RU2456743C1 - Method for adaptive noise suppression - Google Patents

Abstract

FIELD: radio engineering.

SUBSTANCE: method for adaptive noise suppression consists in the fact that a transfer function is generated automatically as adaptive to noise, which efficiently suppresses arbitrary noise of unknown spectrum, to generate such a transfer function, the inlet mixture of a signal and noise is digitised, a count sequence is separated into two flows, in every of which a count sequence is multiplied by a weight function, discrete Fourier transforms are calculated, spectra are averaged, a discrete suppression function is generated, this function is multiplied with spectra, reverse discrete Fourier transforms are calculated, results are divided into a weight function, the outlet signal is generated by an alternate selection of a half of counts in each flow.

EFFECT: efficient suppression of both narrow band and wide band noise at an outlet of a receiving device, when noise capacity of an unknown spectrum substantially exceeds signal capacity.

3 cl, 8 dwg

Description

The invention relates to the field of radio engineering and is used in systems for the reception and processing of radio signals, in particular in digital information transmission systems operating under radio interference (hereinafter referred to as interference).

A known method of suppressing interference in receivers of complex signals [1]. With this method, the input signal in the mixture with interference is processed by an equalizing filter and then processed by a filter matched with the signal. The principle of operation of the device in case of interference with an arbitrary spectrum is to use an equalization filter, which leads to interference with an uneven energy spectrum G _p (ω) to interference with a uniform spectrum and spectral density Q _p .

The interference spectrum is aligned by a filter, the module of the function whose transmission | K (ω) | satisfies the condition

G _p (ω) | K (ω) | ² = Q _n, whence | K (ω) | ² = Q _p / G _p (ω).

The phase characteristic of this filter can be any. The filter providing this operation is called an amplitude-frequency equalizer, or simply an equalizer or a whitening filter. The signal s _o (t) at the equalizer output has an energy spectrum

G _sf (ω) = G _with (ω) | K (ω) | ² = | S _c (ω) | ² | K (ω) | ² where

S _c (ω) is the spectral density or simply the spectrum of the signal;

G _c (ω) is the energy spectrum of the signal after filtering.

Thus, at the output of the equalizer, the interference receives a uniform power spectral density Q _p , and the signal - the spectrum G _cf (ω). For such a signal and interference with uniform intensity, the matched filter has a transfer function complex conjugate to the signal spectrum.

Such a receiver passes an elementary frequency interval with amplification proportional to the amplitude of the spectral component of the signal and inversely proportional to the spectral intensity of the interference in this frequency interval. The signal-to-noise ratio at the output of this filter is higher, the greater the difference in the spectra of the signal and interference. In the general case, the intensity maxima and minima in the interference spectrum have a random distribution along the frequency axis. Therefore, the equalizer should the interference have a random distribution along the frequency axis. Therefore, the equalizer should be a device with a frequency response that varies over time. If you imagine an equalizer in the form of a set of parallel-connected narrow-band filters with non-overlapping frequency characteristics

,

,

the shape and band of the i-th narrow-band filter should be determined by the waveform and the spectral density of the interference in the i-th frequency interval.

The disadvantage of this method is the difficulty of implementation, consisting primarily in the fact that the calculation of the spectrum of the signal must be performed on the interval of its existence. Interference suppression will not be effective, for example, in information transmission systems that receive continuously incoming signals against a background of constantly acting interference for a sufficiently long time.

Signs of the invention, coinciding with the signs of the first analogue: linear amplification of the sum of the useful signal and interference.

A known method of compensating for in-channel additive radio interference in receivers of amplitude-modulated, frequency and phase-shifted radio signals and a device for its implementation [2], which produce the compensation of the interference signal from the received mixture of the useful signal and the interference signal by generating samples of the interference signal at times of zero values of the useful signal in the received additive mixture.

The disadvantage of this method is the need for synchronization, i.e. knowledge of the real phase of the signal. Synchronization schemes complicate the achievement of the goal and are inoperative in the conditions of a phase drift of the signal, which is inevitable.

Signs of the invention, coinciding with the signs of the second analogue: linear amplification of the sum of the useful signal and interference, discretization of this sum of signals.

The closest to the operations performed is a method of compensating narrowband interference [3], which is taken as a prototype of the invention.

This method of noise compensation is based on the large similarity of the spectra of interference samples taken at zero and non-zero points of the carrier frequency of the signal. To obtain interference samples taken at zero points of the signal, a mirror spectrum is formed. In order to achieve the effect of compensation (suppression) of interference in the method, the input effect is amplified, filtering in the signal band, sampling and calculation of the direct discrete Fourier transform (DFT). As a result, a discrete spectrum S _in (k) is formed. To form its mirror spectrum Z (k), the real part of Re [Z (k)] at frequency k is replaced by Re [S _in (2k ₀ -k)] (k ₀ is the number of the spectral reference corresponding to the carrier frequency of the signal) and the imaginary the part of Im [Z (k)] is replaced by -Im [S _in (2k ₀ -k)]. Next, the module of the spectrum of the output (restored) signal S _o (k) is formed according to the formula

.

.

To obtain the complex spectrum of the reconstructed signal, the module | S _o (k) | multiply by the phase function exp [θ (k)] = S _in (k) / | S _in (k) |. Then calculate the inverse DFT (ODPF) and form samples of the output signal s _o (n) with suppressed interference.

However, this method has disadvantages, the most significant of which are the following: the inability to work in conditions where the interference power significantly exceeds the signal power, and insufficient noise immunity when receiving continuously incoming signals against a background of interference.

Signs of the invention, coinciding with the signs of the prototype:

- linear amplification of the sum of the useful signal and interference;

- discretization of this sum of signals;

- the formation of the spectrum of the sum of the useful signal and interference by means of a discrete Fourier transform (DFT), obtained samples of this sum of signals, and an inverse discrete Fourier transform (DFT).

The present invention is a method of adaptive noise reduction solves the problem of improving noise immunity when receiving digital and analog information transmitted using various types of modulation, including the reception of continuously incoming signals against a background of time-varying interference with an unknown spectrum and in conditions where the interference power is significant exceeds signal strength.

The technical result of the invention is the effective suppression of both narrowband and wideband interference at the output of the receiving device, when the interference power with an unknown spectrum significantly exceeds the signal power.

The proposed method does not require the introduction of phase or frequency synchronization and can be used for adaptive interference suppression in the amplification paths of high or intermediate frequency signals.

The developed adaptive interference suppression method has various options for its implementation. The choice of a specific implementation of the method is due to functional tasks and the field of its practical use.

The adaptive interference suppression method in the absence of external information about the interference characteristics is implemented as follows. To effectively suppress interference, an input signal processing procedure is being built that automatically adapts the interference suppression parameters to the current interference environment. When solving the problem, data on the energy spectrum of the received signal G _c (ω), which are known, are used. To form the required transfer function, the additive sum of the useful signal and interference is amplified, its discretization is performed, as a result of which a continuous stream of samples s _in (n) of the mixture of the useful signal and interference is formed.

This stream is distributed in two streams: into the first stream (stream A) of packets s _A (r, j) of samples continuously following each other, samples of N samples

s _A (r, j) = s _in (rN + j), where

r = -∞, ..., 0,1,2, ..., + ∞ - packet number;

j = 0,1,2, ..., N-1 is the reference number in each packet, and in the second stream (stream B), which includes the samples of the input signal shifted relative to stream A by N / 2 samples

Then, in each package, the samples are multiplied by the values of the weight function W (j) for j = 0, ..., N-1, using these products u _{A, B} (r, j) 0 = s _{A, B} (r, j) W (j) for each packet, calculate the N-point DFTs, obtain the spectral components S _{A, B} (r, k), the squares of the modules | S _A (r, k) | ² for k = 0, ..., N-1, accumulate and average, for example, by the formula

As weight functions use, for example, one of the functions

Hamming [4, p. 92]

W (j) = 0.54-0.46cos (27πj / (N-1));

or Blackman [4, p. 187]:

W (j) = 0.42-0.5cos (2πj / (N-1)) + 0.08cos [4πj / (N-1)];

or Kaiser [5, p. 109]:

, где

where

I ₀ (x) is the modified zero-order Bessel function;

β is a parameter that determines the characteristics of the weight function.

и известный энергетический спектр сигнала G_с(ω), на основании выражения для оптимального по критерию максимума отношения сигнал - помехи на выходе подавителя помех [6, стр.134] передаточной функции подавителя помех K_опт(ω)=G_с(ω)/[G_с(ω)+G_п(ω)], где G_с(ω) - энергетический спектр сигнала, G_п(ω) -энергетический спектр помех, формируют текущую функцию подавления P_r(k) помех по формулеFurther, using the obtained averaged squares of the modules of the spectral components

and the well-known energy spectrum of the signal G _c (ω), based on the expression for the optimum signal-to-noise ratio according to the criterion maximum, at the output of the interference suppressor [6, p.134] the transfer function of the interference suppressor K _opt (ω) = G _s (ω) / [G _c (ω) + G _p (ω)], where G _c (ω) is the energy spectrum of the signal, G _p (ω) is the energy spectrum of interference, form the current interference suppression function P _r (k) according to the formula

, где

where

k = 0, ..., N-1, is the number of the spectral component;

ω _k = 2πF _d k / N - reference frequency values;

F _d - sampling rate.

Further, in each stream in each packet, the components of the suppression functions P _r (k) are multiplied with the spectral samples S _{A, B} (r, k) and N-point ODPFs from these products s _odfpA (r, j) and s _odpfV (r , j), which are divided by the weight function, receiving in each stream continuous streams of packets of samples of the sum of the useful signal and interference s _{output А} (r, j) = s _odpfV (r, j) / W (j) and s _output (r, j ) = s _odpfB (r, j) / W (j).

The output signal is formed in the form of continuously following one after another its samples, alternately choosing from each stream according to N / 2 samples according to the formula

, где

where

r = -∞, ..., 0,1,2, ..., + ∞ - packet number;

- номер отсчета в пакете.

- reference number in the packet.

In a specific implementation of the present invention, the values of the sampling frequency F _d , the value of N, the weight function W (j), the value of M are selected based on the hardware capabilities and available computing resources. Moreover, to form an undistorted spectrum of the input, useful signal, the requirement F _d ≥2ΔF, where ΔF is the spectrum width of the useful signal, must be met.

A second embodiment of a patented adaptive interference suppression method is an input signal processing procedure when the spectral position of a rapidly changing narrowband interference is known. In this case, the spectrum is completely suppressed in a given frequency range of the interference, for which the interference suppression function calculated according to the first embodiment is modified, equating it to zero in the given spectral range of fluctuating interference.

The third promising direction for the implementation of the patented method of adaptive interference cancellation is an option that allows combining the interference cancellation with increasing selectivity on the adjacent channel of the carrier and intermediate frequency amplification paths. In this case, the interference suppression function calculated according to the first embodiment is modified, equating it to zero outside the frequency range of the useful signal. At the same time, suppression of the interference spectrum remains effective from 5 to 90 dB.

Features of the invention

Samples of the discretized sum of the useful signal and interference s _in (n) are distributed in two streams of packets of samples of the sum of the useful signal and interference in N samples continuously continuously following each other: into stream A with samples s _A (r, j) = s _in (rN + j) and stream B with samples

shifted relative to stream A by N / 2 samples (r is the packet number, j is the sample number in the packet), in both streams for each packet the samples of the sum of the useful signal and interference are multiplied by the samples of the weight function W (j). For each packet of the obtained products, the spectrum of the sum of the useful signal and interference is calculated using the N-point DFT, for each component of the spectrum at number k, the components of the interference suppression function P _r (k) are calculated, as the quotient of the known squares of the modules of the components of the spectrum of the useful signal by the average square modules of the components of the calculated spectrum, multiply the spectrum by the interference suppression function, calculate the IFFT products and divide the result by the weight function. From the obtained samples, an output useful signal is generated with suppressed interference in the form of continuously following one after another samples, alternately choosing from each stream N / 2 samples according to the formula

Modify the calculated interference suppression function P _r (k), equal to zero in a given spectral range of fluctuating interference.

Modify the calculated interference suppression function P _r (k), equating it to zero outside the frequency range of the signal.

Implementation of the adaptive interference suppression method

An adaptive jammer implementing the patented method can be constructed, for example, on the basis of one programmable logic integrated circuit (FPGA) EP3SE50 from ALTERA.

To confirm the operability and effectiveness of the proposed method, computer simulation of adaptive interference suppression was carried out, the results of which are illustrated by graphic material.

List of graphic materials

Figure 1. A fragment of a useful signal without interference.

Figure 2. Spectra of useful signal and harmonic interference.

Figure 3. Harmonic suppression function.

Figure 4. A fragment of a signal with suppressed interference.

Figure 5. The dependence of the signal-to-noise ratio at the output on the signal-to-noise ratio at the input for harmonic interference.

6. Spectra of signal and broadband interference.

7. Broadband Suppression Function.

Fig. 8. Dependence of the signal-to-noise ratio at the output on the signal-to-noise ratio at the input for broadband interference.

In the simulation, a continuous sequence of phase-manipulated pulses with a phase rotation of 180 degrees was selected as the input signal (message). Phase manipulation is carried out by a random sequence of equiprobable characters. Such a message is typical, for example, for digital information transmission systems based on signals with binary phase shift keying. The energy spectrum of such a signal or the power spectral density is represented by the expression [7, p.176]

где

Where

T _c - the duration of the information symbol;

ω ₀ - carrier or intermediate frequency of the signal.

The results are shown for a message with the following parameters: amplitude of the input signal 1 mV; the center frequency F ₀ = 76.8 MHz the duration of the information symbol T _c = 78.125 ns. A fragment of the sequence of samples of the input signal without interference (points in figure 1) is presented in figure 1. For clarity, figure 1 and the following figures produced a linear interpolation of values between samples.

The effectiveness of the present invention was tested for parameters: sampling frequency F _D = 102.4 MHz, equal to twice the width of the processed signal spectrum ΔF = 51.2 MHz; the number of DFT and ODPF points N = 1024; Blackman's weighting function, the number of servings (N samples in each) of the accumulation of spectra N _nak = 36.

For comparison, a simulation of the method proposed in the prototype was carried out with the same input signals and with the same signal processing parameters.

Figure 2, figure 3, figure 4, figure 5 presents the results when exposed to harmonic interference with a frequency of F _p = 80.73 MHz.

Figure 2 presents the spectra of the signal (line 1) and harmonic interference (line 2), the amplitude of which significantly exceeds the amplitude of the useful signal.

Figure 3 shows the interference suppression function P (k), which shows that in the area where the interference spectrum is concentrated, the suppression is 60 dB.

Figure 4 shows a fragment of a signal processed according to the present invention with an interference amplitude U _p = 100 mV, i.e. when the amplitude of the interference is 100 times the amplitude of the signal. As can be seen, the output signal with the suppression of interference practically does not differ from the input signal without interference, shown in figure 1.

Figure 5 shows the dependence of the signal-to-noise power ratio at the output after suppressing H _o from the signal-to-noise ratio at the input H _in . The value of H _{o was} calculated by the formula

где

Where

s _с (n) - samples of the useful signal without interference;

s _vyx (n) - samples of the output, useful signal.

Dependencies are given for the present invention (line 1) and prototype (line 2). There is a noticeable gain in the noise immunity of the present invention with a relatively large signal-to-noise ratio (gain of 15 dB at H _in = 0, when the signal and noise amplitudes are U _c = U _p ), which increases with increasing noise amplitude, reaching a very impressive value at H _in ≤-50 dB, when the method proposed in the prototype is practically not operational (H _output ≤-30 dB).

The dependence shown in FIG. 5 also makes it easy to estimate the gain in noise immunity H in the form of a signal-to-noise ratio after suppression and a signal-to-noise ratio before suppression

H = H _o / H _in .

With H _in = -30 dB, the gain for the invention is H = 59 dB, and for the prototype - H = 27 dB.

In Fig.6, Fig.7, Fig.8 presents the results when exposed to broadband interference.

Figure 6 presents the spectra of the signal (line 1) and broadband interference (line 2). With such interference, the proposed method generates a noise suppression function P (k) to the form shown in Fig.7. The dependence of the signal-to-noise ratio of the output H _o on the input H _i for the present invention (line 1) and the prototype (line 2) are shown in Fig. 8. Here, as for narrowband interference, when using the present invention there is a very large gain in noise immunity from 5 to 90 dB.

Thus, with the help of modeling, it was proved that the patented method of adaptive interference suppression provides the declared technical result - effective suppression of both narrowband and wideband interference at the output of the receiving device, when the interference power with an unknown spectrum significantly exceeds the signal power.

Information sources

1. Interference immunity of radio systems with complex signals. / Tuzov G.I., Sivov V.A., Prytkov V.I. and etc.; Ed. G.I. Tuzova. - M .: Radio and communications, 1985. S.207-218.

2. Patent RU 2100903 C1, M. cl. Н04В 1/10, publ. 12/27/1997.

3. Patent RU 2269201, M. cl. Н04В 1/10, publ. 07/27/2005 (the closest analogue).

4. Lyons R. Digital signal processing. Second edition. Per. from English - M .: Bikom-Press LLC, 2006. P.92,187.

5. Rabiner L., Gould B. Theory and application of digital signal processing. Per. from English, Moscow: Mir, 1978, p. 109.

6. Kharkevich A.A. Fighting interference. M .: Nauka, 1965. S.130-138.

7. Prokis D. Digital communications. - M .: Radio and communications, 2000. S.173-189.

Claims (3)

1. The method of adaptive interference suppression, which consists in linearly amplifying the sum of the useful signal and interference, its discretization, forming the spectrum of the sum of the useful signal, and interference through the discrete Fourier transform (DFT) of the obtained samples of this sum of the useful signal, and interference and the inverse discrete Fourier transform ( ODPF), characterized in that the samples of the discretized sum of the useful signal and interference s _in (n) are distributed into two streams of packets of samples of the sum of the useful signal and interference, continuously following each other N samples: to stream A with samples s _A (r, j) = s _bx (rN + j) and stream B with samples

shifted relative to stream A by N / 2 samples, where r is the number of the packet, j is the number of the sample in the packet, in both flows for each packet the samples of the sum of the useful signal and interference are multiplied by the samples of the weight function W (j), for the packet of the obtained products by means of an N-point DFT, the spectrum of the sum of the useful signal and interference is calculated, for each component of the spectrum at number k, the components of the interference suppression function P _r (k) are calculated, as the quotient of dividing the known squares of the modules of the components of the spectrum of the useful signal by the average square of the modules components of the calculated spectrum, multiply the spectrum by the interference suppression function, calculate the IFFT products and divide the result by the weight function, form the output useful signal with the suppressed noise from the obtained samples in the form of continuously following one after another samples, alternately choosing N / 2 from each of the streams counts by the formula:

2. The method according to claim 1, characterized in that they modify the interference suppression function calculated according to claim 1, equating it to zero in a given spectral range of fluctuating interference. 3. The method according to claim 1, characterized in that they modify the interference suppression function calculated according to claim 1, equating it to zero outside the frequency range of the useful signal.

Images