RU2713693C1 - Device for automatic recognition of radio signal keying types - Google Patents

Abstract

FIELD: radio equipment.

SUBSTANCE: invention relates to processing and recognition of radio signal manipulation type and can be used in radio devices. Device comprises a series-connected antenna, an analogue-to-digital converter, a unit for generating in-phase and quadrature signal components, unit for calculating cumulants and a neural network unit, between units for generating in-phase and quadrature signal components and calculating cumulants includes a frequency detuning elimination unit, unit for calculating cumulants calculates values of cumulant modules C₄₂ and C₆₁, and neural network unit provides determination of types of manipulation of radio signals with 4-level pulse amplitude manipulation, binary phase, quadrature phase, phase manipulation of 8th order, quadrature manipulation of 8th, 16th and 32th orders.

EFFECT: wider alphabet of recognizable types of radio signal manipulation from 4 to 7.

1 cl, 4 dwg, 1 tbl

Description

The invention relates to the field of processing of radio signals, in particular to recognition of the type of manipulation of radio signals, and can be used in radio engineering means and systems for the recognition of amplitude, amplitude-phase and phase manipulation of radio signals.

A device for the automatic recognition of types of radio signal manipulation [1] is known, in which the received analog radio signal is sampled by time, quantized by level, in-phase and quadrature components of the radio signal are generated in-phase and quadrature components of the digitized radio signal samples, in the cumulant and quad-phase calculation block component of N samples of the radio signal, cumulants C ₄₀ , C ₂₀ , C _{22 are} calculated, the values of which are fed to the neural network block, which ensure nt determination of the type of radio signal manipulation, the neural network unit consists of i neurons of the input layer, j neurons of the intermediate layer, neurons of the output layer, and each neuron of the previous layer is connected by weight connection to each neuron of the subsequent one. Each neuron consists of a series-connected adder and a nonlinear transducer. The output of each neuron of the output layer is an indicator of the type of manipulation and is connected to the logic device MAX, which provides a comparison of the signals of the neurons of the output layer and the output of the neural network block identification radio signal corresponding to the recognized type of manipulation.

The disadvantage of this device is the limited alphabet of recognizable types of radio signal manipulation by four: AM, KAM, FM and FM.

The known device [1] is the closest to the proposed invention and is selected as a prototype.

The device [1] does not require high-precision synchronization by the carrier frequency, but it only recognizes four types of radio signal manipulation - AM, KAM, FM and FM, which does not allow achieving the technical result indicated below.

The technical result of the invention is the expansion of the alphabet of recognizable types of manipulation of radio signals from 4 to 7, including:

4-level pulsed amplitude-shift keying (4-RAM);

binary phase shift keying (PBSK);

quadrature phase shift keying (QBSK);

8th order phase shift keying (8-PSK);

8th order quadrature manipulation (8-QAK);

16th order quadrature key manipulation (16-QAK);

32nd order quadrature key manipulation (32-QAK).

The specified technical result is achieved due to the fact that the device automatically recognizes the types of manipulation of radio signals, containing a series-connected antenna, an analog-to-digital converter, a block for generating in-phase and quadrature components of the signal, a block for calculating cumulants and a block for neural network, between blocks for generating in-phase and quadrature components signal and calculation of cumulants, the unit for eliminating frequency detuning is included, the unit for calculating cumulants calculates the values of the cumulative modules comrade C ₄₂ and C ₆₁ and the neural network unit provides to determine the types of manipulation signals with 4-level pulse amplitude shift keying, binary phase, quadrature phase shift, phase shift keying of order 8, the quadrature manipulation 8 th, 16 th and 32 th orders.

To make a decision, as in the prototype, a pre-trained neural network of the multilayer perceptron type is used, which shows the high efficiency of constructing multidimensional separating surfaces. In accordance with the corollary of the Kolmogorov-Arnold-Hecht-Nielsen theorem, any multidimensional function of several variables can be represented using a two-layer neural network with direct full connections of fixed dimension [2].

A device for automatically recognizing types of radio signal manipulation is illustrated by the following figures.

FIG. 1. The block diagram of the device for automatic recognition of types of manipulation of radio signals.

The numbers in FIG. 1 marked:

1 - antenna 1;

2 - analog-to-digital converter ADC 2;

3 - block forming the in-phase and quadrature components of the BRS 3 radio signal;

4 - block elimination of frequency detuning BEAM 4;

5 - block calculations cumulants DBK 5;

6 - block neural network 6;

FIG. 2. The block diagram of the algorithm of operation of the BCHR 4.

The notation used in the block diagram of the algorithm:

index - a function that returns the number of an array element;

max - a function that returns the value of the maximum element of the array;

Re is the real part of the complex argument;

Im is the imaginary part of a complex argument.

FIG. 3. Block calculation of cumulative DBK.

The numbers in FIG. 3 are indicated:

5.1 - block calculation of the moment M ₄₂

5.2 - block calculation of the moment M ₂₀

5.3 - block calculation of the moment M ₂₁

5.4 - block calculation of the moment M ₆₁

5.5 - block calculation of the moment M ₄₁

5.6 - block calculation of the moment M ₄₀

5.7 - block calculation module cumulant C ₄₂

5.8 - block calculation module cumulant C ₆₁

FIG. 4. A neural network such as a multilayer perceptron.

The numbers in FIG. 4 are indicated:

6.1 - input layer;

6.2 - hidden layer;

6.3 - output layer;

6.4 - logical device MAX.

The claimed device for automatic recognition of types of manipulation of radio signals works as follows:

1. An analog radio signal having a duration Ts received by antenna 1 is fed to ADC 2, where it is sampled in time and quantized in level. Then, from the output of the ADC 2, the digitized radio signal enters the BRS 3, where the in-phase I 'and quadrature Q' components of the digitized samples of the radio signal are formed, resulting in an array of N complex samples of the radio signal.

2. From the output of the BRS 3, the in-phase I 'and quadrature Q' components of the digitized samples of the radio signal are received at BEAM 4, where the values of N digitized samples eliminate the mismatch in the carrier frequency of the input radio signal, for which a forced detuning of its frequency with a fixed mismatch step ƒ _Δ in the range of the expected detuning with the calculation of the cumulant module C _{40 of the} radio signal for each detuning value. The criterion for the minimum frequency detuning is the maximum value of the C ₄₀ cumulant module.

The expected detuning range has a lower f _min and an upper ƒ _max boundary and coincides with the frequency analysis band of the device. The step of analyzing the frequency detuning search range is ƒ _Δ .

where Δ _ϕ is the absolute change in the phase of the radio signal during the time Ts, selected in the range from 0.1 to 0.2.

The frequency detuning is carried out by element-wise multiplication of the array of samples of the input radio signal by the discrete values of the complex exponent defining the frequency shift of the radio signal dƒ

For each radio signal so, the cumulant C _{40 is} calculated [3]

where M _rr are the joint moments of orders rr calculated by the formula

where s is the original radio signal, consisting of N samples.

The method of complete enumeration of detuning values from a given range of values determines the detuning to which the maximum value of the C ₄₀ cumulant module corresponds. This value is an estimate of the frequency mismatch Δ _ƒ .

Partial detuning is eliminated by element-wise multiplication of the array of samples of the input radio signal by discrete values of the complex exponent having a frequency equal to the estimate of the frequency detuning

The flowchart of the frequency detuning elimination unit is shown in FIG. 2.

3. The radio signal with the eliminated mismatch in the carrier frequency from the output of the DRU 4 is fed to the input of the DBK 5, in which the modules of the values of the cumulants C ₄₂ and C ₆₁ are calculated [3].

where M are joint moments.

4. The calculated values of the cumulant modules C ₄₂ and C ₆₁ are fed to the input of the BNS 6. The BNS 6 is a multilayer perceptron neural network (Fig. 3), consisting of two neurons of the input layer 6.1, ten neurons of the hidden layer 6.2 and seven neurons of the output layer 6.3. Each neuron of the layer 6.2, 6.3 consists of a series-connected adder and a nonlinear transducer. A nonlinear converter implements a function of the sigmoid type. For each neuron of the hidden layer, the values of the cumulant modules C ₄₂ and C ₆₁ are received in the form of signals of the corresponding level (amplitude), which are added to the adder with the corresponding weights, the received signal from the adder output goes to a non-linear converter, from which the signal received after conversion is input each neuron of the output layer. The signal from the output of each neuron of the output layer 6.3 is fed to the logic device MAX 6.4.

5. The logic device MAX 6.4 compares the signals from the outputs of the neurons of the output layer 6.3. Each neuron of the output layer 6.3 corresponds to a certain type of manipulation. At the output of the logic device MAX 6.4, an identification signal is generated corresponding to the neuron number of the output layer with the maximum output signal.

The results of evaluating the effectiveness of recognition of the types of manipulation of the proposed device are shown in table 1. The recognition results are obtained by the simulation method of the device with the following initial data:

the number of samples of radio signals for each type of manipulation 1000;

the number of samples of the radio signal N = 2000;

signal to noise ratio 5 dB;

frequency detuning - a random variable with a uniform distribution;

the initial phase of the radio signal is random with a uniform distribution.

An advantage of the invention is that, compared with the prototype, the claimed device allows to solve the problem of automatic recognition of types of manipulation of radio signals for a wider alphabet of types of manipulation. The inventive device allows you to recognize various types of manipulation (BPSK, QPSK, 8-PSK, 4-RAM, 8-QAM, 16-QAM, 32-QAM) with a probability of at least 0.85 with a signal to noise ratio of at least 5 dB.

Literature

1. RU 2510077.

2. Hecht-Nielsen R. Kolmogorov's Mapping Neural Network Existence Theorem // IEEE First Annual Int. Conf. on Neural Networks, San Diego, 1987, Vol. 3, pp. 11-13.

3. William H. Clark IV. Blind comprehension of waveforms through statistical observations. Robert W. McGwier, Chair T. Charles Clancy R. Michael Buehrer. Blacksburg, Virginia, 2014. pp. 12.

Claims (1)

A device for the automatic recognition of types of manipulation of radio signals, containing a series-connected antenna, an analog-to-digital converter, a block for generating in-phase and quadrature components of the signal, a block for calculating cumulants that calculates the values of cumulants, and a block for neural network, characterized in that the device is between the blocks for generating in-phase and the quadrature components of the signal and the cumulant calculation, the frequency detuning elimination unit is included, the cumulant calculation unit calculates the values of cumulant modules C ₄₂ and C ₆₁ , and the neural network block provides determination of signal manipulation types with 4-level pulse amplitude manipulation, binary phase, quadrature phase, 8th-order phase manipulation, 8th, 16th and 32- quadrature manipulations th orders.

Images