CN104052703B - A kind of microsampling data digital Modulation Identification method - Google Patents
A kind of microsampling data digital Modulation Identification method Download PDFInfo
- Publication number
- CN104052703B CN104052703B CN201410317188.2A CN201410317188A CN104052703B CN 104052703 B CN104052703 B CN 104052703B CN 201410317188 A CN201410317188 A CN 201410317188A CN 104052703 B CN104052703 B CN 104052703B
- Authority
- CN
- China
- Prior art keywords
- signal
- parameter
- snr
- instantaneous
- centralization
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000003066 decision tree Methods 0.000 claims abstract description 5
- 238000001914 filtration Methods 0.000 claims abstract description 4
- 230000001052 transient effect Effects 0.000 claims abstract description 4
- 238000004891 communication Methods 0.000 abstract description 5
- 238000002474 experimental method Methods 0.000 description 9
- 238000005070 sampling Methods 0.000 description 7
- 230000000694 effects Effects 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 230000004069 differentiation Effects 0.000 description 3
- 238000012512 characterization method Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 241000196324 Embryophyta Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Landscapes
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
The invention belongs to communication technical field, and in particular to a kind of 2ASK, 4ASK, 2PSK, 4PSK, 2FSK and 4FSK totally 6 kinds of microsampling data digital Modulation Identification methods of modulation system.The present invention includes:Bandpass filtering treatment is carried out to sampled signal;Signal transient information, including instantaneous amplitude, nonlinear phase and instantaneous frequency are obtained using Hilbert transform;Calculate signal amplitude centralization parameter, phase center parameter and center frequency parameter;5 characteristic parameters needed for calculating Modulation recognition;Classification judgement is carried out to signal using decision tree classifier.New centralization parameter proposed by the present invention and classification thresholding automatic adjusting method can efficiently reduce requirement of the recognizer to sample chips quantity, allow that recognition methods is applied on the less hardware platform of resource, and adaptivity of the method to signal to noise ratio can be improved.
Description
Technical field
The invention belongs to communication technical field, and in particular to a kind of 2ASK, 4ASK, 2PSK, 4PSK, 2FSK and 4FSK totally 6
Plant the microsampling data digital Modulation Identification method of modulation system.
Background technology
In information countermeasure field, in order to successfully be intercepted to signal, primary work is exactly to identify the modulation class of signal
Type.With the fast development of radio communication, the modulation system of signal of communication also becomes various all the more, and tradition is based on artificial signal
The need for recognition methods can not meet reality, the modulation classification and tune of automatic identification signal of communication are capable of in the urgent need to making
The equipment of parameter processed.
A.K.Nandi and E.E.Azzouz propose the Modulation Identification side based on signal transient parameter attribute in nineteen ninety-five
Method, the method first does Hibert conversion to signal, and instantaneous amplitude, instantaneous phase and the instantaneous frequency of signal are then calculated successively,
Then 7 identification features being made up of these three parameters are chosen, to 2ASK, 4ASK, 2PSK, 4PSK and 2FSK, 4FSK signals
Recognized.The calculating of this method it is simple, it is necessary to prior information it is less, thus be widely applied.
But the above method is still present some defects in actual hardware implementation process:
1. realized for hardware, due to single-chip microcomputer or DSP resource with respect to be for PC it is very rare and valuable,
The points of the sampled data used by signal identification can not be too many, causes the he number in one group of sampled data less.Work as sampling
Point is less and in the case that code element is random, and the centralization calculation method of parameters of traditional recognition methods occurs relatively large deviation,
Cause the decline of discrimination.
2. because the value of characteristic parameter can change with the change of SNR, particularly sampled point is few, code element is random.Such as
Fruit chooses single classification thresholding, and when SNR changes, recognition performance can drastically deteriorate.
The content of the invention
The feelings that sampled point is less and code element is random are not suitable for it is an object of the invention to be directed to traditional recognition methods
Condition, proposes, using a kind of new centralization parameter and classification thresholding adjust automatically, to be applicable microsampling data and unknown signal to noise ratio
Modulation Signals Recognition method.
The object of the present invention is achieved like this:
(1) bandpass filtering treatment is carried out to sampled signal;
(2) signal transient information, including instantaneous amplitude, nonlinear phase and instantaneous frequency are obtained using Hilbert transform
Rate;
(3) signal amplitude centralization parameter, phase center parameter and center frequency parameter are calculated;
(4) 5 characteristic parameters needed for calculating Modulation recognition:σda、σaa、σdf、σapAnd σaf;Wherein, σdaCentered on change wink
When amplitude standard deviation, σaaCentered on change instantaneous amplitude absolute value standard deviation, σdfIt is non-weak signal section instantaneous frequency
Standard deviation, σapCentered on change instantaneous phase absolute value standard deviation, σafCentered on change instantaneous frequency absolute value standard deviation
Difference;The circular of centralization instantaneous amplitude, centralization instantaneous phase and centralization frequency is:
Centralization instantaneous amplitude aan(i)=a (i)-moa, wherein a (i) is the instantaneous amplitude of signal, moaIt is amplitude centralization
Parameter;
Centralization instantaneous phaseWhereinIt is the nonlinear phase of signal, mopIt is phase center
Change parameter;
Centralization instantaneous frequency fcn(i)=f (i)-mof, wherein f (i) is the instantaneous frequency of signal, mofIt is center frequency
Parameter;
(5) classification judgement is carried out to signal using decision tree classifier:
(5.1) if parameter σda< th1, then signal belong to class 0 { 2FSK, 4FSK, 2PSK, 4PSK }, otherwise signal belongs to
Class 1 { 2ASK, 4ASK };Wherein th1It is parameter σdaDecision threshold;
(5.2) rule out signal and belonged to class 1, perform SNR rough estimates module 1:Work as σap< th5, select th2=th2It is high
SNR, otherwise selects th2=th2Low SNR;If parameter σaa< th2, then signal is 2ASK, and otherwise signal is 4ASK;
Wherein th5It is to utilize σapCarry out the decision threshold of SNR rough estimates, th2It is parameter σaaDecision threshold, th2SNR high
It is the th in the case of high s/n ratio2The thresholding of selection, th2Low SNR is the th in the case of low signal-to-noise ratio2The thresholding of selection;
(5.3) rule out signal and belong to class 0, and parameter σdf< th3, then signal belong to class 01 { 2PSK, 4PSK }, otherwise
Signal belongs to class 00 { 2FSK, 4FSK };Wherein th3It is parameter σdfDecision threshold;
(5.4) rule out signal and belonged to class 01, perform SNR rough estimates module 2:Work as σda< th6, select th5=th5It is high
SNR, otherwise selects th5=th5Low SNR;If parameter σap< th5, then signal belong to 2PSK, otherwise signal is 4PSK;
Wherein th6It is to utilize σdaCarry out the thresholding of SNR rough estimates, th5It is σapDecision threshold, th5SNR high is to believe in height
Make an uproar the th than in the case of5The thresholding of selection, th5Low SNR is the th in the case of low signal-to-noise ratio5The thresholding of selection;
(5) rule out signal and belonged to class 00, perform SNR rough estimates module 3:Work as σda< th6, select th4=th4It is high
SNR, otherwise selects th4=th4Low SNR;If parameter σaf< th4, then signal is 2FSK, and otherwise signal is 4FSK;
Wherein th6It is to utilize σdaCarry out the thresholding of SNR rough estimates, th4It is σafDecision threshold, th4SNR high is to believe in height
Make an uproar the th than in the case of4The thresholding of selection, th4Low SNR is the th in the case of low signal-to-noise ratio4The thresholding of selection.
Amplitude centralization parameter is by moa={ E [a (i) > ma]+E [a (i) < ma]/2 acquisitions, wherein a (i) is signal
Instantaneous amplitude, maIt is the average value of a (i), i.e. ma=E [a (i)], E [] are to take average symbol.
Phase center parameter is by mof={ E [f (i) > mf]+E [f (i) < mf]/2 acquisitions, wherein f (i) is signal
Instantaneous frequency, mfIt is the average value of f (i), i.e. mf=E [f (i)].
Center frequency parameter byObtain, whereinIt is signal
Nonlinear phase, mpForAverage value, i.e.,
Use σapAs characteristic parameter σaaSignal to noise ratio rough estimate parameter, in order to characteristic parameter σaaThresholding adjust automatically
It is whole;Work as σap< th5, select th2=th2SNR high, otherwise selects th2=th2Low SNR.
Use σdaAs characteristic parameter σafSignal to noise ratio rough estimate parameter;Work as σda< th6, select th4=th4SNR high, it is no
Then select th4=th4Low SNR.
Use σdaAs characteristic parameter σapSignal to noise ratio rough estimate parameter, in order to characteristic parameter σapThresholding adjust automatically
It is whole;Work as σda< th6, select th5=th5SNR high, otherwise selects th5=th5Low SNR.
The beneficial effects of the present invention are:New centralization parameter proposed by the present invention and classification thresholding automatic adjusting method
Requirement of the recognizer to sample chips quantity can be efficiently reduced so that recognition methods can be flat in the less hardware of resource
It is applied on platform, and adaptivity of the method to signal to noise ratio can be improved.
Brief description of the drawings
Fig. 1:Traditional recognition methods flow chart.
Fig. 2:Recognition methods flow chart of the invention.
Fig. 3:During using two kinds of centralization methods, parameter σ in 100 experimentsaaDistribution value contrast (16 code elements,
15dB)。
Fig. 4:During using two kinds of centralization methods, parameter σ in 100 experimentsafDistribution value contrast (16 code elements,
15dB)。
Fig. 5:During using two kinds of centralization methods, parameter σ in 100 experimentsapDistribution value contrast (16 code elements,
15dB)。
Specific embodiment
The present invention is described further below in conjunction with the accompanying drawings, in wherein Fig. 3-5, (a) is partly centralization of the invention
Method, (b) is partly traditional centralization method
The present invention includes:
Step 1:The sample sequence of the modulated signal for receiving is x (n), and bandpass filtering treatment is carried out to x (n).
Step 2:Instantaneous amplitude, nonlinear phase and the instantaneous frequency of signal are extracted using Hilbert transform.
Step 3:Calculate signal frequency centralization parameter and amplitude centralization parameter;
Amplitude centralization parameter is by formula
moa={ E [a (i) > ma]+E [a (i) < ma]/2 acquisitions, wherein a (i) is the instantaneous amplitude of signal, maIt is a (i)
Average value, i.e. ma=E [a (i)] (E [] is to take average symbol).
Center frequency parameter is by formula
mof={ E [f (i) > mf]+E [f (i) < mf]/2 acquisitions, wherein f (i) is the instantaneous frequency of signal, mfIt is f (i)
Average value, i.e. mf=E [f (i)].
Phase center parameter is by formula
Obtain, whereinIt is the nonlinear phase of signal, mpForAverage value, i.e.,
Step 4:5 characteristic parameters needed for calculating Modulation recognition:σda、σaa、σdf、σapAnd σaf.Wherein, σdaCentered on change
The standard deviation of instantaneous amplitude, σaaCentered on change instantaneous amplitude absolute value standard deviation, σdfIt is non-weak signal section instantaneous frequency
Standard deviation, σapCentered on change instantaneous phase absolute value standard deviation, σafCentered on change instantaneous frequency absolute value standard
Deviation;The circular of centralization instantaneous amplitude, centralization instantaneous phase and centralization frequency is:
Centralization instantaneous amplitude aan(i)=a (i)-moa, wherein a (i) is the instantaneous amplitude of signal, moaIt is amplitude centralization
Parameter;
Centralization instantaneous phaseWhereinIt is the nonlinear phase of signal, mopIt is phase center
Change parameter;
Centralization instantaneous frequency fcn(i)=f (i)-mof, wherein f (i) is the instantaneous frequency of signal, mofIt is center frequency
Parameter;
Step 5:Classification process using decision tree classifier is classified to signal.
Described signal carries out classification specific method:
(1) if parameter σda< th1, then signal belong to class 0 { 2FSK, 4FSK, 2PSK, 4PSK };Otherwise signal belongs to class 1
{2ASK,4ASK};Wherein th1It is σdaDecision threshold.
(2) rule out signal and belonged to class 1, perform SNR rough estimates module 1:Work as σap< th5, select th2=th2It is (high
SNR), th is otherwise selected2=th2(low SNR).If parameter σaa< th2, then signal is 2ASK;Otherwise signal is 4ASK.
Wherein th5It is to utilize σapCarry out the decision threshold (SNR is divided to for high and low two kinds of situations) of SNR rough estimates, th2For
σaaDecision threshold, th2(SNR high) is th2The thresholding chosen under high s/n ratio, th2(low SNR) is th2Under low signal-to-noise ratio
Selection thresholding.
(3) rule out signal and belong to class 0, and parameter σdf< th3, then signal belong to class 01 { 2PSK, 4PSK };Otherwise believe
Number belong to class 00 { 2FSK, 4FSK }.Wherein th3It is σdfDecision threshold.
(4) rule out signal and belonged to class 01, perform SNR rough estimates module 2:Work as σda< th6, select th5=th5It is (high
SNR), th is otherwise selected5=th5(low SNR).If parameter σap< th5, then signal belong to 2PSK;Otherwise signal is 4PSK.
Wherein th6It is to utilize σdaCarry out the decision threshold of SNR rough estimates, th5It is σapDecision threshold, th5(SNR high) is
The th under high s/n ratio5The thresholding of selection, th5(low SNR) is the th under low signal-to-noise ratio5The thresholding of selection.
(5) rule out signal and belonged to class 00, perform SNR rough estimates module 3:Work as σda< th6, select th4=th4It is (high
SNR), th is otherwise selected4=th4(low SNR).If parameter σaf< th4, then signal is 2FSK;Otherwise signal is 4FSK.
Wherein th6It is to utilize σdaCarry out the decision threshold of SNR rough estimates, th4It is σafDecision threshold, th4(SNR high) is
The th under high s/n ratio4The thresholding of selection, th4(low SNR) is the th under low signal-to-noise ratio4The thresholding of selection.
The centralization parameter proposed in the traditional recognition method only simply average of the number of winning the confidence, i.e. E [h (i)], h (i)
It is the prompting message (instantaneous amplitude, nonlinear phase or instantaneous frequency) of signal.Centralization parameter proposed by the present invention is:
moh={ E [h (i) > mh]+E [h (i) < mh]/2, mh=E [h (i)] (1)
When sampled data is less, the number of symbols of sampling is less and during lack of balance (when the quantity of code element 1 and 0 is different)
The centralization calculation method of parameters of traditional recognition method occurs larger deviation.The σ of 2ASK is pointed out in traditional recognition methodsaa
Parameter should be equal to the σ of 0,4ASKaaParameter is not 0.Reason is that the normalization range value of 2ASK only has 0 and 1, subtracts centralization ginseng
After numerical value 0.5, range value is -0.5 and 0.5.Take absolute value again, only 0.5 1 range values, it is 0 to take variance.But actually
It is not necessarily such, the selection of centralization parameter value is it is critical only that, if not 0.5, then centralization amplitude cannot be ensured
Absolute value only one of which value.
Assuming that Baud Length is 8, there was only 2 sampled points to calculate simply clearly assume between each code element, one has 16
Individual sampled point, the amplitude centralization parametric results that two methods obtain 2ASK are as shown in table 1.
The comparing of the conventional method of table 1 and centralization parameter value of the invention
By the result of calculation of table 1 can be seen that when in code element 1 as 0 number when obtained using previous literature algorithm
Heart parameter value is just equal to 0.5, will not otherwise be equal to 0.5.When 0 and 1 number difference is larger, such as in testing
10000000, it is 0.125 that previous literature calculates centralization value, is differed greatly with 0.5.And set forth herein centralization parameter
Value, no matter whether code element balances, and can obtain centralization value 0.5, is consistent with actual demand all the time.And in the condition of random code element
Under, code element is fewer to be more difficult to meet code element condition in a balanced way, and the selection thus for less sampled data centralization parameter value shows
Obtain particularly important.It is similar with the analysis of phase center parameter for frequency, it is not repeated.
To the σ of coherent signal under the conditions of 15dB signal to noise ratios, random code elementaa、σafAnd σapCarry out 100 experiments, value
As shown in Figure 3-Figure 5.From the figure 3, it may be seen that during using conventional center algorithm, σaaParameter is to two kinds of differentiations of modulation of 2ASK and 4ASK
, there is more aliasing situation in DeGrain, it is difficult to choose a suitable decision threshold.And use centralization of the present invention to calculate
During method, σaaParameter is more satisfactory to the effect that two kinds of modulation are distinguished, and aliasing situation is less, can easily set classification door
Limit.For parameter σafAnd σapAnalysis be similar to, be not repeated.
The apparent σ for working as 2ASKaaThe σ of mean parameter and 4ASKaaMean parameter difference is bigger, easier setting thresholding area
Separate two signals.While the σ of two signalsaaThe smaller selection for also getting over the differentiation and thresholding that are conducive to signal of parameter fluctuation, i.e. σaaGinseng
Several standards is smaller more easily to distinguish two signals.For this can define characterization parameter to signal distinguishing performance quality parameter Kaa、
KapAnd Kaf:
In formula,It is the σ of 2ASK signals in 100 experimentsaaValue result,It is the σ of 4ASK signals in 100 experimentsaaValue result,It is the σ of 2FSK signals in 100 experimentsafValue result, other parameters be similar to.std[·]
It is the standard deviation of parameter, E [] is the average of parameter.
KaaWithStandard deviation size be directly proportional,Standard deviation size is directly proportional, withAverage
WithThe order of magnitude of equal value difference be inversely proportional.KaaThe smaller characterization parameter σ of valueaaThe effect distinguished to { 2ASK, 4ASK }
Fruit is better;KaaValue it is bigger, to { 2ASK, 4ASK } distinguish effect it is poorer.For parameter Kaf,KapAnalysis is similarly.
By experiment it can be found that working as Ka□<Can (wherein takes a, p or f) preferably to { 2 SK, 4 SK } when 5
Distinguish.As shown in Table 2, the data for only needing to 16 code-element periods of sampling of the invention can reach Ka□<5 condition, and it is traditional
Method needs 112 code-element periods of sampling to can be only achieved same recognition effect.
Influence (signal to noise ratio 15dB) of the he number of table 2 to sorting parameter performance
As shown in Table 3, in the data of 16 code-element periods of fixed sample, the present invention is when signal to noise ratio is slightly larger than 12dB
Can reach Ka□<5 condition.And traditional method is in the data of 16 code-element periods of sampling, whatsoever signal to noise ratio all without
Method reaches same differentiation effect.
Influence (16 code element) of the signal to noise ratio of table 3 to sorting parameter performance
Find in an experiment, it is reliable for signal distinguishing between major class when SNR is not less than 10dB, such as adjust
Identification between width, phase modulation and FM signal.But for being distinguished between (2ASK, 4ASK), distinguished between (2PSK, 4PSK) or
The degree of accuracy distinguished between person (2FSK, 4FSK) is very low.Main cause be for distinguish (2ASK, 4ASK), (2PSK, 4PSK) or
The parameter σ of person (2FSK, 4FSK)aa、σapOr σafChange more violent with SNR when signal to noise ratio is smaller.Single thresholding is (such as
Chosen during 20dB), it is impossible to ensure correctly to be recognized in whole SNR >=10dB situations.Thus need to estimate letter first before judgement
Make an uproar and compare, and select suitable thresholding.The method complex structure of traditional signal-to-noise ratio (SNR) estimation, operand is big, is not suitable for quick knowledge
Other system.The method that the present invention is used is, to parameter σaa、σapOr σafSignal to noise ratio reference parameter is set so that decision threshold can be with
The adjust automatically with the change of SNR reference parameters.
Selecting the principle of signal to noise ratio reference parameter is:For (2FSK, 4FSK), (2PSK, 4PSK) or (2ASK, 4ASK)
The difference of its relevant parameter value under the same conditions should be smaller, and the parameter should be able to reflect signal to noise ratio size.The present invention
Selection σdaAs the reference parameter of (2FSK, 4FSK) and (2PSK, 4PSK) signal-to-noise ratio (SNR) estimation, σapAs (2ASK, 4ASK) noise
The reference parameter of compared estimate.(2FSK, 4FSK), (2PSK, 4PSK) are constant envelope signal, its σ under noise-free casedaIt is 0, when depositing
In noise, σdaWhat is reflected is in fact the mean square amplitude size of noise.(2ASK, 4ASK) is permanent in non-weak section of nonlinear phase
It is fixed, its σ under noise-free caseapIt is 0, when there is noise, σapWhat is reflected is also the mean square amplitude size of noise.
Embodiment
Implementation method is comprised the following steps:
Step 1:The sample sequence of the modulated signal for receiving is x (n), and treatment is filtered to x (n).
Step 2, instantaneous amplitude, nonlinear phase and instantaneous frequency that signal is extracted using Hilbert transform.
Modulated signal u (t) for receiving can be expressed as:
Analytic signal form z (t) of input signal:
Z (t)=u (t)+jv (t) (4)
Wherein, u (t) is former modulated signal, and v (t) is the Hilbert transform of u (t).I.e.:
Then there is instantaneous amplitude:
Instantaneous phase:
The instantaneous frequency of signal can be tried to achieve by the differential to instantaneous phase, after obtaining instantaneous phase, in addition it is also necessary to it
Phase being converted by folding is carried out, to obtain the nonlinear phase of signal.
Can be represented by formula (8) by instantaneous phase θ (i) after A/D sampling digitizings, wherein fsIt is sample frequency,
It is nonlinear phase.
What formula (7) was calculated phase is the value of the π of actual phase mould 2, in order to be obtained from the θ (i) for having phase to foldWrapped phase first must be reverted to original unfolded phase (i).Linear phase composition is subtracted from φ (i) again,
Nonlinear phase component can be obtained.Therefore, calculating orrection phase place sequence C (i) first:
Then unfolded phase (i) is:
φ (i)=θ (i)+C (i) (10)
The composition of linear phase is subtracted from φ (i) again, you can obtain nonlinear phase:
Step 3:Calculate signal amplitude centralization parameter, center frequency parameter and phase center parameter;
Amplitude centralization parameter is by formula
moa={ E [a (i) > ma]+E [a (i) < ma]/2 acquisitions, wherein ma=E [a (i)], a (i) are the instantaneous width of signal
Degree.
Center frequency parameter is by formula
mof={ E [f (i) > mf]+E [f (i) < mf]/2 acquisitions, wherein mf=E [f (i)], f (i) are the instantaneous frequency of signal
Rate.
Phase center parameter is by formula
Obtain, wherein It is the non-thread of signal
Property phase.
5 characteristic parameters needed for step 4, calculating Modulation recognition:
(1) standard deviation of centralization instantaneous amplitudeda
In formula, NsIt is number of sampling, acnChange instantaneous amplitude centered on (i):
acn(i)=a (i)-moa (13)
Wherein:moa={ E [a (i) > ma]+E [a (i) < ma]}/2,ma=E [a (i)], a (i) are instantaneous amplitude.σdaWith
It is that constant envelope signal is also non-permanent envelope to distinguish, { 2PSK, 4PSK, 2FSK, 4FSK } is constant envelope signal, its σdaIn preferable feelings
Should be 0 under condition;{ 2ASK, 4ASK } signal is non-constant envelope signal, its σdaMore than 0;
(2) standard deviation of non-weak signal section centralization instantaneous frequencydf:
In formula:atIt is an amplitude decision threshold level for judging weak signal, c is in gross sample data NsIn belong to non-
The number of weak signal value, foChange instantaneous frequency centered on (i):
fo(i)=f (i)-mof (15)
Wherein, mof={ E [f (i) > mf]+E [f (i) < mf]}/2,mf=E [f (i)], f (i) are the instantaneous frequency of signal.
σdfIt is FM signal { 2FSK, 4FSK } or other signals { 2PSK, 4PSK } for distinguishing signal.
(3) standard deviation of centralization instantaneous amplitude absolute valueaa
σaaTo be used for distinguishing signal is 2ASK or 4ASK.
(4) standard deviation of centralization instantaneous phase absolute valueap
φNLChange instantaneous phase centered on (i):
It is the nonlinear phase of signal.
σaaIt is 2PSK or 4PSK to be mainly used to distinguish signal.
(5) standard deviation of centralization instantaneous frequency absolute valueaf:
σafIt is 2FSK or 4FSK for distinguishing signal.
Step 5, signal is classified using decision tree classifier.
Described signal carries out classification specific method:
(1) if parameter σda< th1, then signal belong to class 0 { 2FSK, 4FSK, 2PSK, 4PSK };Otherwise signal belongs to class 1
{2ASK,4ASK};
(2) rule out signal and belonged to class 1, perform SNR rough estimates module 1:Work as σap< th5, select th2=th2It is (high
SNR), th is otherwise selected2=th2(low SNR);If parameter σaa< th2, then signal is 2ASK;Otherwise signal is 4ASK;
(3) rule out signal and belong to class 0, and parameter σdf< th3, then signal belong to class 01 { 2PSK, 4PSK };Otherwise believe
Number belong to class 00 { 2FSK, 4FSK };
(4) rule out signal and belonged to class 01, perform SNR rough estimates module 2:Work as σda< th6, select th5=th5It is (high
SNR), th is otherwise selected5=th5(low SNR);If parameter σap< th5, then signal belong to 2PSK;Otherwise signal is 4PSK;
(5) rule out signal and belonged to class 00, perform SNR rough estimates module 3:Work as σda< th6, select th4=th4It is (high
SNR), th is otherwise selected4=th4(low SNR);If parameter σaf< th4, then signal is 2FSK;Otherwise signal is 4FSK.
2ASK, 4ASK, 2FSK, 4FSK, 2PSK and 4PSK totally 6 kinds of modulated signals are considered, to identification in Matlab7 environment
Method is emulated.The carrier frequency f of input modulating signalcIt is 5MHz, sample frequency fsIt is 32MHz.The base band code of digital modulation
Unit uses random code element, chip rate fdIt is 1M character per seconds.
Under 10,15 and 20dB signal to noise ratios, 6 kinds of modulation types are carried out with 300 identification emulation respectively, statistic algorithm is just
True discrimination, experimental result is as shown in table 4.As shown in Table 4, the present invention sampled point 256, in signal to noise ratio 10dB when, averagely
Recognition correct rate reaches 82.66%, and in SNR >=15dB, average recognition rate is more than 91.21%;In sampled point 512, in letter
Make an uproar during than 10dB, Mean accurate rate of recognition reaches 92.00%, reached the requirement of Project Realization, and realize decision threshold
The function of adjust automatically.
When the sampled point of table 4 is 256/512/1024, the accuracy rate % of recognition methods
Claims (7)
1. a kind of microsampling data digital Modulation Identification method, it is characterised in that:
(1) bandpass filtering treatment is carried out to sampled signal;
(2) signal transient information, including instantaneous amplitude, nonlinear phase and instantaneous frequency are obtained using Hilbert transform;
(3) signal amplitude centralization parameter, phase center parameter and center frequency parameter are calculated;
(4) 5 characteristic parameters needed for calculating Modulation recognition:σda、σaa、σdf、σapAnd σaf;Wherein, σdaCentered on change instantaneous width
The standard deviation of degree, σaaCentered on change instantaneous amplitude absolute value standard deviation, σdfIt is the standard of non-weak signal section instantaneous frequency
Deviation, σapCentered on change instantaneous phase absolute value standard deviation, σafCentered on change instantaneous frequency absolute value standard deviation;
The circular of centralization instantaneous amplitude, centralization instantaneous phase and centralization frequency is:
Centralization instantaneous amplitude aan(i)=a (i)-moa, wherein a (i) is the instantaneous amplitude of signal, moaFor amplitude centralization is joined
Number;
Centralization instantaneous phaseWhereinIt is the nonlinear phase of signal, mopFor phase centerization is joined
Number;
Centralization instantaneous frequency fcn(i)=f (i)-mof, wherein f (i) is the instantaneous frequency of signal, mofFor center frequencyization is joined
Number;
(5) classification judgement is carried out to signal using decision tree classifier:
(5.1) if parameter σda< th1, then signal belong to class 0 { 2FSK, 4FSK, 2PSK, 4PSK }, otherwise signal belongs to class 1
{2ASK,4ASK};Wherein th1It is parameter σdaDecision threshold;
(5.2) rule out signal and belonged to class 1, perform SNR rough estimates module 1:Work as σap< th5, select th2=th2SNR high, it is no
Then select th2=th2Low SNR;If parameter σaa< th2, then signal is 2ASK, and otherwise signal is 4ASK;
Wherein th5It is to utilize σapCarry out the decision threshold of SNR rough estimates, th2It is parameter σaaDecision threshold, th2SNR high be
Th in the case of high s/n ratio2The thresholding of selection, th2Low SNR is the th in the case of low signal-to-noise ratio2The thresholding of selection;
(5.3) rule out signal and belong to class 0, and parameter σdf< th3, then signal belong to class 01 { 2PSK, 4PSK }, otherwise signal
Belong to class 00 { 2FSK, 4FSK };Wherein th3It is parameter σdfDecision threshold;
(5.4) rule out signal and belonged to class 01, perform SNR rough estimates module 2:Work as σda< th6, select th5=th5SNR high,
Otherwise select th5=th5Low SNR;If parameter σap< th5, then signal belong to 2PSK, otherwise signal is 4PSK;
Wherein th6It is to utilize σdaCarry out the thresholding of SNR rough estimates, th5It is σapDecision threshold, th5SNR high is in high s/n ratio
In the case of th5The thresholding of selection, th5Low SNR is the th in the case of low signal-to-noise ratio5The thresholding of selection;
(5) rule out signal and belonged to class 00, perform SNR rough estimates module 3:Work as σda< th6, select th4=th4SNR high, it is no
Then select th4=th4Low SNR;If parameter σaf< th4, then signal is 2FSK, and otherwise signal is 4FSK;
Wherein th6It is to utilize σdaCarry out the thresholding of SNR rough estimates, th4It is σafDecision threshold, th4SNR high is in high s/n ratio
In the case of th4The thresholding of selection, th4Low SNR is the th in the case of low signal-to-noise ratio4The thresholding of selection.
2. a kind of microsampling data digital Modulation Identification method according to claim 1, it is characterized in that:In the amplitude
Heart parameter is by moa={ E [a (i) > ma]+E [a (i) < ma]/2 acquisitions, wherein a (i) is the instantaneous amplitude of signal, maIt is a
The average value of (i), i.e. ma=E [a (i)], E [] are to take average symbol.
3. a kind of microsampling data digital Modulation Identification method according to claim 1, it is characterized in that:In the phase
Heart parameter is by mof={ E [f (i) > mf]+E [f (i) < mf]/2 acquisitions, wherein f (i) is the instantaneous frequency of signal, mfIt is f
The average value of (i), i.e. mf=E [f (i)].
4. a kind of microsampling data digital Modulation Identification method according to claim 1, it is characterized in that:In the frequency
Heart parameter byObtain, whereinIt is the nonlinear phase of signal, mp
ForAverage value, i.e.,
5. a kind of microsampling data digital Modulation Identification method according to claim 1, it is characterized in that:Use σapAs
Characteristic parameter σaaSignal to noise ratio rough estimate parameter, in order to characteristic parameter σaaThresholding adjust automatically;Work as σap< th5, select th2
=th2SNR high, otherwise selects th2=th2Low SNR.
6. a kind of microsampling data digital Modulation Identification method according to claim 1, it is characterized in that:Use σdaAs
Characteristic parameter σafSignal to noise ratio rough estimate parameter;Work as σda< th6, select th4=th4SNR high, otherwise selects th4=th4It is low
SNR。
7. a kind of microsampling data digital Modulation Identification method according to claim 1, it is characterized in that:Use σdaAs
Characteristic parameter σapSignal to noise ratio rough estimate parameter, in order to characteristic parameter σapThresholding adjust automatically;Work as σda< th6, select th5
=th5SNR high, otherwise selects th5=th5Low SNR.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410317188.2A CN104052703B (en) | 2014-07-04 | 2014-07-04 | A kind of microsampling data digital Modulation Identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410317188.2A CN104052703B (en) | 2014-07-04 | 2014-07-04 | A kind of microsampling data digital Modulation Identification method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104052703A CN104052703A (en) | 2014-09-17 |
CN104052703B true CN104052703B (en) | 2017-06-20 |
Family
ID=51505079
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410317188.2A Expired - Fee Related CN104052703B (en) | 2014-07-04 | 2014-07-04 | A kind of microsampling data digital Modulation Identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104052703B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104363194B (en) * | 2014-11-04 | 2016-05-11 | 武汉大学 | PSK Modulation Identification method based on waveform transformation |
CN105207965B (en) * | 2015-08-14 | 2018-07-24 | 成都中安频谱科技有限公司 | A kind of Automatic modulation classification method of VHF/UHF frequency ranges |
CN108270703A (en) * | 2016-12-30 | 2018-07-10 | 中国航天科工集团八五研究所 | A kind of signal of communication digital modulation type recognition methods |
CN110035025A (en) * | 2019-04-22 | 2019-07-19 | 桂林电子科技大学 | A kind of detection recognition method of the multicarrier mixed signal based on direct feature extraction |
CN114374436A (en) * | 2022-01-11 | 2022-04-19 | 北京鼎普科技股份有限公司 | Visible light signal detection device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0544991A2 (en) * | 1991-11-29 | 1993-06-09 | Daimler-Benz Aerospace Aktiengesellschaft | Method for the automatic classification of digitally modulated signals, and apparatus to carry out the method |
CN101834819A (en) * | 2010-05-14 | 2010-09-15 | 哈尔滨工业大学 | Analog-digital mixing modulation recognition device and digital modulation recognition device based on parallel judgment |
CN102710572A (en) * | 2012-07-06 | 2012-10-03 | 江苏省邮电规划设计院有限责任公司 | Feature extraction and modulation identification method of communication signals |
-
2014
- 2014-07-04 CN CN201410317188.2A patent/CN104052703B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0544991A2 (en) * | 1991-11-29 | 1993-06-09 | Daimler-Benz Aerospace Aktiengesellschaft | Method for the automatic classification of digitally modulated signals, and apparatus to carry out the method |
CN101834819A (en) * | 2010-05-14 | 2010-09-15 | 哈尔滨工业大学 | Analog-digital mixing modulation recognition device and digital modulation recognition device based on parallel judgment |
CN102710572A (en) * | 2012-07-06 | 2012-10-03 | 江苏省邮电规划设计院有限责任公司 | Feature extraction and modulation identification method of communication signals |
Non-Patent Citations (2)
Title |
---|
Automatic identification of digital modulation types;E.E.Azzouz,A.k.Nandi;《SIGNAL PROCESSING》;19941128;全文 * |
数字调制信号识别***的DSP硬件实现;何林飞,张晓林;《应用科技》;20130831;第40卷(第4期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN104052703A (en) | 2014-09-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104052703B (en) | A kind of microsampling data digital Modulation Identification method | |
CN107124381B (en) | Automatic identification method for digital communication signal modulation mode | |
CN104052702B (en) | The recognition methods of digital modulation signals under a kind of Complex Noise | |
CN107948107B (en) | Digital modulation signal classification method based on joint features | |
CN101834819B (en) | Analog-digital mixing modulation recognition device and digital modulation recognition device based on parallel judgment | |
CN101764786B (en) | MQAM signal recognition method based on clustering algorithm | |
CN108134753A (en) | The recognition methods of broadband signal modulation system | |
CN104301056A (en) | Frequency spectrum monitoring method based on signal feature analysis | |
CN102882819A (en) | Digital demodulation signal identification method under non-gaussian noise | |
CN103199945B (en) | Method for identifying modulation mode of cognitive radio signal under low signal-to-noise ratio condition | |
CN106330805B (en) | A kind of signal modulation mode automatic identifying method and system | |
CN104869096B (en) | Bootstrap-based BPSK signal blind processing result credibility test method | |
CN113225282A (en) | Communication signal modulation identification method based on BP neural network | |
CN106169070A (en) | The communication specific emitter identification method and system represented based on cooperation | |
CN106357575A (en) | Multi-parameter jointly-estimated interference type identification method | |
CN112003803B (en) | Detection and reception equipment for VHF and UHF band aviation radio station signals | |
CN109076038B (en) | Method for estimating parameters of a signal contained in a frequency band | |
CN103997475B (en) | A kind of method of digital modulation signals under identification Alpha Stable distritation noises | |
CN108650203B (en) | Modulation mode identification method based on reconnaissance receiver | |
CN101764785B (en) | Quadrature amplitude modulation signal identifying method based on mixed moment and fisher discrimination | |
CN108768563A (en) | A kind of cooperative frequency spectrum sensing method and relevant apparatus | |
CN102006252A (en) | Single-tone signal identification method | |
CN116506273A (en) | Novel MPSK modulation signal identification and classification method | |
CN111540381A (en) | Voice simulation modulation characteristic recognition method based on random forest | |
CN111245756B (en) | Composite signal modulation recognition method based on cascade SVM and full digital receiver |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170620 |