RU2613733C1 - Method for recognizing radio signals - Google Patents

Abstract

FIELD: radio engineering, communication.

SUBSTANCE: method for recognizing radio emissions, which consists in the fact that the signal information part is selected to form the recognition features of the radio signals by the structural and temporary construction. The selected signal part is digitized and processed. Then the number of the most informative approximation coefficients is estimated for the continuous wavelet transformation, the signal vector is formed based on the number of the most informative coefficients. The integral indicator is calculated by using the energy characteristic indicators of the signal and the wavelets. The wavelet function, providing the maximum contrast of the radio signal feature - scalograms, is selected by the maximum value of the integral indicator. The signal scalogram is constructed by the continuous wavelet transformation, which is taken as the recognition feature.

EFFECT: enhancing the contrast of the radio recognition features.

8 dwg

Description

The invention relates to the recognition of radio signals based on the analysis of the signal scalogram and can be used in the technical means of signal recognition in real reception conditions in the presence of noise and interference.

There is a known method of signal recognition according to the patent of the Russian Federation No. 2133501 MPK5 G06K 9/00 of 04/04/98, which consists in comparing on a sample of length Q channel symbols a "sliding window" of all combinations of length K with each other, where K / 2 is the estimated memory of the noise-resistant convolutional encoder , and the subsequent calculation of the frequency of occurrence of events consisting in the fact that in two segments of the code sequence there is a single pair of mismatched characters, and these symbols do not belong to two different sub-ensembles of the partition at the same time.

The disadvantage of this method is the narrow scope of its application, because when using this method, only signals using signal-code constructs can be recognized.

A known method for the recognition of radio signals based on the singular decomposition of the pseudo-frequency-time-distribution (pseudo-HRM) Wigner-Ville [N.M. Marinovic, G. Eichmann. An expansion of Wigner distribution and its applications. - Proc. IEEE ICASSP-85, 1985, pp. 1021-1024]. In this method, preliminary based on the Wigner pseudo-FWM, energy distribution matrices (REs) of the reference radio signals are formed, their spectral decomposition is performed, parameters of the reference radio signals are formed, then a recognized radio signal is received, sampled and quantized, the received energy signal distribution matrix is formed, and signs of received the radio signal, compare them with the parameters of the reference radio signals and the results of the comparison identify the received radio signal.

The disadvantage of this method is the low probability of correct recognition [J. Tu, R. Gonzalez. Pattern recognition principles. Per. from English - M .: Mir, 1978. - p. 142] of radio signals of complex time-frequency structure, as well as radio signals when exposed to noise and interference, which is due to the features of Wigner-Ville [Cohen L. Time-frequency used for recognition of pseudo-CVMs] distribution. Review // TIIER, 1989, v. 77, No. 10, p. 72-121].

The probability of correct recognition is reduced due to the appearance of interference background and false power peaks in the Wigner-Ville pseudo-FWM, which distort the real picture of the signal energy distribution in frequency-time coordinates.

The closest in technical essence to the claimed one is a method for recognizing radio signals (according to RF patent 2261476 IPC: G06K dated 01/26/2004) based on a continuous wavelet transform of a sequence of discrete quantized samples of radio signals. In this method, when generating the energy distribution matrix of the reference radio signals, a continuous wavelet transform of the sequence of discrete quantized samples of the reference radio signals is calculated, the energy distribution vectors are generated, the total covariance matrix is calculated, the truncated matrix of eigenvectors is formed, the truncated matrix of eigenvectors is multiplied by the energy distribution vectors of the reference radio signals, from the generated matrix of the energy distribution of the received radio signal to form the energy distribution vector is multiplied and the truncated matrix of eigenvectors is multiplied by the energy distribution vector, and the calculation results are taken as signs of recognition of the received radio signal.

The disadvantage of the prototype method is the difficulty of choosing the most suitable from the point of view of the effectiveness of identifying the characteristics of the analyzed wavelet function signal, which leads to a decrease in the contrast of signs of recognition of the radio signal.

The purpose of the claimed technical solution is to increase the contrast of signs of recognition of radio signals - wavelet spectrograms (scalograms) in the conditions of receiving a real radio signal by selecting the most optimal wavelet.

This goal is achieved by the fact that in the known method of recognizing radio signals, which consists in the fact that to generate the reference energy distribution matrices (RE), continuous wavelet transform (NVP) is calculated, the RE vectors of the reference radio signals are formed, for which the total covariance matrix is calculated, the matrices are calculated eigenvalues and eigenvectors of the covariance matrix, form a truncated matrix of eigenvectors, form the parameters of the reference radio signals by multiplying the truncated matrix of the eigenvectors to the RE vectors of the reference radio signals, the calculation results are averaged, the RE matrix of the received signal is formed by calculating the NVP and the RE vector, the truncated matrix of eigenvectors is multiplied by the RE vector of the received radio signal, the calculation results are taken as signs of recognition of the received radio signal, in the claimed method for the formation of signs of recognition of radio signals by structural-temporal construction, the information part of the signal is selected. The selected signal section is entered into the Matlab computer mathematics system, where the signal is digitized and processed. Then, the number of the most informative approximation coefficients for the NIR is estimated, the signal vector is formed taking into account the number of the most informative coefficients. Using indicators of the energy characteristics of the signal and wavelets, an integral indicator is calculated. Based on the maximum value of the integral indicator, the wavelet function is selected that provides the maximum contrast of the sign of the radio signal - the scalogram. A scalogram of the signal is constructed by NVP, which is taken as a sign of recognition.

The claimed method is illustrated by drawings and tables, which show:

 FIG. 1 is a spectrogram of a signal of standard MIL-STD-188-110A;

 FIG. 2 - an oscillogram of the information part of the Li signal of the MIL-STD-188-110A standard;

 FIG. 3 is a waveform of informative approximation coefficients (ni) for direct continuous wavelet transform;

 FIG. 4 - contour plots and 3D plots of matrix (external) multiplication of the signal vector by the wavelet vector Meyer (a) and by the wavelet vector Db6 (b);

 FIG. 5 is a table of the results of calculating the mutual energy of the information part Li of the signal S and the wavelet;

FIG. 6 - the table of the results of calculating the integral indicator of the efficiency of the wavelet K _Σ ;

 FIG. 7 is a block diagram of a radio signal recognition algorithm;

FIG. 8 is a diagram of the entropy assessment of the sign contrast of the scalogram for various types of wavelet (a) and the integral index K _Σ for various types of wavelet (b).

The recognition process includes the following procedures: selection from the structural-temporal region L of the analyzed signal S (Fig. 1) the information (from the point of view of recognizing and identifying features) part of Li, processing and digitization of the selected structural-temporal region in the computer mathematics system (SCM ) Matlab (Lic), evaluation and selection of the most informative approximation coefficients for direct continuous wavelet transform - ni, forming Lcin length signal vector S with the ni, selections based on the integral index K _Σ most e optimal from the standpoint of improving the contrast scalograms analyte indicative signal wavelet - Wi, direct continuous wavelet transform structurally Lic time domain signal and forming scalograms Swi signal. Signal recognition is carried out by comparing the parameters of the matrices of the scalogram of the standard and the analyzed signal using one of the known methods [Method for the recognition of radio signals (RU 2464633), method for the recognition of radio signals (RU 2423735)].

An important condition for increasing the contrast of the signs of recognition of radio signals - wavelet spectrograms - is a reasonable choice of the wavelet function for the analysis of the signal under study.

The signal energy is equal to the integral of the power over the entire interval of the signal [Baskakov S.I. Radio engineering circuits and signals: Textbook for universities. - M.: Higher School, 1988, p. 74]:

where E _s is the signal energy s (t);

w (t) is the signal power s (t).

One of the main metric parameters of linear space in mathematics is the signal norm. For the analysis of signals, quadratic norms are often used [S. Baskakov. Radio engineering circuits and signals: Textbook for universities. - M.: Higher School, 1988, p. 25]:

For a discrete signal case:

where s (n) are the discrete samples of the signal, n = 1, 2 ... ∞.

Based on the expressions (1) and (2), the signal energy is determined:

The energy of interaction of the signals s (t) and v (t) is determined by the expression [S. Baskakov Radio engineering circuits and signals: Textbook for universities. - M.: Higher School, 1988, pp. 72-73]:

Expression (6) corresponds to the doubled scalar product of the signals s (t) and v (t) [Davydov A.V. Signals and linear systems: Thematic lectures. - Yekaterinburg: USMU, IGiG, 1, T2, p. 9] and determines the interaction energy in the time domain:

When choosing a wavelet for signal analysis, it is necessary to ensure the greatest coincidence of the time-frequency characteristics of the signal and the wavelet, while the mutual energy of the signal and the wavelet should tend to the full signal energy with an ideal coincidence of the signal and wavelet functions.

To assess the coincidence of the time-frequency characteristics of the wavelet with the signal under study, expression (6) is used, which determines the mutual energy through the scalar product of the signals.

Using the Matlab matrix laboratory and expression (6), the energy of a real signal representing the information part of the MIL-STD-188-110A standard signal was determined (Fig. 2).

Listing on the Matlab command line:

y = wavread ('V12687_all', [80 365]); % signal conditioning

>> y1 = y (1:64, 1: 1); % cropping vector c 1 to 64 count

z = trace (abs (y1 * y1 '))% scalar product of signals (if they coincide completely, it corresponds to the signal energy);

z = 0.0824.

In accordance with expression (4), the signal energy is equal to the square of the second norm of this signal.

Listing on the Matlab command line:

z1 = norm (y1,2)% the second norm of the signal y1 in accordance with expression (5)

z1 = 0.2870

z2 = z1 ^∧ 2

z2 = 0.0824.

The total energy of the signal, calculated through the square of the second norm, is 0.0824 and completely coincides with the energy calculated through the scalar product [S. Baskakov Radio engineering circuits and signals: Textbook for universities. - M.: Higher School, 1988, p. 29].

The mutual energy of the signal and the wavelet Db6 is determined [Dyakonov V.P. - Wavelets. From theory to practice. Solon-R. 2002, p. 155-158].

Listing on the Matlab command line:

[phi, psi, x] = wavefun ('db6', 4); % wavelet formation

Y = [psi, x];

Y1 = Y (1: 1.46: 109); % cropping the wavelet vector according to the length of the signal vector.

Y1n = Y1 * (max (y1)) / (max (Y1)); % normalization of the wavelet vector by the maximum amplitude of the signal.

z = trace (abs (y1 * Y1n '))% scalar product of the signal vector and wavelet:

z = 0.0576.

The mutual energy of the signal and the wavelet Db6 - E _sν is 0.0576, which is much less than the total signal energy.

The results of calculating the mutual energy E _{sν of a} signal and various wavelets most frequently used in wavelet analysis [V. Dyakonov - Wavelets. From theory to practice. Solon-R. 2002. - pp. 155-158] are given in the table (Fig. 5).

A good estimate of the coincidence of the time-frequency characteristics of the wavelet with the signal under study is provided by the matrix (external) multiplication of the signal vector by the wavelet vector in accordance with expression (7) [E.B. Winberg. Algebra course: 2nd ed., Rev. and add. - M .: Publishing house "Factorial Press", 2001, p. 318]:

where x ₁ , x ₂ , ... x _N - discrete samples of the values of the wavelet function;

y ₁ , y ₂ , ... y _N - discrete samples of signal values.

The matrix of external multiplication carries generalized information about the signals: the diagonal of the matrix corresponds to the multiplication of the signal function by the wavelet function in the coordinate plane, which in the frequency domain corresponds to the convolution of the Fourier transform of the functions and characterizes the degree of coincidence of the signal and wavelet functions, the sum of the elements of the matrix diagonal vector is scalar signal multiplication and a wavelet, which corresponds to the mutual energy of the signal and wavelet functions.

In FIG. Figure 4 shows the results of matrix (external) multiplication of a signal vector by a wavelet vector by a contour graph and a 3D graph using the Meyer and Db6 wavelets. For the Meyer wavelet, the matrix is weakly filled, this is also clearly visible on the three-dimensional graph. For the Db6 wavelet, the matrix filling is much higher.

In accordance with expression (3), to evaluate the matrix, the second norm is calculated, which finds the largest singular number.

From the graphs shown in FIG. 4, it is obvious that the matrix norm of the product of the signal vector by the wavelet vector will be larger for the Db6 wavelet, since the absolute values of the matrix elements have large values and the matrix is better filled. This is confirmed by the results of the calculation of the second matrix norm for the Db6 and Meyer wavelets: N1 _Db6 = 0.0883; N1 _Mtyer = 0.0551.

Direct continuous wavelet transform with the use of one or another wavelet allows to obtain the matrix of spectrograms (scalograms) of the analyzed signal. The matrix energy is estimated by the second norm N2, which will characterize the efficiency of the selected wavelet.

The analysis of the energy characteristics of the signal and wavelets allows you to generate indicators for their assessment: N1, N2, E _sv . The values of these indicators for a number of the most frequently used wavelets are shown in FIG. 6. The table shows the norm values of the matrix of the external multiplication of the signal and wavelet vector — N1, the norm values of the wavelet spectrogram matrix — N2, the mutual energy of the signal and the wavelet — E _sv and their reduced values.

The integral indicator allows a comprehensive assessment of the effectiveness of the wavelet transform using one or another wavelet (8):

where N1, N2, E _sv - reduced values of the norms of the corresponding matrices and mutual energy.

The integral wavelet efficiency indicator justifies the choice of the wavelet that provides the greatest coincidence of the time-frequency characteristics of the signal and the wavelet, while the mutual energy of the signal and the wavelet tends to the total energy of the analyzed signal. To analyze the information part of the Li signal of the MIL-STD-188-110A standard, such a wavelet is Db6, this is confirmed by the graphs shown in FIG. 4, the wavelet spectrogram of FIG. 6.

The results of calculating the integral wavelet efficiency index are shown in FIG. 6.

The claimed method is implemented as follows (Fig. 7).

Initially, the informational (from the point of view of recognizing and identifying features) part of the Li signal S is selected according to the structural-temporal construction (Fig. 1).

The selected signal section is entered into the Matlab computer mathematics system for digitization (Lic) and further processing (Fig. 2).

The number of the most informative approximation coefficients (ni) for the direct continuous wavelet transform is estimated (Fig. 3).

The signal vector Licn is formed taking into account the number of the most informative coefficients ni.

The integral indicator K _{Σ is formed} on the basis of the energy characteristics of the signal and the wavelets used for analysis.

Based on the maximum value of the integral index K _Σ , the type of wavelet Wi is selected, which ensures the maximum contrast of the recognition sign - Swi scalograms.

A sign of recognition is formed - Swi scalograms.

It is decided that Swi belongs to one of the Swi standards.

The choice of a wavelet for signal analysis using an integral indicator increases the sign contrast, scalograms have more details, and more data can be extracted from the analysis.

For a quantitative characteristic of the characteristic contrast of the scalograms, allowing to evaluate their information content, the entropy is chosen.

Entropy for continuous wavelet transform characterizes the fuzziness of the analyzed signal in the basic directions [A. Solonina Digital signal processing. Modeling in MATLAB / A.I. Solonina, S.M. Watermelons. - St. Petersburg: BHV-Petersburg, 2008. - S. 687]. It takes smaller values when the bulk of the signal energy is distributed over a small number of basis functions and, conversely, takes large values when the main signal energy is distributed over a large number of basis functions. In general, this allows us to evaluate the informativeness of the signal scalogram.

In FIG. Figure 8 shows a diagram of the entropy assessment of the sign contrast of the scalogram for various types of wavelet (Fig. 8a) and the value of the integral index K _Σ (Fig. 8b). The diagrams show that a larger value of the integral indicator corresponds to a larger value of entropy, i.e. the scalogram has a greater sign contrast.

The calculations (Fig. 8) allow us to make a conclusion about the following assessments of the contrast of signs.

When choosing features according to the algorithm implemented in the prototype, any of the available set of wavelet function - W _{i is} equally likely to be used, then we have i features with different contrast. The average contrast can be calculated by the formula:

- среднее значение энтропии;Where

- the average value of entropy;

E _i is the entropy of the i-th attribute, i = 1, 2, ... n;

P _i - the probability of choosing the i-th attribute.

, что составляет 82% от максимального значения контрастности.For the feature selection option shown in FIG. 8, the average contrast will matter

, which is 82% of the maximum contrast value.

The proposed method for the recognition of radio signals allows, due to the optimal choice of the wavelet function, to uniquely ensure the maximum contrast of the feature - 26.3. Thus, for the considered variant of the recognition of radio signals, the contrast of the feature increases by an average of 18%. When choosing the worst sign in contrast, the loss in contrast of the sign for the prototype will be 58%.

When constructing scalograms based on a continuous wavelet transform to solve the recognition problem according to the algorithm implemented in the prototype, signs are not evaluated depending on the selected wavelet, which leads to the selection of low-contrast signs.

The proposed recognition method through the use of an integral indicator to select the optimal wavelet function allows you to reasonably choose the most contrasting signs for the signal under study.

Claims (1)

The method of recognition of radio signals, which consists in the fact that for the formation of signs of recognition of radio signals according to the structural-temporal construction, the information part of the signal is selected, which is entered into the Matlab computer mathematics system, where the signal is digitized and processed, the number of the most informative approximation coefficients for direct continuous wavelet is estimated transformations (NVP), a signal vector is formed taking into account the number of the most informative coefficients, then the integ is calculated the total wavelet efficiency indicator by the greatest coincidence of the time-frequency characteristics of the signal and the wavelets used to analyze the wavelet, the wavelet function is selected according to the maximum value of the integral wavelet efficiency indicator for the analysis of the signal under study, by which a waveform of the signal is received by continuous wavelet transform, which is received in as a sign of recognition of the radio signal.

Images