new frequency invariance error variance f according to the stepping of frequency invariance error search precision
2,1← f
2,1+ f
j, and jump to step 2034;
2037 produce new time difference variable τ according to time difference search precision stepping
t, 2,1← τ
t, 2,1+ T
j, and jump to step 2032.
Time difference τ corresponding to target function peak value is asked in 2038 search
t, 2,1with frequency invariance error f
2,1be designated as the estimated result τ finally searched
for estimating the time difference, enter end step 204.
Step 204: TDOA estimation is complete, stops TDOA estimation algorithm.
Two-dimensional search is not limited to and first carries out the searching method that coarse search then carries out precise search, also only can carry out the search in a stage in actual applications, can according to Practical Project need hunting zone and search precision are set.
Embodiment 3
The two-dimensional search problem of three-dimensional search be converted into above greatly reduces operand to realize TDOA estimation, if also need compaction algorithms amount to improve estimating speed, it is the method for linear search that the present embodiment provides further for two-dimensional search dimensionality reduction:
When not considering noise, the derivation from above:
Consider the impact of noise, have approximate expression below:
Observation data can be utilized above on the right of equation directly to calculate, describe in order to convenient below, be called indirect operation data here. this equation forms a linear regression model (LRM), unknown frequency conformity error f
2,1for model parameter to be estimated, utilize least square method can estimate this parameter as follows:
Here
for indirect operation data.So just certain time difference place obtain the estimation of frequency invariance error
and do not need to estimate f by the mode of search
2,1, thus two-dimensional search algorithm is above down to 1 dimension search.
Composition graphs 2 illustrates that linear search TDOA estimation algorithm flow is as follows:
Step 301, starts linear search TDOA estimation algorithm.
Step 302, arranges time difference hunting zone and precision, as arranged time difference hunting zone is
arranging time difference search precision is T
j, and with the lower bound initialization time difference variable of hunting zone, namely
here
for the estimated result of two-dimensional search TDOA estimation algorithm or the estimated result of a front linear search TDOA estimation algorithm or default initial value.
Step 303, judges current time difference τ
t, 2,1whether in hunting zone, namely judge
whether set up, perform step 304 if set up, otherwise perform step 309.
Step 304, calculates y
1[n-τ
t, 2,1t
δ], its computational process is identical with two-dimensional search TDOA estimation algorithm.
Step 305, calculates indirect operation amount z [n]:
Step 306, estimated frequency conformity error:
Here intermediate quantity
Step 307, utilizes current time difference τ
t, 2,1frequency invariance error is obtained with estimation
calculating target function L
3,
Step 308, produces new time difference variable τ according to time difference search precision stepping
t, 2,1← τ
t, 2,1+ T
j, and jump to step 303.
Step 309, TDOA estimation value corresponding to target function peak value is asked in search
Step 310, TDOA estimation is complete, stops TDOA estimation algorithm.
Embodiment 4
Not only in order to accurate but also realize TDOA estimation rapidly in engineering, this example is in conjunction with two-dimensional search method and linear search method, and process is as follows:
Step 401, prepares the I/Q base band data needed for TDOA estimation:
Step 4011 receiver 1 intercepted signal is also down-converted to baseband I/Q data;
Step 4012 receiver 2 intercepted signal is also down-converted to baseband I/Q data;
Baseband I/Q transfer of data arrival time difference the measuring equipment at step 4013 receiver 1 and receiver 2 place.
Step 402, starts TDOA estimation.
Step 403, select the concrete grammar of TDOA estimation:
4031 judge whether it is first TDOA estimation, if perform step 4033, if not perform step 4032;
4032 judge whether base band signal to noise ratio (SNR) is greater than 12dB, if perform step 4034, if not perform 4033;
Step 4033 is two-dimensional search time difference estimation method; Step 4034 is linear search time difference estimation method.