CN109274626A

CN109274626A - A kind of Modulation Identification method based on planisphere orthogonal scanning feature

Info

Publication number: CN109274626A
Application number: CN201811389534.2A
Authority: CN
Inventors: 王卫东; 甘露; 廖红舒
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2018-11-21
Filing date: 2018-11-21
Publication date: 2019-01-25
Anticipated expiration: 2038-11-21
Also published as: CN109274626B

Abstract

The invention belongs to fields of communication technology, particularly relate to a kind of Modulation Identification method based on planisphere orthogonal scanning feature.The present invention utilizes orthogonal scanning, compared to the way that planisphere is converted to colour density spectrogram picture using Statistics of Density window, by the time complexity in signal characteristic abstraction stage from O (n²) it has been reduced to O (n), and since the signal characteristic extracted is one-dimensional, it is possible to more simple neural network classifier is selected, occupancy resource is less, and calculation amount is smaller, and faster, recognition accuracy is also promoted recognition speed.

Description

A kind of Modulation Identification method based on planisphere orthogonal scanning feature

Technical field

The invention belongs to fields of communication technology, particularly relate to a kind of modulation based on planisphere orthogonal scanning feature Recognition methods.

Background technique

Since four scholars such as C.S.Weaver delivered first in 1969 on the technical report of Stanford University Since the article for studying signal of communication Automatic Modulation Recognition, the Automatic Modulation Recognition technology of signal of communication is always the communications field Research hotspot is suffered from the fields such as electronic reconnaissance and confrontation, spectrum monitoring and management and is widely applied, for communication intelligence Change is of great significance.Existing Modulation identification technology is broadly divided into two major classes: based on the assumption that the maximum likelihood method and base examined In the mode identification method of feature extraction.

Based on the assumption that the maximum likelihood method examined is derived using probabilistic model, reach maximum in the probability density of observation When find out parameter Estimation amount the most reasonable, for the angle of Bayesian Estimation, as a result, optimal, but such method pair It is more sensitive in parameter error and model mismatch, it is difficult to be widely applied in the complex communication environment of reality.Compared to based on vacation If the maximum likelihood method examined, the mode identification method based on feature extraction is more stable, and practicability is stronger.Currently, being used for The signal characteristic of Modulation Identification mainly includes the temporal signatures such as instantaneous amplitude, frequency and phase, planisphere geometrical characteristic, when frequency division Cloth feature, Higher-Order Statistics Characteristics, cyclostationary characteristic etc..

The planisphere of digital modulation signals can intuitively reflect the modulation type of signal, but tradition is based on planisphere geometry The recognition methods of feature such as Fuzzy C-Means Clustering etc. counts the way of constellation symbols cluster, has higher want for signal-to-noise ratio It asks.For this problem, there is researcher to propose that common planisphere the close of colour can be converted to using Statistics of Density window Degree spectrum view, the i.e. statistical picture about symbol sampler point distribution density on planisphere, and the neural network recognization image is utilized, So that applicability of the method based on planisphere geometrical characteristic under low signal-to-noise ratio greatly enhances.But this method is mentioned in signal characteristic The time complexity for taking the stage is O (n²), and in order to reach higher image recognition precision, use depth convolutional Neural net Network, such as GoogLeNet, AlexNet model, the number of parameters of these models all in million and ten million magnitude, occupy resource compared with More, calculation amount is larger, thus recognition speed is also relatively slow.

Summary of the invention

The purpose of the present invention provides a kind of modulation knowledge based on planisphere orthogonal scanning feature aiming at the above problem Other method, the distribution that essence is still the symbol sampler point on to planisphere carry out Statistics of Density.

The technical solution adopted by the invention is as follows:

A kind of Modulation Identification method based on planisphere orthogonal scanning feature is mainly used for identifying digital modulation mode (PSK/QAM), which is characterized in that the Modulation Identification method the following steps are included:

S1: ready signal data

Assuming that signal modulation mode to be identified has v kind, current class label t indicates each modulation system, then to be identified Signal collection is represented by T={ t | t=1,2 ..., v }.Emulation signal is generated by MATLAB, then simulates actual signal hair It send, transmit and receive process, or actual signal is directly acquired by signal receiver.Finally obtained is baseband signal, if M symbol of acquisition every time, obtains symbol sebolic addressingSequential element (x_i, y_i) it is symbol s_iIn planisphere On coordinate.

S2: orthogonal scanning is executed to planisphere

Set the width w and sweep length n of Statistics of Density window, it is assumed that scan, then scan since the upper left corner of planisphere Initial position isScanning process is as shown in Figure 1.

The X-direction of planisphere is scanned to obtain the Statistics of Density vector that length is nY direction is scanned It arrivesSpecific scanning process is as follows:

X=x₀, y=y₀

Fork=1,2 ..., ndo

X '=x+w/2

Y '=y-w/2

d_x(k)=count_x(S, x, x ')

d_y(k)=count_y(S, y, y ')

X=x '

Y=y '

end for

Wherein, function count_x(S, x, x ') meets abscissa more than or equal to x and less than x ' for searching in symbol sebolic addressing S Symbol, return meet condition symbol sampler point quantity；Similarly, function count_y(S, y, y ') is for searching symbol Meet the symbol that ordinate is greater than y ' less than or equal to y in sequence S.

Executing the signal characteristic that orthogonal scanning obtains to planisphere is d_xy=(d_x, d_y)。

S3: training neural network classifier

Step S1, S2 are repeated, for every kind of modulation system that preset signals are concentrated, obtains its a large amount of signal characteristic sample d_xy, then construction trains data set D={ (d used_xy, t) }, while neural network as shown in Figure 2 is built, it is calculated using Adam Method optimizes the neural network.

S4: identification signal of communication

Letter is extracted using digital communication signal of the signal characteristic extracting methods mentioned in step S2 to unknown modulation system Number feature d_xy, it is then input to the neural network classifier that training finishes in step S3, obtains recognition result t.

Beneficial effects of the present invention are that the present invention is based on the Modulation Identification methods of planisphere orthogonal scanning feature, are utilized The advantage of orthogonal scanning, compared to the way that planisphere is converted to colour density spectrogram picture using Statistics of Density window, by density The sliding number of statistic window is from n²It reduces to n, so that the time complexity in signal characteristic abstraction stage is from O (n²) have decreased to O (n)；It, can be with and because the obtained signal characteristic of this method is one-dimensional, three-dimensional relative to identification color image feature Simpler neural network classifier is selected, occupancy resource is less, and calculation amount is smaller, and recognition speed is faster；In addition, by constellation Figure executes the Statistics of Density feature that orthogonal scanning obtains, and is equivalent to and does in X-direction and Y direction to the density spectra of planisphere Secondary statistics, the signal characteristic of formation can more be stablized and effectively, help to promote final recognition accuracy, experiment is also demonstrate,proved This point is illustrated.

Detailed description of the invention

Fig. 1 is that the process proposed by the present invention for executing orthogonal scanning to signal constellation (in digital modulation) figure is illustrated；

Fig. 2 is the structural representation for the neural network classifier that the present invention selects；

Fig. 3 is the case where average recognition accuracy of signal in the embodiment of the present invention changes with signal-to-noise ratio；

Fig. 4 is signal identification result when signal-to-noise ratio is 5dB in the embodiment of the present invention.

Specific embodiment

The present invention will be described in detail with reference to the accompanying drawings and examples, so that those skilled in the art can be more preferable Ground understands the present invention.

Embodiment

The purpose of the present embodiment be the digital communication signal of different modulating mode is identified, and to recognition accuracy into Row verifying.For data source in the present embodiment in actual wireless communication signals, signal transmission rate is 2M Baud, includes Five kinds of digital modulation modes { BPSK, QPSK, 8PSK, 16QAM, 64QAM }.It is responsible for receiving signal by signal receiver, through a system Column signal pretreatment process, obtained baseband signal signal-to-noise ratio is about 35dB.Sampling obtains 8192 symbols every time, i.e., M=8192 constitutes symbol sebolic addressing S, sets the width w=0.15 of Statistics of Density window, sweep length n=256, discribed to S Planisphere executes orthogonal scanning, obtains signal characteristic d_xy, 20,000 operations are repeated to the signal of every kind of modulation system, are obtained total 100,000 signal characteristic samples are counted, data set D={ (d is then constructed_xy, t) }, t=1 here, 2 ..., 5, correspond to aforementioned five Kind modulation system.Then with data set D training neural network classifier.In order to test the proposed method of the present invention under low signal-to-noise ratio Performance, by directly adding the white Gaussian noise of varying strength to baseband signal so that signal-to-noise ratio changes 0 between 9dB, And under every kind of signal-to-noise ratio, average recognition accuracy is taken using cross validation, as a result as shown in figure 3, when wherein signal-to-noise ratio is 4dB Specific identification situation it is as shown in Figure 4.

Claims

1. a kind of Modulation Identification method based on planisphere orthogonal scanning feature, which comprises the following steps:

S1, signal data is obtained:

Assuming that signal modulation mode to be identified has v kind, each modulation system is indicated with class label t, then by signal to be identified Set representations are T={ t | t=1,2 ..., v }；

If m symbol of acquisition every time, collects symbol sebolic addressingSequential element (x_i, y_i) it is symbol s_iCoordinate on planisphere；

S2, orthogonal scanning is executed to planisphere:

The width w and sweep length n of Statistics of Density window are set, and is scanned since the upper left corner of planisphere, i.e., then scans starting Position is

The X-direction of planisphere is scanned to obtain the Statistics of Density vector that length is nY direction is scanned to obtainSpecific scanning process are as follows:

Wherein, function count_x(S, x, x '), which is used to search, meets the symbol that abscissa is less than x ' more than or equal to x in symbol sebolic addressing S Number, return to the quantity of the symbol sampler point for the condition that meets；Function count_y(S, y, y ') is indulged for searching to meet in symbol sebolic addressing S Coordinate is greater than the symbol of y ' less than or equal to y；

Executing the signal characteristic that orthogonal scanning obtains to planisphere is d_xy=(d_x, d_y)；

S3, training neural network classifier:

Step S1, S2 are repeated, for every kind of modulation system that preset signals are concentrated, obtains its a large amount of signal characteristic sample d_xy, Then construction training data set D={ (d used_xy, t) }, while neural network is built, and carry out to neural network according to data set Training；

S4, identification signal of communication:

Signal characteristic d is extracted using digital communication signal of the signal characteristic extracting methods in step S2 to unknown modulation system_xy, Then it is input to the neural network classifier that training finishes in step S3, obtains recognition result t.