CN107135176A

CN107135176A - Figure field communication signal modulate method based on fractional lower-order Cyclic Spectrum

Info

Publication number: CN107135176A
Application number: CN201710546645.9A
Authority: CN
Inventors: 阎啸; 刘冠男; 吴孝纯; 王茜
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2017-07-06
Filing date: 2017-07-06
Publication date: 2017-09-05
Anticipated expiration: 2037-07-06
Also published as: CN107135176B

Abstract

The invention discloses a kind of figure field communication signal modulate method based on fractional lower-order Cyclic Spectrum.Utilize the three-dimensional fractional lower-order Cyclic Spectrum for receiving signal, it will be transformed into by the modulated signal of α Stable distritation noise jammings on figure domain, then effective characteristic parameters line index arrangement set is extracted in the sparse adjacency matrix that can be represented from figure as the feature of modulation type, according to training signal and the line index arrangement set Hamming distance for receiving signal, to realize under α Stable distritation noise jammings, the identification of more stable more effective signal of communication modulation type.

Description

Image domain communication signal modulation identification method based on fractional low-order cyclic spectrum

Technical Field

The invention belongs to the technical field of signal processing, and particularly relates to an image domain communication signal modulation identification method based on a fractional low-order cyclic spectrum.

Background

Automatic Modulation Classification (AMC), also called communication signal Modulation identification, can identify the Modulation type of a received signal under the condition of little or no prior knowledge, is an essential important step between signal detection and demodulation, and is widely applied to many military and civil communication fields.

Classical Automatic Modulation Classification (AMC) methods can be generally divided into two categories: (i) a likelihood-based (LB) decision theory approach and (ii) a feature-based (FB) Pattern Recognition (PR) approach. However, the LB approach inevitably has some disadvantages, such as lack of closed form solution, intolerable high computational complexity, and mismatch of probabilistic models. The performance of FB methods is not optimal, however they can be implemented very efficiently, and therefore many studies utilize different features and different classification algorithms in pursuit of robust performance of FB methods.

It should be noted that both the LB method and the FB method are applied to the assumption of gaussian noise channel, however, various studies have shown that in practical wireless communication, frequency channels are usually multiple access interference caused by significant pulses, low frequency atmospheric noise, electromagnetic interference, etc. These physical noises exhibit sharp pulse characteristics and probability density distributions with heavy tails. These non-gaussian distributed noises, which are the main sources of errors in wireless communication systems, can be modeled as alpha stationary distributed noises according to the central limit theorem. In a channel where α -stable distributed noise occurs, the conventional AMC method may have significant deterioration in performance.

Automatic Modulation Classification (AMC) based on graph domain_G) AMC techniques were introduced to the graphics domain for the first time and have achieved superior performance over existing PR and LB-based decision theory algorithms, but the method extracts the graphics domain features by mapping the graphics domain to the second order cyclic spectrum of the received signal, however, second and higher order statistics are not present in α stationary distributed noise, and therefore, existing AMC techniques do not provide for such statistics_GThe method also fails in α stable distributed noise, and therefore, new, more stable and effective AMC techniques suitable for α stable distributed noise are urgently to be discovered.

Disclosure of Invention

The invention aims to overcome the defects of the prior art and provides an image domain communication signal modulation identification method based on a fractional low-order cyclic spectrum so as to adapt to alpha stable distribution noise and realize more stable and effective identification of a communication signal modulation type.

In order to achieve the above object, the method for modulating and identifying an image domain communication signal based on a fractional low-order cyclic spectrum according to the present invention is characterized by comprising the following steps:

(1) feature extraction of modulation type training signal

1.1) map Domain mapping based on fractional Sum-valent Cyclic Spectrum

Training signal x for noiseless class k modulation type^k(t), K is 1,2, …, K is the type number of modulation types; dividing a sampling sequence into L sections, and mapping each section by an image domain:

using FAM algorithm ((Fast Fourier transform) Accumulation Method): FFT accumulation algorithm for calculating the cyclic spectrum density) to calculate the flo cs (Fractional Low-Order cyclic spectrum) of the l training signals, and obtain a domain set:

wherein,h-1, 2.. H, which represents the cyclic frequency at which the segment l of the training signal of the type k modulation remains_hCorresponding time domain smooth cycle periodic diagram is extracted to obtain H cycle frequencies_hAnd smoothing the adjacent matrix of the cyclic period diagram by the corresponding time domain to obtain an adjacent matrix set:

wherein, the time domain smoothing cycle periodic diagram is obtained according to the following mode:

a1) normalizing and quantizing the calculated FLOCS (fractional low-order cyclic spectrum) to obtain a discrete fractional low-order cyclic spectrum with the maximum value of 1Wherein, is the cycle frequency, f is the frequency;

in the FAM algorithm, the frequency resolution of the FLOCS is Δ f ═ f_sN', cyclic frequency resolution Δ α ═ 1/Δ t ═ f_sN, wherein f_sFor sampling frequency, N 'is the number of data used for complex demodulation, N is the number of data points input within Δ t time, and thus the FLOCS calculated by the FAM algorithm is a matrix of (N' +1) × (2N + 1);

a2) discrete-based fractional low-order cycles due to FLOCS symmetryThe quarter quadrant establishes the corresponding map domain mapping:

defining a stable cycle frequency_p,p＝1,2...N，_pThe conditions are satisfied:

will stabilize the cycle frequency_pAnd taking the frequency value corresponding to the N as a vertex to obtain a vertex set:

taking the amplitude difference value between the two vertexes as an edge, obtaining an edge set:

wherein:

thus, at each stable cycle frequency_pThen, a corresponding map domain map is obtained, i.e. the domain smoothing cycle chart is:

deleting the circulation frequency with the fraction low-order circulation spectrum of 0 to obtain H reserved circulation frequencies_hThe corresponding time domain smoothing cycle chart is as follows:

1.2) extraction of the line index sequence

For each adjacency matrixL1, 2, …, extracting non-zero entries (elements) of the secondary diagonal line directly above the primary diagonal line, and extracting the row index sequences corresponding to the non-zero entries (elements)The principle of row index sequence extraction is as follows:

b1) checking the nonzero values of the secondary diagonal lines, listing the row indexes corresponding to the nonzero values, performing descending arrangement on the row indexes according to the absolute values of the nonzero values, and then sequentially extracting the row indexes according to the descending order;

b2) if two or more non-zero entries have the same absolute value, extracting the row index closest to the previously extracted row index, and discarding the others;

b3) if two or more non-zero entries have the same absolute value and are the largest, then the largest row index is selected, and the others are discarded;

thus obtaining the cycle frequency_hCorrespondingly obtaining L row index sequences, and selecting the row indexes with the occurrence probability of more than 95 percent in the L row index sequences to form a stable row index sequence

For a training signal of a k-th modulation type, H circulation frequencies are extracted_hA stable row index sequence, forming a stable row index sequence set:and as a characteristic of the kth class modulation type;

(2) identification of modulation type of communication signal

For a received signal, acquiring the characteristics of the modulation type of the received signal according to the method in the step (1), and collecting the row index sequenceWherein V is the number of reserved cycle frequencies;

computing a set of row index sequencesCharacteristic of modulation type of class kThe Hamming distances of the K Hamming distances are obtainedK is 1,2, …, K, and then the minimum hamming distance is found, and the corresponding modulation type is the modulation type of the received communication signal.

The object of the invention is thus achieved.

In order to deal with alpha stable distributed noise, the invention discloses an image domain communication signal modulation identification method based on fractional low-order cyclic spectrum. The modulation signal interfered by alpha stable distribution noise is converted to a graph domain by utilizing a three-dimensional fractional low-order cyclic spectrum of a received signal, then an effective characteristic parameter row index sequence set can be extracted from a sparse adjacent matrix shown in the graph as the characteristic of a modulation type, and more stable and effective identification of the modulation type of the communication signal under the interference of the alpha stable distribution noise is realized according to the Hamming distance of the training signal and the row index sequence set of the received signal.

Drawings

FIG. 1 is a schematic block diagram of an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention is provided in order to better understand the present invention for those skilled in the art with reference to the accompanying drawings. It is to be expressly noted that in the following description, a detailed description of known functions and designs will be omitted when it may obscure the subject matter of the present invention.

For convenience of description, the related terms appearing in the detailed description are explained:

amc (automatic modulation classification): automatic modulation classification;

FB (feature-based): based on statistical features

Pr (pattern recognition): pattern recognition

LB (Likelihood-based influence): based on likelihood functions

AMC_G(graph-based automatic modulation classification): automatic modulation classification of the image domain;

pdf (robustness diversity function): a probability density function;

cf (charateristic function): a feature function;

FLOCS (fractional low-order cyclic spectrum): fractional low order cyclic spectrum;

FLOC (fractional low-order correlation): fractional low order autocorrelation functions;

FLOCC (fractional low-order cyclic correlation): fractional low order cyclic autocorrelation;

FAM (FFT (fast Fourier transform) accumulation method): an FFT accumulation algorithm for calculating the cyclic spectral density;

BPSK (binary phase-shift keying): binary phase shift keying;

QPSK (quaternary phase-shift keying): quadrature phase shift keying;

oQPSK (offset quadrature phase-shift keying): offset quadrature phase shift keying;

2FSK (binary frequency-shift keying): binary frequency shift keying;

4FSK (square frequency-shift keying): quaternary frequency shift keying;

msk (minimum shift keying): minimum frequency shift keying;

1. alpha stable distribution

The alpha stable distribution is also called non-gaussian stable distribution and heavy tail distribution, and is a generalized gaussian distribution, and the distribution model can accurately simulate the statistical characteristics of noise in an actual wireless communication environment.

The model of the alpha stable distribution is the only model satisfying the stability and generalized central limit theorem, the alpha stable distribution does not have a uniform and closed Probability Density Function (PDF), but has a uniform Characteristic Function (CF), which can be expressed as:

ψ(u)＝exp{jau-γ|u|^α[1+jβsgn(u)ω(u,α)]} (9)；

wherein sgn (·) is a sign function. Alpha (alpha is more than 0 and less than or equal to 2) is a characteristic index which determines the pulse characteristic degree of the distribution, and the smaller the alpha value is, the thicker the tail of the corresponding distribution is, so the more remarkable the pulse characteristic is; beta (-1 is more than or equal to beta is less than or equal to 1) is a deviation parameter and is used for determining the symmetrical degree of distribution; γ (γ > 0) is the dispersion coefficient, also known as the scale parameter, which is a measure of how dispersed a sample deviates from its mean, similar to the variance in a gaussian distribution; α (- ∞ < a < + ∞) is a position parameter, corresponding to the mean or median of a stable distribution, and u is a random variable of the feature function.

When α is 2, the α stable distribution degrades to a gaussian distribution;

when α is 1 and β is 0, α has a stable distribution of cauchy;

when β is 0, the α stable distribution is a symmetric distribution about mean α, and we refer to such a distribution as a symmetric α stable (S α S) distribution.

2. Fractional Low Order Cyclic Spectrum (FLOCS) analysis

Since the modulation signal s (t) is contaminated by noise n (t) subject to a stationary distribution, the received signal x (t) can be modeled as:

x(t)＝s(t)+n(t) (11)；

wherein, n (t) is noise complying with the S α S distribution, because the α -stationary-distribution noise has significant spike characteristics and does not have second-order or higher-order statistics, the conventional AMC algorithm based on the second-order or higher-order cyclic statistics fails in the α -stationary-distribution noise, and the fractional low-order cyclic spectrum (FLOCS) obtained by performing nonlinear transformation on the received signal can effectively suppress the α -stationary-distribution noise, so for the AMC technology, corresponding information can be extracted from the FLOCS of the received signal to identify the modulated signal.

FIG. 1 is a schematic block diagram of an embodiment of the present invention.

In this embodiment, the input data is modulated by the modulator in the transmitter to obtain a modulated signal s (t), and then α stable distributed noise n (t) mixed in the channel becomes a received signal x (t) of the receiver.

First, in an automatic modulation classifier, a received signal x (t) is sampled at a sampling frequency F_s＝1/T_sWith uniform sampling, the fractional low-order autocorrelation Function (FLOC) of the sampled discrete signal x (n) can be expressed as:

FLOC(n,m)＝E{[x(n+m)]^{b}[x^*(n)]^{b}} (12)；

x(n)^{b}＝|x(n)|^b-1x^*(n) (13)；

where equation (12) is a b-order nonlinear transformation of the dispersion signal x (n), 0 < b < α/2, E (-) is desired, x^*(n) is the conjugate of x (n). Then, the fractional low-order cyclic autocorrelation (FLOCC) of the signal is:

wherein < · > represents time averaging, and it is worth noting that the b-order nonlinear transformation only changes the amplitude of the signal and does not change the period information, so the cyclic frequency defined under the second-order cyclic correlation is also suitable for fractional low-order cyclic correlation; if b is 1, then FLOCC degenerates to a second order cyclic autocorrelation. FLOCS is a Fourier transform of FLOCC, which can be expressed as:

in practice, the amount of the liquid to be used,the time domain smoothing algorithm, FAM algorithm, can be used to estimate that for a given frequency f and cycle frequency, the time domain smoothing cycle periodogram can be represented by the following equation:

wherein g (n) is NT in width_sUniform weight function of seconds, f₁And f₂Is the center frequency, T, of the filter in the FAM algorithm_sIs the sampling period, where f₁＝f+α/2，f₂＝f-α/2，X_T(r,f₁) And X_T(r,f₂) The complex demodulation of x (n) can be calculated by the following equation.

Wherein a (r) is a duration of T ═ N' T_sA second cone of data window whose width is the frequency resolution Δ f of the FLOCS, which can be estimated unbiased from a time-domain smoothed periodogram if a (r) is normalized, as follows:

3. map domain mapping

Calculated by FAM algorithmThe amplitude of the three-dimensional graph is non-negative, and the calculated FLOCS is subjected to normalization and quantization processing to obtain a discrete fraction low-order cyclic spectrum with the maximum value of 1In the FAM algorithm, the frequency resolution of the FLOCS is Δ f ═ f_sN', cyclic frequency resolution Δ α ═ 1/Δ t ═ f_sN, wherein f_sFor the sampling interval, N 'is the number of data used for complex demodulation, and N is the number of data input during Δ t time, i.e., the FLOCS matrix (N' +1) × (2N +1) calculated by FAM algorithm.

Due to the symmetry of FLOCS, the separation isScattered music scoreThe quadrant of the map establishes a corresponding map domain map. Defining a stable cycle frequency_p,p＝1,2...N，_pThe conditions are satisfied:

the frequency value corresponding to the stable cycle frequency is set as the vertex:taking the amplitude difference between two vertices as an edge, set as:q₁,q₂0,1, N'/2 }, wherein:

to this end, a corresponding map domain map can be obtained at each stable cycle frequencyp is 0, 1.. times.n, it is apparent that the graph at each cycle frequency has cyclicity, and thus the adjacency matrix of the corresponding graph can be extractedAs a discriminating characteristic for different signals.

4. Extracting features

Let the modulation type set beWherein,denotes the K-th modulation type, K being 1,2. In this embodiment, 6 types of modulation type signals, that is, BPSK, 2FSK, 4FSK, QPSK, OQPSK, and MSK, may be identified, and for a k-th type of modulation type training signal without noise, its FLOCS may be calculated, and an image domain set may be constructed according to the method in section 3.

Dividing the sampling sequence of the training signal into L segments, L times of map domain mapping can be established, H maps can be obtained for each time of map domain mapping, and for the ith time of map domain mapping, the set of map domains can be expressed asWhereinH-1, 2.. H, which represents the cyclic frequency at which the training signal of the k-th modulation type remains_hCorresponding diagram, and adjacent matrix set extracted from the corresponding diagram is expressed as

Because FLOCS represents a weighted directed loop, an arbitrary adjacency matrix, in the graph domainIs a sparse matrix of the following nature

Wherein,is a contiguous matrixFor each adjacency matrixExtracting non-zero entries of a secondary diagonal right above a main diagonal of the adjacency matrix, and extracting row index sequences corresponding to the non-zero entriesThe principle of row index sequence extraction is as follows:

For a training signal of a k-th modulation type, extracting an H cycle frequency_hA stable row index sequence, forming a stable row index sequence set:and as a characteristic of the k-th type of modulation.

Note that these row index sequencesIt is not necessary to have the same number of elements,because the length of each sequence is determined by the corresponding adjacency matrixIs determined.

5. Identification of communication signal modulation type

For the received signal, the modulation type is characterized by the method according to the parts 3 and 4, and the row index sequence is collectedWherein V is the number of reserved cycle frequencies;

In this embodiment, as shown in fig. 1, the received signal x (t) is preprocessed and then sent to the classifier for modulation and identification of the communication signal according to the method in the above section 5, and the modulation type is sent to the demodulator, and the preprocessed received signal is demodulated according to the corresponding modulation type, so as to obtain the output data.

As shown in fig. 1, the modulation signal interfered by α stable distributed noise is converted to an image domain by calculating a fractional low-order cyclic spectrum FLOCS, then features of a modulation type training signal are obtained by image domain mapping and feature extraction, and then image domain classification is performed according to the features, so that more stable and effective identification of a modulation type of a communication signal under the interference of α stable distributed noise is realized.

Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, and various changes may be made apparent to those skilled in the art as long as they are within the spirit and scope of the present invention as defined and defined by the appended claims, and all matters of the invention which utilize the inventive concepts are protected.

Claims

1. A map domain communication signal modulation identification method based on a fractional low-order cyclic spectrum is characterized by comprising the following steps:

(1) feature extraction of modulation type training signal

1.1) fractional low order cyclic spectrum based map domain mapping

wherein,cyclic frequency with reserved segment of training signal representing k-th type of modulation_hCorresponding time domain smooth cycle periodic diagram is extracted to obtain H cycle frequencies_hAnd smoothing the adjacent matrix of the cyclic period diagram by the corresponding time domain to obtain an adjacent matrix set:

a1) normalizing and quantizing the calculated FLOCS (fractional low-order cyclic spectrum) to obtain a discrete fractional low-order cyclic spectrum with the maximum value of 1

a2) discrete-based fractional low-order cycles due to FLOCS symmetryIs established correspondingly to the quarter quadrantMap domain mapping:

<mrow> <munderover> <mo>&Sigma;</mo> <mrow> <mi>q</mi> <mo>=</mo> <mn>0</mn> </mrow> <mrow> <msup> <mi>N</mi> <mo>&prime;</mo> </msup> <mo>/</mo> <mn>2</mn> </mrow> </munderover> <msubsup> <mover> <mi>S</mi> <mo>&OverBar;</mo> </mover> <msub> <mi>&epsiv;</mi> <mi>p</mi> </msub> <mrow> <mi>k</mi> <mo>,</mo> <mi>l</mi> </mrow> </msubsup> <mrow> <mo>(</mo> <msubsup> <mi>f</mi> <mi>q</mi> <mrow> <mi>k</mi> <mo>,</mo> <mi>l</mi> </mrow> </msubsup> <mo>)</mo> </mrow> <mo>&NotEqual;</mo> <mn>0</mn> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> <mo>;</mo> </mrow>

<mrow> <msubsup> <mi>V</mi> <msub> <mi>&epsiv;</mi> <mi>p</mi> </msub> <mrow> <mi>k</mi> <mo>,</mo> <mi>l</mi> </mrow> </msubsup> <mo>=</mo> <mo>{</mo> <msubsup> <mi>f</mi> <mi>q</mi> <mrow> <mi>k</mi> <mo>,</mo> <mi>l</mi> </mrow> </msubsup> <mo>,</mo> <mi>q</mi> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mn>1</mn> <mo>,</mo> <mo>...</mo> <mo>,</mo> <msup> <mi>N</mi> <mo>&prime;</mo> </msup> <mo>/</mo> <mn>2</mn> <mo>}</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>4</mn> <mo>)</mo> </mrow> <mo>;</mo> </mrow>

<mrow> <msubsup> <mi>E</mi> <msub> <mi>&epsiv;</mi> <mi>p</mi> </msub> <mrow> <mi>k</mi> <mo>,</mo> <mi>l</mi> </mrow> </msubsup> <mo>=</mo> <mo>{</mo> <mtable> <mtr> <mtd> <mrow> <mi>&Delta;</mi> <msubsup> <mover> <mi>S</mi> <mo>&OverBar;</mo> </mover> <msub> <mi>&epsiv;</mi> <mi>p</mi> </msub> <mrow> <mi>k</mi> <mo>,</mo> <mi>l</mi> </mrow> </msubsup> <mrow> <mo>(</mo> <msubsup> <mi>f</mi> <msub> <mi>q</mi> <mn>1</mn> </msub> <mrow> <mi>k</mi> <mo>,</mo> <mi>l</mi> </mrow> </msubsup> <mo>,</mo> <msubsup> <mi>f</mi> <msub> <mi>q</mi> <mn>2</mn> </msub> <mrow> <mi>k</mi> <mo>,</mo> <mi>l</mi> </mrow> </msubsup> <mo>)</mo> </mrow> </mrow> </mtd> <mtd> <mrow> <msub> <mi>q</mi> <mn>1</mn> </msub> <mo>,</mo> <msub> <mi>q</mi> <mn>2</mn> </msub> <mo>=</mo> <mn>0</mn> <mo>,</mo> <mn>1</mn> <mo>,</mo> <mo>...</mo> <mo>,</mo> <msup> <mi>N</mi> <mo>&prime;</mo> </msup> <mo>/</mo> <mn>2</mn> </mrow> </mtd> </mtr> </mtable> <mo>}</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>5</mn> <mo>)</mo> </mrow> <mo>;</mo> </mrow>

wherein:

1.2) extraction of the line index sequence

For each adjacency matrixExtracting non-zero entries (elements) of a secondary diagonal line directly above the primary diagonal line, and extracting the row index sequence corresponding to the non-zero entries (elements)The principle of row index sequence extraction is as follows:

(2) identification of modulation type of communication signal

computing a set of row index sequencesCharacteristic of modulation type of class kThe Hamming distances of the K Hamming distances are obtainedThen, the minimum Hamming distance is found, and the corresponding modulation type is the modulation type of the received communication signal.