RU2552534C1 - Method of processing hydroacoustic noise-like phase-shift keyed signals - Google Patents

Abstract

FIELD: radio engineering, communication.

SUBSTANCE: method of processing hydroacoustic noise-like phase-shift keyed signals includes receiving a signal s(t), digitising the signal, obtaining y_k, pre-equalising the amplitude $${\dot{y}}_k=sign[y_k],$$

where $$sign[x]=\{\begin{array}{l}+1приx\ge 0\\ -1приx<0\end{array},$$

performing shift into the low-frequency region and determining the real component

and the imaginary component of the signal

(f_s is the centre frequency of the noise-like phase-shift keyed signal being processed, f_d is the sampling frequency of the signal processing system, N_s is the processing window length, which must be equal to an integer number of periods in sampling frequency readings, i.e. N_s=n·T_s·f_d, where n=1, 2, 3…), for the obtained signal $${\tilde{y}}_j=A_j+i{B}_j$$

( $$i=\sqrt{-1}$$

- imaginary unit) using a low-pass filter, suppressing high-frequency components,

is the pulse response of the filter, N_f if the length of the pulse response of the filter), decimating the sampling frequency with an interval N_s of the signal

where N_s is the sampling interval, which is equal to the radio of the sampling frequency f_d of the source signal and the doubled cutoff frequency $$N_s=\frac{f_d}{2f_{cp}}=\frac{f_d}{\Delta f},$$

after which the sampling frequency of the signal becomes equal to f_d2=2f_cut=Δf, repeating signal amplitude equalisation $${\dot{y}}_j^s=sign[y_j^s]$$

and for the obtained signal $${\dot{y}}_j^{д}$$

, calculating the value of a correlation function $$Y_j={\Sigma }_{k=1}^{N_{cut}}{\dot{y}}_j^s\cdot m_k,$$

where N_cut is the duration of the processed signal in sampling frequency readings f_d2, m_k is the reference correlator signal in symbol form, calculating a threshold value $${{\rm Y}}_{thres}=\frac{n-2k}{n},$$

where n is the number of symbols in the modulating pseudorandom sequence, k is an integer defined by a given false response probability ρ_false, wherein k≤n and is selected as the greatest number for which the condition $${\rho }_{false}\approx {0.5}^k{\Sigma }_{j=k}^n{C}_n^i$$

is satisfied, where $$C_n^i$$

is the number of combinations i through $$n:C_n^i=\frac{n!}{i!(n-i)!}),$$

comparing the value of the correlation function Y_j with the threshold value Y_thres, and presence of a signal is determined when the value of the correlation function exceeds the threshold value.

EFFECT: improved noise-immunity when detecting a hydroacoustic signal in real operating conditions with low hardware computational power.

Description

The invention relates to the field of sonar, and in particular to methods for processing sonar signals in a real distribution channel, and can be used in sonar communication systems, control and positioning. The processing method can be used to detect a hydroacoustic noise-like phase-shifted signal of a known shape. It is also possible to use when organizing a multi-subscriber system with channel separation by encoding signals with various binary pseudorandom sequences from the same family.

A known method of classical linear correlation processing of noise-like signals in order to assess their parameters [1, 2]. The basis of this method is the calculation of the scalar product (u, ν) of two signals u (t), ν (t), which is also called correlation and indicates the similarity (similarity) of the signals:

where u (t) is the initial observed signal in the channel;

ν (t) is the reference signal identical to the received one.

Then the rule of maximum correlation is used:

where k is the signal number from the set of considered signals;

j is the number of the received signal;

H _j is the hypothesis that the signal number j is the one sought.

The rule of maximum correlation means, in particular, that of the M possible signals with the same energy, the one that has the maximum correlation with observation ν (t) is considered to be actually accepted. Preference is given to the one of the signals that is most similar to the observation of ν (t) in comparison with the others, provided that the correlation value is taken as a criterion of similarity.

The disadvantage of this method is the low noise immunity in conditions of non-white noise (since this method is not optimal for these conditions), as well as the fact that the method cannot be used in asynchronous systems when the moment and interval of signal emission are unknown.

A known method of demodulating signals with relative phase modulation described in the patent of the Russian Federation No. 2271071, 2006, IPC H06L 27/22. In this method, a signal S (t) is received, its amplitude is filtered and equalized, a reference signal S ₀ (t) is generated, a correlation function Y (t) is calculated. Then, the result of the product of signals S ₀ (t) and S _C (t) is filtered in the block for reducing the noise level due to a change in the polarity of the video signal at the output of the low-pass filter over time, i.e. time during which the correlation function Y (t) is formed on the duration of the signal element T. Then, the correlation function Y (t) is integrated sequentially over the duration T and its value Y _{n is} fixed at the end of the signal element. The absolute value of the difference | ΔY _n | between the current and previous values of the correlation functions Y _n and Y _n-1, respectively, at the nth and (n-1) -th time intervals T. The obtained value of the difference modulus is compared with a previously generated threshold Y _then according to the rule: if the inequality | ΔY _n |> Y _then , the decision on the demodulated symbol is taken equal to "one", and in case of inequality, they are taken equal to "zero".

The disadvantage of this method is the relatively low noise immunity, due to the fact that the decision on the demodulated symbol is made by comparing with the constant _pore threshold Y _then preformed in the absence of interference, which does not take into account phase changes of the demodulated signal as a result of interference.

порогового значения корреляционной функции от предварительно заданного его значения Y_пop и рассчитывают значение $$\Delta Y_{пор}^{кор}$$

путем алгебраического сложения предварительно заданного порогового значения корреляционной функции Y_пop и вычисленного ее отклонения $$\Delta Y_{пор}^{п}$$

на n-м временном интервале Т, $$\Delta Y_{пор}^{кор}=Y_{пор}+\Delta Y_{пор}^{п}$$

, после чего все действия по демодуляции сигнала S(t) на последующем (n+1)-м временном интервале Т повторяют с учетом откорректированного значения $$\Delta Y_{пор}^{кор}$$

. А отклонение $$\Delta Y_{пор}^{п}$$

порогового значения корреляционной функции вычисляют по формуле:

There is also a method of demodulating signals with relative phase modulation, described in RF patent No. 2454014, 2010, IPC H04L 27/00, H04L 13/18, the first option. In the known method of demodulating signals with relative phase modulation, a signal S (t) is received, filtered and its amplitude is equalized, a reference signal S is generated₀(t), calculate the correlation function Y (t) between the reference signal S₀(t), and a filtered signal with equalized amplitude S_c(t) by multiplying them Y (t) = S_c(t)₀(t), integrate the correlation function Y (t) sequentially at time intervals of duration T and fix its value Y_n at the end of the n-th time interval T, where n = 1, 2, ..., the modulus of difference | ΔY_n| values of correlation functions Y_n and Y_n-1 respectively, at the nth and (n-1) -th time intervals T, the obtained value of the difference modulus | ΔY_n| compared with a predetermined threshold value Y_popcorrelation function and when the condition | ΔY_n|> Y_pop assign the value of "one" to the received information element, otherwise - "zero". A random L-element sequence with an equal number of unit and zero elements in it is preformed, where L is an integer, and then this sequence is changed, for which the demodulated information element received at the nth time interval T is written as the first element in the L-element sequence shifting all its elements by one bit while maintaining its total length L, adjust the threshold value of the correlation function Y_sincewhy calculate the number of "units" in the changed L-element sequence, calculate the deviation $$\Delta Y_{P\mathrm{about}R}^P$$

 threshold value of the correlation function from its predetermined value Y_pop and calculate the value $$\Delta Y_{P\mathrm{about}R}^{\mathrm{to}\mathrm{about}R}$$

 by algebraic addition of a predetermined threshold value of the correlation function Y_pop and its calculated deviation $$\Delta Y_{P\mathrm{about}R}^P$$

 on the n-th time interval T, $$\Delta Y_{P\mathrm{about}R}^{\mathrm{to}\mathrm{about}R}=Y_{P\mathrm{about}R}+\Delta Y_{P\mathrm{about}R}^P$$

after which all the actions for demodulating the signal S (t) on the subsequent (n + 1) -th time interval T are repeated taking into account the adjusted value $$\Delta Y_{P\mathrm{about}R}^{\mathrm{to}\mathrm{about}R}$$

. A deviation $$\Delta Y_{P\mathrm{about}R}^P$$

 the threshold value of the correlation function is calculated by the formula:

where k (l) is the number of "units" in the L-element sequence. This method is closest to the claimed and is hereinafter referred to as the prototype method.

The disadvantages of the prototype method are also relatively low noise immunity (reception is carried out at a signal-to-noise ratio of more than 6 dB) and redundancy of computational operations.

The problem to which the invention is directed is to increase the noise immunity when solving the problem of detecting a hydroacoustic signal in real operating conditions (signal power is much less than the level of hydroacoustic noise) with low computing power of hardware.

где $$sign[x]=\{\begin{array}{l}+1приx\ge 0\\ -1приx<0\end{array},$$

выполняют смещение в область низких частот и определяют реальную составляющую

и мнимую составляющую сигнала

(f_s - средняя частота обрабатываемого шумоподобного фазоманипулированного сигнала, f_d - частота дискретизации системы обработки сигнала, N_s - длина окна обработки, должна равняться целому числу периодов в отсчетах частоты дискретизации, т.е. N_s=n·T_s·f_d, где n=1, 2, 3…), для полученного сигнала $${\tilde{y}}_j=A_j+i{B}_j$$

$$(i=\sqrt{-1}$$

- мнимая единица) фильтром нижних частот подавляют высокочастотные составляющие,

- импульсная характеристика фильтра, N_ф - длина импульсной характеристики фильтра), проводят операцию децимации частоты дискретизации с шагом N_д сигнала

где N_д - шаг дискретизации, равный отношению частоты дискретизации f_d исходного сигнала и удвоенной частоты среза $$N_{д}=\frac{f_d}{2f_{cp}}=\frac{f_d}{\Delta f},$$

после чего частота дискретизации сигнала становится равна f_d2=2f_cp=Δf , вторично выполняют выравнивание амплитуд сигнала $${\dot{y}}_j^{д}=sign[y_j^{д}]$$

и для полученного сигнала $${\dot{y}}_l^{д}$$

вычисляют значение корреляционной функции $$Y_j={\displaystyle {\sum }_{k=1}^{N_{cp}}{\dot{y}}_j^{д}\cdot m_k},$$

где N_cp - длительность обрабатываемого сигнала в отсчетах частоты дискретизации f_d2, m_k - опорный сигнал коррелятора в знаковой форме, вычисляют пороговое значение $$Y_{пор}=\frac{n-2k}{n}$$

, где n - количество знаков в модулирующей псевдослучайной последовательности, k - это целое число, определяемое заданной вероятностью ложных срабатываний ρ_лож (при этом k<n и выбирают как наибольшее число, при котором выполняется условие $${\rho }_{лож}\approx {0.5}^k{\Sigma }_{i=k}^n{C}_n^i$$

, где $$C_n^i$$

- число сочетаний i по $$n:C_n^i=\frac{n!}{i!(n-i)!})$$

, с пороговым значением Y_пор сравнивают значение корреляционной функции Y_j>Y_пop, определяют наличие сигнала в случае, если значение корреляционной функции превышает пороговое значение.The technical result is achieved by receiving a signal s (t) with an average frequency f _s and a frequency band Δf, filtering and equalizing its amplitude, generating a reference signal s ₀ (t), and calculating the correlation function Y (t) between the reference signal s ₀ ( t) and a filtered signal with aligned amplitude s _c (t) by multiplying them Y (t) = s _c (t) s ₀ (t), integrate the correlation function Y (t), according to the invention, the signal s (t) is received, digitized the signal is obtained at _k , the amplitudes are pre-aligned $${\dot{y}}_k=sign[y_k],$$

Where $$sign[x]=\{\begin{array}{l}+\mathrm{one}PR\mathrm{and}x\ge 0\\ -\mathrm{one}PR\mathrm{and}x<0\end{array},$$

perform an offset to the low-frequency region and determine the real component

and the imaginary component of the signal

(f _s is the average frequency of the processed noise-like phase-shifted signal, f _d is the sampling frequency of the signal processing system, N _s is the length of the processing window, should be an integer number of periods in the samples of the sampling frequency, i.e., N _s = n · T _s · f _d , where n = 1, 2, 3 ...), for the received signal $${\tilde{y}}_j=A_j+i{B}_j$$

$$(i=\sqrt{-\mathrm{one}}$$

- imaginary unit) high-frequency components suppress the low-pass filter,

- impulse response of the filter, N _f - length of the impulse response of the filter), carry out the operation of decimation of the sampling frequency in increments of N _d signal

where N _d is the sampling step equal to the ratio of the sampling frequency f _{d of the} original signal and the double cutoff frequency $$N_d=\frac{f_d}{2f_{cp}}=\frac{f_d}{\Delta f},$$

after which the sampling frequency of the signal becomes equal to f _d2 = 2f _cp = Δf, the signal amplitudes are equalized again $${\dot{y}}_j^d=sign[y_j^d]$$

and for the received signal $${\dot{y}}_l^d$$

calculate the value of the correlation function $$Y_j={\displaystyle {\sum }_{k=\mathrm{one}}^{N_{cp}}{\dot{y}}_j^d\cdot m_k},$$

where N _cp is the duration of the processed signal in samples of the sampling frequency f _d2 , m _k is the reference signal of the correlator in sign form, the threshold value is calculated $$Y_{P\mathrm{about}R}=\frac{n-2k}{n}$$

, where n is the number of characters in the modulating pseudorandom sequence, k is an integer determined by the given probability of false positives ρ _boxes (in this case, k <n and is chosen as the largest number under which the condition $${\rho }_{l\mathrm{about}\mathrm{well}}\approx {0.5}^k{\Sigma }_{i=k}^n{C}_n^i$$

where $$C_n^i$$

is the number of combinations i by $$n:C_n^i=\frac{n!}{i!(n-i)!})$$

, with the threshold value of Y _then compare the value of the correlation function Y _j > Y _pop , determine the presence of a signal if the value of the correlation function exceeds the threshold value.

The inventive method of processing hydroacoustic noise-like phase-shifted signals includes:

- alignment of the amplitudes of the signal, the result of which is a discrete signal that takes one of two values (1 and -1);

- signal shift to low frequencies;

- filtering high-frequency components, starting with a cutoff frequency f _cf , which is defined as half the signal bandwidth;

- decimation of the sampling frequency of the signal in an integrated way to a double value of the cutoff frequency f _cp to reduce the number of computational operations for processing the signal;

- secondary alignment of the amplitudes of the signal, the result of which is a discrete signal that takes one of two values (1 and -1);

- Sign correlation processing of the received signal with the reference signal;

- comparison of the obtained correlation values with a given threshold.

This method is intended for processing a hydroacoustic noise-like phase-shifted signal modulated by the direct sequence method in order to detect a signal of a given shape and evaluate its time delay (time of arrival).

When implementing the method, equalization of the amplitudes of the received signal is presented, presented in a discrete form, which implements a sign function:

$${\dot{y}}_k=sign[y_k],$$

where _k are discrete samples of the received signal,

sign [] - sign function:

The operation of equalizing the amplitudes is performed in order to reduce the bit depth of digital blocks performing filtering, shifting the frequency band of the signal and decimating the sampling frequency. In the prototype method, amplitude equalization before filtering is absent and is performed only once after the filter.

Then the received signal is shifted to the low frequency region and is represented as a quadrature sum:

, $${\tilde{y}}_j=A_j+i{B}_j$$

,

where A _j is the real component of the signal,

B _j is the quadrature component of the signal,

- мнимая единица. $$i=\sqrt{-\mathrm{one}}$$

- imaginary unit.

The real and imaginary components of the signal are calculated as follows:

where f _s is the average frequency of the processed noise-like phase-shifted signal,

f _d is the sampling frequency of the signal processing system,

N _s is the length of the processing window, must be an integer number of periods in the samples of the sampling frequency, i.e. N _s = n · T _s · f _d , where n = 1, 2, 3 ...

A shift to the low-frequency region allows you to lower the sample rate of the signal most effectively. In the prototype method, the offset to the low frequency region is not used.

смещен в область низких частот и проходит через фильтр низких частот:Received signal $${\tilde{y}}_j$$

shifted to the low-frequency region and passes through the low-pass filter:

where h _j is the impulse response of the low-pass filter,

N _f - the length of the impulse response of the filter.

The cutoff frequency of the filter is selected equal to half the frequency band of the signal.

Filtering improves noise immunity by cutting off high-frequency noise.

Next, the doubled cutoff frequency 2f _cp (signal bandwidth) is selected as the system sampling frequency after decimating the frequency in an integrated way:

where N _d is the sampling step equal to the ratio of the sampling frequency f _{d of the} original signal and the double cutoff frequency (N _d = f _d / 2f _cp ).

Decimation of the sampling frequency also distinguishes the claimed method for processing a hydroacoustic complex phase-shifted signal from the prototype method. The decimation operation can significantly reduce the number of computational operations for calculating the correlation, provided that the frequency band of the received signal is much less than the upper frequency of the signal.

The amplification equalization operation is performed again on the received signal:

The operation of equalization of amplitudes is performed in order to reduce the bit depth of digital blocks performing correlation processing. In addition, the sign correlation processing is more noise-resistant in conditions of non-white noise, which also includes hydroacoustic noise.

Next, a low-frequency signal with a reduced sampling frequency undergoes sign correlation processing in order to extract a useful signal:

where N _cp is the duration of the processed signal in samples of the sampling frequency f _cp,

m _k is the correlator reference signal in sign form.

, где n - количество знаков в модулирующей псевдослучайной последовательности, k - это целое число, определяемое заданной вероятностью ложных срабатываний ρ_лож (при этом k≤n и выбирают как наибольшее число, при котором выполняется условие $${\rho }_{лож}\approx {0.5}^k{\Sigma }_{i=k}^n{C}_n^i$$

, где $$C_n^i$$

- число сочетаний i по $$n:C_n^i=\frac{n!}{i!(n-k)!}),$$

с пороговым значением Y_пор сравнивают значение корреляционной функции Y_j>Y_пop, определяют наличие сигнала в случае, если значение корреляционной функции превышает пороговое значение.Next, at a given threshold, _Ypop , the presence of a signal in the channel is determined. The threshold value is calculated. $$Y_{P\mathrm{about}R}=\frac{n-2k}{n}$$

, where n is the number of characters in the modulating pseudorandom sequence, k is an integer determined by the given probability of false positives ρ _boxes (in this case, k≤n is chosen as the largest number under which the condition $${\rho }_{l\mathrm{about}\mathrm{well}}\approx {0.5}^k{\Sigma }_{i=k}^n{C}_n^i$$

where $$C_n^i$$

is the number of combinations i by $$n:C_n^i=\frac{n!}{i!(n-k)!}),$$

with the threshold value of Y _then compare the value of the correlation function Y _j > Y _pop , determine the presence of a signal if the value of the correlation function exceeds the threshold value.

The described cycle of calculations is performed every moment of time j, with the sampling rate of the system, which allows the signal to be detected in real time.

Distinctive features of the prototype method of the features of the proposed method are: 1) preliminary alignment of the signal amplitudes before the filtering operation; 2) the shift of the signal to the low frequency region; 3) decimation of the sampling frequency of the signal in an integrated way to reduce the number of computational operations for correlation signal processing; 4) a method for calculating a threshold value. The presence of distinctive features from the prototype features allows us to conclude that the proposed method meets the criterion of ″ novelty ″.

A review of the known inventions showed that the claimed method has a new property that allows to minimize the value of the probability of false alarm, to effectively deal with impulse noise and thereby increase the noise immunity of the processing of hydroacoustic signals by performing operations in the proposed sequence. This circumstance allows us to conclude that the developed method meets the criterion ″ significant differences ″.

An example implementation of the method.

Let there be an emitter of a hydroacoustic noise-like phase-shifted signal modulated by the direct sequence method. The average signal frequency f _s = 41666 Hz, the signal bandwidth, determined by the level of 0.1, is Δf = 10 kHz, the lower edge of the spectrum f _n = 36166 Hz, the upper edge of the spectrum f _B = 46166 Hz, the length of one signal symbol is specified by the number periods of the average signal frequency N _p = 8, signal duration τ = 24.3 ms, a modulating code from the family of M-sequences of length 127. These conditions are relevant for short-base sonar systems operating in shallow sea conditions.

At the input of the receiving device, a threshold detector operating at a frequency f _d = 192 kHz (the frequency of modern ADCs) that receives a signal. The binary signal, which is a mixture of the useful signal and the noise component, is shifted from the detector output to the low frequency region (f _cf = 0 Hz, f _H = -5 kHz, f _B = 5 Hz), while the frequency band remains the same. The received signal passes through a low-pass filter with a cutoff frequency f _av = 5 kHz, which suppresses the high-frequency components of the signal (5 (f) → 0 at f> 5 kHz). Then, the sampling rate of the original signal is reduced by an integer number of times N _d , which is defined as the value of the partial f _d / 2f _cp rounded to a smaller integer, that is, in this case 192000/10000 = 19.2 and after rounding N _d = 19. That is, after the decimation operation, the sampling frequency is 10.1 kHz, and the signal duration in the samples of the sampling frequency is 242. After that, correlation processing occurs and each correlation value is compared with a predetermined threshold h = 0.45. This threshold value was determined using methods of mathematical statistics and ensures the correct operation of the receiving device with a probability of at least 95% for SNR ≥ -18 dB in white noise or SNR≥-7 dB in conditions of non-white noise of a real sonar channel.

Information sources

1. Valery P. Ipatov, Spread Spectrum and CDMA. Principles and Applications / John Willy & Sons Ltd, 2005 - 398 p.

2. Linnik M.A., Karabanov I.V., Burdinskiy I.N. Threshold Methods of Sonar Pseudonoise Phase-shift Signal Detection / The First Russia and Pacific Conference on Computer Technology and Applications (Russia Pacific Computer 2010) 6-9 September, 2010 Russian Academy of Sciences, Far Eastern Branch. - Vladivostok, 2010 .-- S.404-408.

Claims (1)

A method for processing noise-like phase-shifted signals, based on the fact that a signal s (t) is received with an average frequency f _s and a frequency band Δf, filter and equalize its amplitude, a reference signal s ₀ (t) is generated, a correlation function Y (t) between the reference signal s ₀ (t) and a filtered signal with aligned amplitude s _c (t) by multiplying them Y (t) = s _c (t) s ₀ (t) integrate the correlation function Y (t), characterized in that they take signal s (t), digitize the signal, get at _k , equalize the amplitudes $${\dot{y}}_k=sign[y_k],$$

Where $$sign[x]={\{}_{-\mathrm{one}PR\mathrm{and}x<0}^{+\mathrm{one}PR\mathrm{and}x\ge 0}$$

, perform an offset to the low frequency region and determine the real component

and the imaginary component of the signal

(f _s is the average frequency of the processed noise-like phase-shifted signal, f _d is the sampling frequency of the signal processing system, N _s is the length of the processing window, should be an integer number of periods in the samples of the sampling frequency, i.e., N _s = n · T _s · f _d , where n = 1, 2, 3 ...), for the received signal $${\tilde{y}}_j=A_j+i{B}_j$$

( $$i=\sqrt{-\mathrm{one}}$$

- imaginary unit) high-frequency components suppress the low-pass filter,

^_ impulse response of the filter, N _f - length of the impulse response of the filter), carry out the operation of decimation of the sampling frequency in increments of N _d signal

where N _d ^{_ the} sampling step equal to the ratio of the sampling frequency f _{d the} original signal and the double cutoff frequency $$N_d=\frac{f_d}{2f_{cp}}=\frac{f_d}{\Delta f},$$

after which the sampling frequency of the signal becomes equal to f _d2 = 2f _cf = Δf, the signal amplitudes are again aligned $${\dot{y}}_j^d=sign[y_j^d]$$

and for the received signal $${\dot{y}}_j^d$$

calculate the value of the correlation function $$Y_j={\Sigma }_{k=\mathrm{one}}^{N_{cp}}{\dot{y}}_j^d\cdot m_k,$$

where N _cp ^_ processed signal duration in samples sampling frequency f _d2, m _k - reference signal correlator in symbolic form, calculated threshold value $${{\rm Y}}_{P\mathrm{about}R}=\frac{n-2k}{n}$$

, where n is the number of characters in the modulating pseudorandom sequence, k is an integer determined by the given probability of false positives ρ _boxes (in this case, k <n and is chosen as the largest number under which the condition $${\rho }_{l\mathrm{about}\mathrm{well}}\approx {0.5}^k{\Sigma }_{i=k}^n{C}_n^i$$

where $$C_n^i$$

is the number of combinations i by $$n:C_n^i=\frac{n!}{i!(n-i)!}),$$

with the threshold value of Y _then compare the value of the correlation function Y _j > Y _then , determine the presence of a signal if the value of the correlation function exceeds the threshold value.