CN103490824A - Reference signal recovery method for EVM analysis - Google Patents

Abstract

The invention discloses a reference signal recovery method for EVM analysis. The reference signal recovery method for EVM analysis is characterized by comprising the following steps that a if the symbol rate of a measurement signal is unknown, the symbol rate is roughly estimated according to wavelet transformation, then, the step b is executed, and if the symbol rate is known, the step b is directly executed; b a code element jump point of the measurement signal is extracted according to wavelet transformation, and then the step c is executed; c a reference signal is worked out based on the extracted code element jump point. According to the reference signal recovery method for EVM analysis, the symbol jump point of the signal is detected according to wavelet alignment, so that accurate alignment of a reference signal code element and a measurement signal code element in time domain is achieved, the jump distance is calculated, the symbol length is estimated, the reciprocal value of the symbol length is obtained, and therefore the symbol rate is accurately obtained; then a extreme point is used as a symbol switching point, the phase mean value between two jump points is judged, the phase mean value is used as a phase value between jumps, and therefore the reference signal is recovered. By the adoption of the reference signal recovery method for EVM analysis, the influence of accumulated errors caused by inaccurate estimation of the initial symbol rate can be eliminated, and the high accuracy and the high anti-noise performance can be achieved.

Description

The reference signal restoration methods that a kind of EVM analyzes

Technical field

The present invention relates to the reference signal restoration methods that a kind of EVM analyzes.

Background technology

In communication system, an important parameter weighing the modulation signal quality is Error Vector Magnitude (EVM:error vector magnitude), and the modulation quality that need not use a plurality of parameters to transmit to transmitter is assessed.EVM has characterized the error vector of measuring-signal with reference signal, carrys out the modulation accuracy of measuring-signal by calculating.Usually adopt the interior chip of one or more time slots or the error vector mean-square calculation EVM of symbol, the EVM value is less, illustrates that the modulation quality that transmits is better, otherwise, illustrate that modulation quality is poorer.The calculating object of EVM also can be according to the difference of modulation system and is different.As QPSK (quarternary phase-shift keying (QPSK)) modulation adopted in communication system, it is a kind of constant enveloped modulation, also there is error on amplitude due to it, phase error and frequency difference have been not enough to react its modulation accuracy, therefore, the index that needs a kind of gauge signal range error comprehensively and phase error.On planisphere, error vector is the degree of injury of reflected signal clearly, so can obtain EVM by the error of calculating amplitude.And the signal of picture GMSK (Gaussian minimum shift keyed) modulation, signal intensity only is embodied in the angle variation, so EVM obtains by planisphere upper angle error.Measurement signal Z (t) is exactly the signal that actual measurement arrives that picks, and reference signal R (t) carries out ideal Modulated to measurement signal Z (t) through the result of demodulation to obtain again.In conjunction with reference to figure 3, meaned the relation of they and EVM in this figure.

Realize the EVM of signal is analyzed, must be according to the symbol recovery reference signal of demodulation, EVM method of measurement is traditionally generally measured EVM by steps such as reception, modulation, calculating.At first, receive the signal that signal transmitter sends, according to receiving signal, determine measured signal.Then, the signal from measured signal is determined modulation, by modulation, obtain reference vector signal.Finally, utilize measured signal and reference vector signal, the Error Vector Magnitude index of transmitter computes.Generally, be all that the parameter informations such as symbol rate according to measuring-signal, amplitude carry out ideal Modulated again and obtain reference signal.

Traditional EVM analyzes, and is that the information such as symbol rate according to measuring-signal, amplitude are carried out ideal Modulated and obtained reference signal.At present, the algorithm of estimate symbol rate is broadly divided into: envelope square spectrometry, Wavelet Transform, the distortion of Circular correlation method and these methods and combination etc.But, which kind of method all needs more accurate estimate symbol rate, and, under non-collaboration communication, if any method is carried out the reference signal recovery by the inadequate closely coincide rate of estimating, error on time domain all can accumulate in time, has a strong impact on the analysis precision of EVM, even obtains wrong analysis result.Carrying out at present the EVM analysis is all the accuracy of striving improving the estimate symbol rate basically, and then recovers reference signal.The method of estimate symbol rate mainly contains:

(1) the psk signal symbol rate estimation method based on instantaneous amplitude.The signals such as general ASK, PSK, because the impact of molding filtration, the envelope temporal evolution, comprised the symbol rate component.The intermediate-freuqncy signal quadrature frequency conversion, obtain in-phase component and the quadrature component of base band, and calculate signal envelope square, calculate frequency spectrum after windowing.Generally, according to maximum alternating current component, corresponding frequency just can the estimate symbol rate.Method based on instantaneous amplitude, it utilizes Hilbert conversion to extract the instantaneous amplitude of signal, utilizes the FFT of instantaneous amplitude to estimate the symbol rate of psk signal.Although this method operand is lower, its noiseproof feature is bad, and practicality is not strong.

(2) utilize small wave converting method estimate symbol rate, as the symbol rate based on the Harr wavelet transformation is estimated.Hopping edge to signal is detected, carry out the different Haar wavelet transformation of yardstick, in the middle of symbol period, amplitude after wavelet transformation is substantially equal, can both produce spike on the symbol period edge that receives signal, and the position of spike is identical, superposeing after the amplitude square after these different scale wavelet transformations, the rated output spectrum, detect spectrum energy, then estimate symbol rate again.The method has good performance when signal to noise ratio is higher, but affected by noise very large under the low signal-to-noise ratio environment.And this method is not suitable for the signal that adopts raised cosine to be shaped, to dissimilar signal, automatically choosing of wavelet transform dimension is a difficult problem always.

(3) symbol rate estimating method based on Cyclic Spectrum.But the amount of calculation of Cyclic Spectrum is larger, and not often global maximum corresponding to the spectral line of symbol rate.And a kind of non-linear filtering method of Background suppression noise carries out filtering to Cyclic Spectrum, the method can make to highlight corresponding to the spectral line of symbol rate, becomes global maximum, is conducive to the estimation of symbol rate.And, for bandwidth efficient channel, usually require auto-correlation γ > 50, its operand is too large, is difficult to practicality.And algorithm does not change the inherent defect of rudimentary algorithm, when signal to noise ratio is low, the symbol rate spectral line, likely lower than the low frequency spectral line, therefore needs the scope of supplementary means qualifying symbol rate, equally to improve correct expectancy rate.

(4) utilize the method for bandpass signal square spectrum estimate symbol rate.Utilize square spectrum to estimate the symbol rate of m-PSK signal, the operand of square spectrum will be far smaller than Cyclic Spectrum.While for signal, adopting raised cosine to be shaped in square spectrum corresponding to the spectral line of the symbol rate distant situation that will become, use one group of filter to carry out filtering to square spectrum, then carry out the method for comprehensive judgement.This method can improve the performance that symbol rate is estimated, but has increased computational complexity.A kind of method of square spectrum estimate symbol rate of intermediate-freuqncy signal of utilizing solves the problem that reduces computational complexity, but has reduced noiseproof feature simultaneously.

Above-mentioned simplified method is all accurately to be estimated as prerequisite, the prerequisite of analyzing as EVM with symbol rate.Although more or less improved the accuracy of estimate symbol rate, the method for these estimate symbol rates has limitation, and symbol rate estimates that noiseproof feature is also lower.Therefore, no matter which kind of method to be recovered reference signal by, the frequency deviation evaluated error, will cause the baseband signal obtained to have phase drift, and the symbol rate error will cause reference signal can't synchronize with measuring-signal.Error on time domain also can accumulate in time, has a strong impact on the analysis precision of EVM, even obtains wrong analysis result.

Summary of the invention

The reference signal restoration methods that provides a kind of new EVM to analyze is provided task of the present invention, it uses a kind of method of small echo alignment, carry out the symbol rate estimation with small echo, then utilize Wavelet Detection symbol trip point, make the reference signal code element realize on time domain with the measuring-signal code element precisely aliging, then the saltus step spacing is carried out to statistical computation, and estimate symbol length, get its inverse and can accurately draw symbol rate.Thereby the impact that the estimation inaccuracy of having eliminated symbol rate is brought.

Its technical solution is:

The reference signal restoration methods that a kind of EVM analyzes comprises the following steps:

If the symbol rate of a measuring-signal is unknown, with wavelet transformation, ask symbol rate, then perform step b; If the symbol rate of measuring-signal is known, directly perform step b,

B extracts the code element trip point of measuring-signal by the method for wavelet transformation; Then perform step c,

C obtains reference signal according to the code element trip point extracted.

In above-mentioned steps b, measuring-signal is carried out to wavelet transformation, transformed value can produce maximum at the symbol trip point, and the integral multiple that the spacing of maximum is symbol period, for this characteristic, wavelet transformation detection method in conjunction with bandwidth estimation, utilize the Singularity Detection characteristic of small echo, the code element trip point detected, carry out statistical computation to the saltus step spacing, estimate symbol length, thus its valuation that draws the signal code rate reciprocal got.

Above-mentioned steps b specifically comprises the steps:

B1 carries out wavelet transformation to measuring-signal

If in the Frequency Hopping Signal time slot, the signal resolution expression formula is:

$x\left(t\right)=s\left(t\right)+n\left(t\right)=\tilde{s}\left(t\right)e^{j({\mathrm{\ω}}_{c}t+{\mathrm{\θ}}_c)}+n\left(t\right)$

Wherein, x (t) is the complex signal in time slot, and s (t) is modulated complex signal, and n (t) is additive white Gaussian noise, ω _cfor time slot carrier frequency angular frequency, θ _cfor time slot carrier frequency initial phase,

for baseband signal;

For the m-PSK digital signal, can be expressed as:

$\tilde{s}\left(t\right)=A\underset{n}{\mathrm{\Σ}}{e}^{j{\mathrm{\φ}}_n}u(t-n{T}_s)$

Wherein, A is signal amplitude, φ _nfor the phase place of n element of signal, u (t) is unit step function, T _sfor symbol lengths;

Adopt continuous wavelet transform:

$\mathrm{CWT}(a,\mathrm{\τ})=\frac{1}{\sqrt{a}}\∫s\left(t\right){\mathrm{\ψ}}^{*}\left(\frac{t-\mathrm{\τ}}{a}\right)\mathrm{dt}$

Wherein, s (t) is measured signal, and ψ (t) is mother wavelet function, ^*mean conjugation, a is yardstick, and τ is displacement;

Above-mentioned small echo is selected the haar small echo, and it is expressed as follows:

During to the wavelet transformation of m-PSK digital signal, be divided in the conversion of phase place continuum with in two kinds of situations of conversion of phase place discontinuity interval,

When the phase place continuum:

$\left|\mathrm{CWT}(a,\mathrm{\τ})\right|=\frac{A}{\sqrt{a}}\frac{4}{{\mathrm{\ω}}_c}{\sin }^2\frac{a}{4}{\mathrm{\ω}}_c$

At the phase place discontinuity interval,

be respectively the phase place of PSK n and n+1 code element, and change is at the d place when (d<0):

Similar result is arranged when d>0;

In one-period or code element in identical cycle, the mould value of wavelet coefficient | CWT (a, τ) | be a constant, when symbol generation saltus step, | CWT (a, τ) | also can change thereupon, and at the trip point place, | CWT (a, τ) | can produce a maximum;

B2 selects wavelet scale and extracts extreme point

Consider symbol rate R and its bandwidth B of signal _wpass be R=kB _w, (1<k<2), before signal is carried out to wavelet transformation, first carry out bandwidth estimation to signal, chooses

as wavelet transform dimension, wherein

estimated value for bandwidth;

The pseudo-extreme point that rises and falls and cause in order to eliminate noise and envelope, extract modulus maximum, chooses a threshold value thresholding T _hright | CWT (a, τ) | carry out denoising, for bpsk signal, choose T _h=max (| CWT (a, τ) |)/2 are as the denoising thresholding, by the signal wavelet module value after the thresholding denoising

can be expressed as:

$\left|\widetilde{\mathrm{CWT}(a,\mathrm{\&tau;})}\right|=\left\{\begin{array}{cc}\left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|& \left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|\&GreaterEqual;T_h\\ 0& \left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|<T_h\end{array}\right.$

With 0.5/B _wfor length of window pair

extract local maximum and can orient symbol trip point P _i, (1<i<N), i means the sequence number of trip point, N means the number of trip point;

B3 estimated signal symbol rate

Length L between note adjacent hop height _i=(P _i+1-P _i), (1<i<N-1), L _ibe the length between i section adjacent hop height, in the ideal situation, L _ifor symbol lengths T _sintegral multiple: L _i=kT _s(k=1,2,3 ...); First trip point is to the total symbol numbers of last trip point

its duration T=P _n-P ₁; Due to L _i=T _smaximum probability, thereby choose the L that maximum probability occurs _ivalue, as the rough estimate evaluation of symbol lengths, is designated as

Employing length is less than

l _icalculate total symbol numbers, statistical duration:

$\tilde{M}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}\mathrm{round}(L_i/{\tilde{T}}_s),(L_i<4{\tilde{T}}_s),$

$\tilde{T}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}{L}_i,(L_i<4{\tilde{T}}_s)$

Wherein round is the round operation,

with

ratio be the estimated value of signal code rate;

B4 extracts the code element trip point

With a=c/R, (0.4<c<1), as wavelet transform dimension, does wavelet transformation again to signal, and process is as b2, and the estimated value that wherein R is the signal code rate, detect the code element saltus step.

Upper step c comprises: using extreme point as the symbol switching point, the phase place average between two trip points is adjudicated as the phase value between saltus step, using the amplitude average of signal as amplitude, obtain reference signal.

Above-mentioned steps b3 also comprises following concrete steps:

B31 asks bandwidth B to signal _w, frequency deviation F _c, go frequency deviation, obtain baseband signal;

B32 with

for yardstick carries out continuous wavelet transform to baseband signal and delivery obtains | CWT (a, τ) |;

B33 is right | CWT (a, τ) | and do the thresholding denoising, obtain new mould value sequence

B34 couple

with 0.5/B _wfor window width location Wavelet Modulus Maxima, obtain symbol trip point P _i, (1<i<N);

B35 asks the distance L between every two adjacent hop heights _i=(P _i+1-P _i), (1<i<N-1), to trip point spacing L _iask histogram, the corresponding L with the histogram maximum _ias symbol lengths rough estimate evaluation

B36 chooses length and is less than

l _iadd up total symbol numbers and duration thereof $\tilde{M}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}\mathrm{round}(L_i/{\tilde{T}}_s),(L_i<{\tilde{T}}_s),$ $\tilde{T}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}{L}_i,(L_i<4{\tilde{T}}_s);$

B37 passes through formula

calculate the estimated value of signal code rate.

The present invention has following useful technique effect:

The present invention uses the method for small echo alignment, wavelet transformation detection method in conjunction with bandwidth estimation, utilize the Singularity Detection characteristic of small echo, and utilize Wavelet Detection symbol trip point, make the reference signal code element realize on time domain with the measuring-signal code element precisely aliging, the saltus step spacing is carried out to statistical computation, estimate symbol length, thereby get its symbol rate that draws reciprocal, using extreme point as the symbol switching point, phase place average between two trip points is judged as the phase value between saltus step, usingd the amplitude average of signal as amplitude, just can be obtained reference signal.Such reference signal, can effectively eliminate because the reference signal that the symbol rate evaluated error causes is recovered mistake.

The present invention is by adopting Wavelet Detection symbol trip point, make the reference signal code element realize on time domain with the measuring-signal code element precisely aliging, thereby recovery reference signal, not only there is very high accuracy and noise immunity, even symbol rate is estimated to certain deviation, also can provide rational reference signal, erasure signal is estimated coarse impact.

The accompanying drawing explanation

Below in conjunction with accompanying drawing and embodiment, the present invention is further described:

Fig. 1 shows a kind of collaboration communication baseband signal EVM analysis process schematic block diagram of the prior art.

Fig. 2 shows a kind of non-collaboration communication baseband signal EVM analysis process schematic block diagram of the prior art.

Fig. 3 shows error vector signal definition schematic diagram of the prior art.

The schematic process flow diagram that Fig. 4 is one embodiment of the present invention.

Embodiment

Realize the EVM of signal is analyzed, must be according to the symbol recovery reference signal of demodulation, and the recovery precision of reference signal has affected the EVM analysis result on large program very much.Reference signal is recovered difficulty, measuring accuracy is not high, this is the difficult problem that puzzlement EVM analyzes always, and traditional EVM analyzes, and is that the information such as symbol rate according to measuring-signal, amplitude are carried out ideal Modulated and obtained reference signal, but, under non-collaboration communication, if carry out the reference signal recovery by the accurate not symbol rate of estimating, the error on time domain can accumulate in time, have a strong impact on the analysis precision of EVM, even obtain wrong analysis result.

As shown in Figure 1, in collaboration communication, input signal 101 obtains baseband signal 102 through down-converted, baseband signal obtains two paths of signals through demodulation process, one tunnel is bit information 103, and bit information obtains reference signal 104 through ideal Modulated, and reference signal is sent into EVM and calculated 105, another road is measuring-signal 106, and measuring-signal is sent into EVM and calculated; The symbol rate adopted in above-mentioned ideal Modulated process is known in advance exact value, can directly by modulation, obtain reference signal.In non-collaboration communication, as shown in Figure 2, input signal 201 is carried out to carrier frequency and estimate 202, the signal after carrier frequency is estimated obtains baseband signal 203 through down-converted, baseband signal is carried out to symbol rate and estimate 204, then the estimated value of symbol rate is carried out to demodulation process and obtain two paths of signals, one tunnel is bit information 205, and bit information obtains reference signal 206 through ideal Modulated, and reference signal is sent into EVM and calculated 207, another road is measuring-signal 208, and measuring-signal is sent into EVM and calculated; The information such as above-mentioned symbol rate are to obtain according to the Signal estimation obtained, and the estimated accuracy of these parameters has influence on the computational accuracy of EVM to a great extent.The frequency deviation evaluated error, will cause the baseband signal obtained to have phase drift, and the symbol rate error will cause reference signal can't synchronize with measuring-signal.

To this, the present invention proposes a kind of new EVM analytical method, use a kind of method of small echo alignment.The method is utilized Wavelet Detection symbol trip point, makes the reference signal code element realize on time domain with the measuring-signal code element precisely aliging, and finally recovers reference signal.Use this method, even, to the certain deviation of estimating at of symbol rate, also can provide rational reference signal, erasure signal is estimated coarse impact.

Below in conjunction with Fig. 4, the present invention is described in further detail.

The present invention roughly flow process is as follows:

Input signal 401 is judged, judge whether it is known symbol rate 402, if be judged as "Yes", to this signal, utilize continuous wavelet transform to search extreme point 403, selected subsequently wavelet scale and extracted extreme point 404, and then carry out detected symbol trip point 405, finally obtain and recover reference signal 406; If be judged as "No", utilize continuous wavelet transform to ask symbol rate 407, then selected wavelet scale and extract extreme point 404, carry out subsequently detected symbol trip point 405, finally obtain and recover reference signal 406.

The present invention adopts the method for wavelet transformation to extract the code element trip point, uses small wave converting method estimate symbol rate, chooses a=c/R, and (0.4<c<1) as wavelet transform dimension, the estimated value that wherein R is the signal code rate.The method has higher accuracy and noise immunity.Can effectively resist the impact of high-frequency noise like this, guarantee to detect the code element saltus step.Using extreme point as the symbol switching point, the phase place average between two trip points is adjudicated, as the phase value between saltus step, using the amplitude average of signal as amplitude, just can obtain reference signal.Using this signal as the reference signal, can effectively eliminate because the reference signal that the symbol rate evaluated error causes is recovered mistake.

The basic step of the recovery reference signal that the present invention realizes is:

(1) if the symbol rate of measuring-signal is unknown, with wavelet transformation, ask symbol rate, then execution step (2).If the symbol rate of measuring-signal is known, directly perform step (2).

(2) extract the code element trip point of measuring-signal by the method for wavelet transformation.Then perform step (3),

(3) obtain reference signal according to the code element trip point extracted.

Be implemented as follows:

(1) when symbol rate the unknown of measuring-signal, first use wavelet transformation estimate symbol rate, then execution step (2).If the symbol rate of known measuring-signal, directly carry out (skipping to) step (2).

Signal is carried out to wavelet transformation, transformed value can produce maximum at the symbol trip point, the integral multiple that the spacing of maximum is symbol period, for this characteristic, in conjunction with the wavelet transformation detection method of bandwidth estimation, utilize the Singularity Detection characteristic of small echo, the code element trip point detected, the saltus step spacing is carried out to statistical computation, estimate symbol length, thus get its estimated value that draws the signal code rate reciprocal.Detailed process is as follows:

(1.1) to the signal wavelet transformation

If in the Frequency Hopping Signal time slot, the signal resolution expression formula is:

$x\left(t\right)=s\left(t\right)+n\left(t\right)=\tilde{s}\left(t\right)e^{j({\mathrm{\&omega;}}_{c}t+{\mathrm{\&theta;}}_c)}+n\left(t\right)---\left(4.1\right)$

Wherein, x (t) is the complex signal in time slot, and s (t) is modulated complex signal, and n (t) is additive white Gaussian noise, ω _cfor time slot carrier frequency angular frequency, θ _cfor time slot carrier frequency initial phase.

for baseband signal.

For the m-PSK signal, can be expressed as:

$\tilde{s}\left(t\right)=A\underset{n}{\mathrm{\&Sigma;}}{e}^{j{\mathrm{\&phi;}}_n}u(t-n{T}_s)---\left(4.2\right)$

Wherein, A is signal amplitude, φ _nfor the phase place of n element of signal, u (t) is unit step function, T _sfor symbol lengths.

Adopt continuous wavelet transform:

$\mathrm{CWT}(a,\mathrm{\&tau;})=\frac{1}{\sqrt{a}}\&Integral;s\left(t\right){\mathrm{\&psi;}}^{*}\left(\frac{t-\mathrm{\&tau;}}{a}\right)\mathrm{dt}.---\left(4.3\right)$

Wherein, s (t) is measured signal, and ψ (t) is mother wavelet function, ^*mean conjugation.A is yardstick, and τ is displacement.

Small echo is selected the haar small echo, and because of it, to transient signal, especially the variation of phase signal has stronger detectability, and it is expressed as follows:

During to the wavelet transformation of m-PSK digital signal, be divided in the conversion of phase place continuum with in two kinds of situations of conversion of phase place discontinuity interval.

When the phase place continuum:

$\left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|=\frac{A}{\sqrt{a}}\frac{4}{{\mathrm{\&omega;}}_c}{\sin }^2\frac{a}{4}{\mathrm{\&omega;}}_c---\left(4.5\right)$

At the phase place discontinuity interval,

be respectively the phase place of PSK n and n+1 code element, and change is at the d place when (d<0):

Similar result is arranged when d>0.

From above-mentioned two formulas, in one-period or code element in identical cycle, the mould value of wavelet coefficient | CWT (a, τ) | be a constant, when symbol generation saltus step, | CWT (a, τ) | also can change thereupon, and at the trip point place, | CWT (a, τ) | can produce a maximum.

(1.2) select wavelet scale and extract extreme point

Correctly choosing wavelet scale is the key of extracting the symbol trip point, and wavelet scale is selected less, and the precision of estimation is just higher, but more easily is subject to the impact of high-frequency noise, and stability is the phase strain differential also.Wavelet scale is selected larger, estimates that stability is just better, yet, when exceeding Baud Length, yardstick just can not detect the symbol saltus step.Consider symbol rate R and its bandwidth B of signal _wpass be R=kB _w, (1<k<2), before signal is carried out to wavelet transformation, first carry out bandwidth estimation to signal, chooses

as wavelet transform dimension, wherein

estimated value for bandwidth.

The pseudo-extreme point that rises and falls and cause in order to eliminate noise and envelope, extract modulus maximum, chooses a threshold value thresholding T _hright | CWT (a, τ) | carry out denoising, for bpsk signal, choose T _h=max (| CWT (a, τ) |)/2 are as the denoising thresholding, by the signal wavelet module value after the thresholding denoising can be expressed as:

$\left|\widetilde{\mathrm{CWT}(a,\mathrm{\&tau;})}\right|=\left\{\begin{array}{cc}\left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|& \left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|\&GreaterEqual;T_h\\ 0& \left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|<T_h\end{array}\right.---\left(4.7\right)$

With 0.5/B _wfor length of window pair

extract local maximum and can orient symbol trip point P _i, (1<i<N), i means the sequence number of trip point, N means the number of trip point.

(1.3) estimate symbol rate

Length L between note adjacent hop height _i=(P _i+1-P _i), (1<i<N-1), L _ibe the length between i section adjacent hop height, known, under perfect condition, L _ifor symbol lengths T _sintegral multiple: L _i=kT _s(k=1,2,3 ...).First trip point is to the total symbol numbers of last trip point

its duration T=P _n-P ₁.Due to L _i=T _smaximum probability, thereby we choose the L that maximum probability occurs _ivalue, as the rough estimate evaluation of symbol lengths, is designated as

To actual signal, due to noise effect, L _ibe not generally symbol lengths T _sintegral multiple, symbol lengths rough estimate evaluation have equally error, we select the round value as the symbol numbers between the adjacent hop height, but a plurality of consecutive identical code elements can cause error accumulation, consecutive identical code element number is more, and symbol numbers evaluated error probability is just larger, for avoiding as far as possible this cumulative errors, we give up length and are more than or equal to l _i, only adopt length to be less than

l _icalculate total symbol numbers, statistical duration:

$\tilde{M}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}\mathrm{round}(L_i/{\tilde{T}}_s),(L_i<4{\tilde{T}}_s),---\left(4.8\right)$

$\tilde{T}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}{L}_i,(L_i<4{\tilde{T}}_s)---\left(4.9\right)$

Wherein round is the round operation.

with

ratio be symbol rate.

Carry out above process, concrete algorithm steps is as follows:

(a) signal is asked to bandwidth B _w, frequency deviation F _c, go frequency deviation, obtain baseband signal;

(b) with for yardstick carries out continuous wavelet transform to baseband signal and delivery obtains | CWT (a, τ) |;

(c) right | CWT (a, τ) | do the thresholding denoising, obtain new mould value sequence

(d) right

with 0.5/B _wfor window width location Wavelet Modulus Maxima, obtain symbol trip point P _i, (1<i<N);

(e) ask the distance L between every two adjacent hop heights _i=(P _i+1-P _i), (1<i<N-1), to trip point spacing L _iask histogram, the corresponding L with the histogram maximum _ias symbol lengths rough estimate evaluation

(f) choosing length is less than

l _iadd up total symbol numbers and duration thereof $\tilde{T}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}{L}_i,(L_i<4{\tilde{T}}_s)$

(g) pass through formula

the compute sign rate.

(2) extract the code element trip point by the method for wavelet transformation.

With a=c/R, (0.4<c<1), as wavelet transform dimension, does wavelet transformation again to signal, and process is as (1.2), the estimated value that wherein R is the signal code rate.Can effectively resist the impact of high-frequency noise like this, guarantee to detect the code element saltus step.

(3) obtain reference signal according to trip point.

Using extreme point as the symbol switching point, the phase place average between two trip points is adjudicated as the phase value between saltus step, using the amplitude average of signal as amplitude, we just can obtain reference signal.Using this signal as the reference signal, can effectively eliminate because the reference signal that the symbol rate evaluated error causes is recovered mistake.

The relevant technologies content of not addressing in aforesaid way is taked or is used for reference prior art and can realize.

It should be noted that, under the instruction of this specification, those skilled in the art can also make such or such easy variation pattern, such as equivalent way, or obvious mode of texturing.Above-mentioned variation pattern all should be within protection scope of the present invention.

Claims (5)

1. the reference signal restoration methods that an EVM analyzes is characterized in that comprising the following steps:

If the symbol rate of a measuring-signal is unknown, with wavelet transformation, ask symbol rate, then perform step b; If the symbol rate of measuring-signal is known, directly perform step b,

B extracts the code element trip point of measuring-signal by the method for wavelet transformation; Then perform step c,

C obtains reference signal according to the code element trip point extracted.

2. the reference signal restoration methods that a kind of EVM according to claim 1 analyzes, it is characterized in that: in above-mentioned steps b, measuring-signal is carried out to wavelet transformation, transformed value can produce maximum at the symbol trip point, the integral multiple that the spacing of maximum is symbol period, for this characteristic, wavelet transformation detection method in conjunction with bandwidth estimation, utilize the Singularity Detection characteristic of small echo, the code element trip point detected, the saltus step spacing is carried out to statistical computation, estimate symbol length, thus get its valuation that draws the signal code rate reciprocal.

3. the reference signal restoration methods that a kind of EVM according to claim 2 analyzes, it is characterized in that: described step b specifically comprises the steps:

B1 carries out wavelet transformation to measuring-signal

If in the Frequency Hopping Signal time slot, the signal resolution expression formula is:

$x\left(t\right)=s\left(t\right)+n\left(t\right)=\tilde{s}\left(t\right)e^{j({\mathrm{\&omega;}}_{c}t+{\mathrm{\&theta;}}_c)}+n\left(t\right)$

Wherein, x (t) is the complex signal in time slot, and s (t) is modulated complex signal, and n (t) is additive white Gaussian noise, ω _cfor time slot carrier frequency angular frequency, θ _cfor time slot carrier frequency initial phase,

for baseband signal;

For the m-PSK digital signal, can be expressed as:

$\tilde{s}\left(t\right)=A\underset{n}{\mathrm{\&Sigma;}}{e}^{j{\mathrm{\&phi;}}_n}u(t-n{T}_s)$

Wherein, A is signal amplitude, φ _nfor the phase place of n element of signal, u (t) is unit step function, T _sfor symbol lengths;

Adopt continuous wavelet transform:

$\mathrm{CWT}(a,\mathrm{\&tau;})=\frac{1}{\sqrt{a}}\&Integral;s\left(t\right){\mathrm{\&psi;}}^{*}\left(\frac{t-\mathrm{\&tau;}}{a}\right)\mathrm{dt}$

Wherein, s (t) is measured signal, and ψ (t) is mother wavelet function, ^*mean conjugation, a is yardstick, and τ is displacement;

Above-mentioned small echo is selected the haar small echo, and it is expressed as follows:

During to the wavelet transformation of m-PSK digital signal, be divided in the conversion of phase place continuum with in two kinds of situations of conversion of phase place discontinuity interval,

When the phase place continuum:

$\left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|=\frac{A}{\sqrt{a}}\frac{4}{{\mathrm{\&omega;}}_c}{\sin }^2\frac{a}{4}{\mathrm{\&omega;}}_c$

At the phase place discontinuity interval,

be respectively the phase place of PSK n and n+1 code element, and change is at the d place when (d<0):

Similar result is arranged when d>0;

In one-period or code element in identical cycle, the mould value of wavelet coefficient | CWT (a, τ) | be a constant, when symbol generation saltus step, | CWT (a, τ) | also can change thereupon, and at the trip point place, | CWT (a, τ) | can produce a maximum;

B2 selects wavelet scale and extracts extreme point

Consider symbol rate R and its bandwidth B of signal _wpass be R=kB _w, (1<k<2), before signal is carried out to wavelet transformation, first carry out bandwidth estimation to signal, chooses

as wavelet transform dimension, wherein

estimated value for bandwidth;

The pseudo-extreme point that rises and falls and cause in order to eliminate noise and envelope, extract modulus maximum, chooses a threshold value thresholding T _hright | CWT (a, τ) | carry out denoising, for bpsk signal, choose T _h=max (| CWT (a, τ) |)/2 are as the denoising thresholding, by the signal wavelet module value after the thresholding denoising

can be expressed as:

$\left|\widetilde{\mathrm{CWT}(a,\mathrm{\&tau;})}\right|=\left\{\begin{array}{cc}\left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|& \left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|\&GreaterEqual;T_h\\ 0& \left|\mathrm{CWT}(a,\mathrm{\&tau;})\right|<T_h\end{array}\right.$

With 0.5/B _wfor length of window pair

extract local maximum and can orient symbol trip point P _i, (1<i<N), i means the sequence number of trip point, N means the number of trip point;

B3 estimated signal symbol rate

Length L between note adjacent hop height _i=(P _i+1-P _i), (1<i<N-1), L _ibe the length between i section adjacent hop height, in the ideal situation, L _ifor symbol lengths T _sintegral multiple: L _i=kT _s(k=1,2,3 ...); First trip point is to the total symbol numbers of last trip point

its duration T=P _n-P ₁; Due to L _i=T _smaximum probability, thereby choose the L that maximum probability occurs _ivalue, as the rough estimate evaluation of symbol lengths, is designated as

Employing length is less than

l _icalculate total symbol numbers, statistical duration:

$\tilde{M}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}\mathrm{round}(L_i/{\tilde{T}}_s),(L_i<4{\tilde{T}}_s),$

$\tilde{T}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}{L}_i,(L_i<4{\tilde{T}}_s)$

Wherein round is the round operation,

with

ratio be the estimated value of signal code rate;

B4 extracts the code element trip point

With a=c/R, (0.4<c<1), as wavelet transform dimension, does wavelet transformation again to signal, and process is as b2, and the estimated value that wherein R is the signal code rate, detect the code element saltus step.

4. the reference signal restoration methods that a kind of EVM according to claim 3 analyzes, it is characterized in that: upper step c comprises: using extreme point as the symbol switching point, phase place average between two trip points is adjudicated as the phase value between saltus step, using the amplitude average of signal as amplitude, obtain reference signal.

5. the reference signal restoration methods that a kind of EVM according to claim 3 analyzes, it is characterized in that: above-mentioned steps b3 also comprises following concrete steps:

B31 asks bandwidth B to signal _w, frequency deviation F _c, go frequency deviation, obtain baseband signal;

B32 with

for yardstick carries out continuous wavelet transform to baseband signal and delivery obtains | CWT (a, τ) |;

B33 is right | CWT (a, τ) | and do the thresholding denoising, obtain new mould value sequence

B34 couple

with 0.5/B _wfor window width location Wavelet Modulus Maxima, obtain symbol trip point P _i, (1<i<N);

B35 asks the distance L between every two adjacent hop heights _i=(P _i+1-P _i), (1<i<N-1), to trip point spacing L _iask histogram, the corresponding L with the histogram maximum _ias symbol lengths rough estimate evaluation

B36 chooses length and is less than l _iadd up total symbol numbers and duration thereof $\tilde{M}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}\mathrm{round}(L_i/{\tilde{T}}_s),(L_i<4{\tilde{T}}_s),$ $\tilde{T}=\underset{i=1}{\overset{N-1}{\mathrm{\&Sigma;}}}{L}_i,(L_i<4{\tilde{T}}_s);$

B37 passes through formula

calculate the estimated value of signal code rate.

Images