RU2103821C1 - Method of noncoherent reception of discrete signals - Google Patents

Abstract

FIELD: radio engineering and communication, multichannel reception of discrete signals under conditions of high level of noise. SUBSTANCE: method includes comparison of input processes in two reception channels, calculation of step function, formation of transfer function and transfer function of addition and their multiplication correspondingly with input processes in first and second reception channels. With this in mind input processes in reception channels are smoothed over within specified time interval corresponding to interval of noise correlation. Pulse functions that are integrated in the course of preset time interval corresponding to interval of signal correlation are formed on their basis in each reception channel. Then amplitudes of pulse functions of first and second reception channels are compared and after this step function is calculated. From it transfer function and additional transfer function are formed, they are multiplied with input processes in reception channels having correspondingly bigger or smaller values of amplitude of pulse function. Processes obtained after multiplication are sent into path of their further processing. It results in suppression of initial random process containing noise component and in extraction of random process carrying signal. EFFECT: diminished probability of error of recognition of random process carrying signal against noise background. 4 dwg, 1 tbl

Description

 The invention relates to radio engineering and communications, in particular to the technique of multichannel reception, and can be used in multichannel receiving devices for distinguishing discrete signals in a high noise level.

 Known methods for receiving incoherent reception of discrete signals [1, 2, 3], including comparing the signals in the receiving channels (KP) and the subsequent suppression of a strong signal weak.

 However, the known reception methods are characterized by a higher probability of error, in cases of small ratios of the signal power to the sum of the interference and noise powers (SINR). The high probability of error is determined by fluctuations of a random process (SP) containing the signal over a time interval exceeding the noise correlation interval, which leads to loss of information [2].

 Closest to the technical nature of the proposed is the method of incoherent reception of discrete signals described in [1].

 The prototype method includes: comparing the powers of the input SP in two reception channels, the formation of a single step function, the formation of a monotonously increasing transfer function and the transfer function from it to complement the single step function, and the multiplication of a monotonously increasing transfer function with a higher input SP, and the transfer function additions with less power, respectively.

 However, the prototype method has the following disadvantages: there is a greater probability of error due to the fact that when implementing the known method, the fluctuations of the SP are smoothed over a time interval corresponding to the noise correlation interval, that is, only the noise component of the input processes is smoothed out and the fluctuations of the SP are not taken into account in the time interval exceeding the noise correlation interval. At the same time, fluctuations of the SP in a time interval exceeding the noise correlation interval can reach 3–12 dB [4]. This in turn leads to loss of information. The use of the prototype method for processing random processes in a time interval determined by relatively slow signal fluctuations leads to an increase in time spent on the analysis of the SP in the receiving channels. For example, in the HF band, such a time interval can be minutes, and in some cases tens of minutes [4].

 The aim of the invention is to develop a method of incoherent reception of discrete signals, which reduces the likelihood of errors in distinguishing between a random process containing a signal against a background of noise.

The goal set in the method is achieved by the fact that in the known method of incoherent reception of discrete signals, which includes comparing the input processes in two reception channels, calculating a single step function, forming the transfer function and the transfer function of the complement and multiplying them respectively with higher and lower power input processes, pre-input processes in the receiving channels smooth out in a given time interval τ _W corresponding to the noise correlation interval. After that, on the basis of smoothed input SPs, pulse functions are formed in each receiving channel. The obtained impulse functions of the receiving channels are integrated with a time constant corresponding to the signal correlation interval τ _c . Then, the amplitudes of the pulse functions of the first and second reception channels are compared. Based on the results of the comparison, a single step function is calculated from which the transfer function and the transfer function of the complement are formed. The generated transfer function and the transfer function of the addition to the unit step function are multiplied with input random processes having correspondingly higher and lower values of the pulse function amplitude, and the processes obtained after multiplication are transferred to the path for their further processing for subsequent use.

 The indicated new set of essential features makes it possible to take into account fluctuations in the power of the SP in a time interval exceeding the noise correlation interval, which reduces the likelihood of improper selection of the process containing the signal component from two SPs, and thereby reduce the probability of signal discrimination against noise. This in turn increases the reliability of the received information.

In FIG. 1 shows timing diagrams explaining the operation of the proposed method. In FIG. 1a shows the initial random X ₁ (t), X ₂ (t) processes in the first and second reception channels, respectively (to simplify the recording below, X ₁ and X ₂ ), 1b _{shows the} processes P _cr1 , P _cr2 in the first and second CP accordingly, smoothed over the noise correlation interval, 1c - the generated impulse functions P _if1 , P _if2 for each of the receiving channels, 1d - impulse functions P _and1 , P _{and 2} integrated on the signal correlation interval, 1d, 1e - unit step function and transfer function with the predecessor function of the complement P _pf , P _pfd, respectively.

In FIG. 2 shows time charts showing the processing of the joint venture in the prototype method and the proposed method, where x $\genfrac{}{}{0ex}{}{\text{(1)}}{\text{1}}$ , x $\genfrac{}{}{0ex}{}{\text{(1)}}{\text{2}}$ , x $\genfrac{}{}{0ex}{}{\text{(2)}}{\text{1}}$ , x $\genfrac{}{}{0ex}{}{\text{(2)}}{\text{2}}$ - the implementation of the converted processes of the prototype method and the proposed method in the first and second reception channels, respectively.

 In FIG. 3 shows a structural diagram of a device that implements the proposed method.

 In FIG. 4 shows the dependence of the probability of error in determining a random process containing a signal component from the ratio of signal power to noise power, explaining the effectiveness of the proposed method.

 The possibility of implementing the proposed method of incoherent reception of discrete signals is explained as follows.

 It is known that joint ventures can be determined by statistical moment functions of order k. In the case of representing SPs as Gaussian processes, they can be described by statistical moments of the second order (correlation function and dispersion). In addition, it is known from the theory of optimal reception that the possibility of separating several random processes is limited by their dispersion of estimates. Therefore, to separate the two SPs, it is necessary to reduce their dispersion. The dispersion of the SP determined by the nature of the propagation of radio waves in the radio channels can be reduced both by smoothing and by suppressing a weak signal by a strong one. In this case, smoothing is carried out, as a rule, on a time interval corresponding to the noise correlation interval. The correlation interval is understood as the length of time in which random processes are considered correlated. However, in real radio channels, signal fluctuations are also observed in a time interval exceeding the noise correlation interval. This, in the case of small SINRs, leads to improper selection of the SP containing the signal, and therefore leads to loss of information.

 As follows from the analysis of statistical data [4, 5], the law of variation of the median value of the SP containing noise will be smoother compared to the SP containing the signal.

 Based on this, in the proposed method, after smoothing the input processes (to eliminate the noise component) in each receiving channel to form a pulse function (Fig. 1d.). For this, the smoothed random processes of each receiving channel are differentiated. As a result of differentiation of smoothed random processes in each receiving channel, we obtain a sequence of pulses. The number of pulses will be determined by the law of variation of the envelope of the SP in each receiving channel. For SP containing the signal and noise component, the number of pulses will be different due to the smoother nature of the change in the envelope of the SP with noise. This allows us to distinguish between the SP containing the signal, since the value of the pulse function in each channel will be determined by the number of pulses and, accordingly, the nature of the SP in each receiving channel (Fig. 1c).

 By integrating the obtained pulse functions (IF) of each receiving channel with a time constant corresponding to the noise correlation interval, it is possible to determine the channel in which the SP contains a signal. To do this, after integration, the values (in the particular case of the amplitude) of the IF of each receive channel are compared and, according to the results of this comparison, a unit step function (ESF) is formed, which takes values from 0.5 to -0.5 [1]. The ESF is formed by multiplying the unit function (obtained according to [6]) and the difference in the amplitudes of the impulse functions. The formation of a single step function from the difference in the IF values can be performed, for example, by an amplifier-limiter. Comparison of impulse functions can be performed by subtracting their amplitudes (for example, in comparing devices (comparators)). A single step function contains information about the difference between the values of the noise and signal components of the input random processes, as well as about the reception channel in which the signal is present.

 On the basis of the ESF, a transfer function and a supplement function are formed, which are multiplied with a larger and lower power SP, respectively (the pulse functions of which have a larger and lower value, respectively (Fig. 1d-1e)). The formation of the transfer function and the transfer function of the complement is carried out by determining the polarity of a single step function and normalizing its level to the value specified by the parameters of the multipliers. The standardization level is set based on a priori known average signal levels in the reception channels. The polarity of a single step function establishes a correspondence between the numbers of the reception channels and the transfer functions (transfer function and transfer function of addition). The polarity of a single step function can be determined by a diode rectifier. In turn, the average signal level depends on the technical characteristics of the radio line and can be determined either by practical measurement or by calculation. The transfer function and the transfer function of the complement are multiplied respectively with receive channels having a greater or lesser value of the amplitude of the impulse function.

 As a result of multiplication, the input SP is suppressed, which has a smoother character of the envelope change (i.e., the SP containing noise).

 Thus, as a result of the formation of impulse functions and their integration, the variance of random processes decreases. A comparison of the integrated IFs with the subsequent formation of the ESF, as well as the transfer function and the complement function, allows us to suppress the input process containing the noise component based on the differences in the initial moments of random processes, while simultaneously isolating the process with a signal.

 The indicated sequence of actions of the proposed method can be implemented by a multi-channel receiving device, one of the variants of which is shown in FIG. 3.

 In FIG. 3 positions denote: 1 - low-pass filter (LPF), 2 - differentiation device, 3 - amplifier-limiter (UO), 4 - integration devices with an integration time constant corresponding to the signal correlation interval, 5 - devices for generating a single step function, 6 - devices for forming the transfer function and the transfer function of the complement, 7 - multipliers.

 Devices 1 smooth out the input processes in each of the reception channels at a time interval corresponding to the noise correlation interval. As a result, the noise component of the input random processes is smoothed out. Devices 2, 3 form pulse functions in each receiving channel, respectively. In this case, the formation of impulse functions is carried out by differentiating the smoothed input random processes of each channel, and the value of the impulse functions at the output of devices 3 in each channel will be determined by the nature of the change in the envelope of the SP. Device 4 integrates impulse functions with a time constant exceeding the noise correlation interval. In the device 5, a unit step function is formed by converting the signal from the output of the integrating device into a pulse with an amplitude corresponding to the value of the channel pulse function with a faster character of the envelope change. In the device 6, which implements the amplifier-limiter, one of the outputs is inverse, the transfer function and the transfer function of the addition are formed with the values necessary for multiplying with random input processes in each of the receiving channels. Multiplication is carried out in devices 7.

где k₁(t-Δt) = (x $\genfrac{}{}{0ex}{}{\text{(1)}}{\text{2}}$ (t-Δt))/(x₂(t-Δt)) - значение коэффициента передачи первого процесса в момент времени (t-Δt), Δt → 0; z₁[t,g((t), k₁(t-Δt)] - передаточная функция первого процесса, существующая при x₁(t) > x₂(t) и монотонно возрастающая в пределах
0 ≤ z₁[t,g(t), k₁(t-Δt)] ≤ 1. (2)
Выражения (1) и (2) описывают динамику подавления одного из СП. Аналогичные функции реализуются выражениями (3) и (4), но в случае обратного хода данных процессов

где k₂(t-Δt) = (x $\genfrac{}{}{0ex}{}{\text{(1)}}{\text{2}}$ (t-Δt))/(x₂(t-Δt)) - значение коэффициента передачи второго процесса в момент времени (t-Δt), Δt → 0; z₂[t,g(t), k₂(t-Δt)] - передаточная функция второго процесса, существующая при x₁(t) > x₁(t) и монотонно возрастающая в пределах
0 ≤ z₂[t,g(t), k₂(t-Δt)] ≤ 1. (4)
Моделировались процессы, представляющие собой огибающие на выходах разделительных фильтров частотного дискриминатора. Предполагалось, что распределение одного из этих процессов (содержащего сигнал) подчинено закону Райса
W₁(x)(x/σ²)exp(-(x²+A²)/(2σ²)I_o(A_x/σ)), (5)
а другого (не содержащего сигнал) - закону Релея
W₂(x)(x/σ²)exp(-(x²)/(2σ²), (6)
где А - амплитуда сигнала, σ - эффективное значение шума на входе дискриминатора.In order to confirm the effectiveness of the proposed method, we used the modeling of random processes on a computer and their processing in accordance with the algorithms

where k ₁ (t-Δt) = (x $\genfrac{}{}{0ex}{}{\text{(1)}}{\text{2}}$ (t-Δt)) / (x ₂ (t-Δt)) - the value of the transmission coefficient of the first process at time (t-Δt), Δt → 0; z ₁ [t, g ((t), k ₁ (t-Δt)] is the transfer function of the first process, which exists for x ₁ (t)> x ₂ (t) and monotonically increases within
0 ≤ z ₁ [t, g (t), k ₁ (t-Δt)] ≤ 1. (2)
Expressions (1) and (2) describe the dynamics of suppression of one of the SPs. Similar functions are realized by expressions (3) and (4), but in the case of a reverse course of these processes

where k ₂ (t-Δt) = (x $\genfrac{}{}{0ex}{}{\text{(1)}}{\text{2}}$ (t-Δt)) / (x ₂ (t-Δt)) is the value of the transmission coefficient of the second process at time (t-Δt), Δt → 0; z ₂ [t, g (t), k ₂ (t-Δt)] is the transfer function of the second process, existing for x ₁ (t)> x ₁ (t) and monotonically increasing within
0 ≤ z ₂ [t, g (t), k ₂ (t-Δt)] ≤ 1. (4)
The processes were simulated, which are envelopes at the outputs of the separation filters of the frequency discriminator. It was assumed that the distribution of one of these processes (containing the signal) is subject to Rice's law
W ₁ (x) (x / σ ² ) exp (- (x ² + A ² ) / (2σ ² ) I _o (A _x / σ)), (5)
and another (not containing a signal) - to Rayleigh's law
W ₂ (x) (x / σ ² ) exp (- (x ² ) / (2σ ² ), (6)
where A is the signal amplitude, σ is the effective value of the noise at the discriminator input.

The indicated processes are represented by numerical values taken through 0.3 τ _k . The volume of statistics for each process amounted to 15 thousand numbers.

 The criterion for the effectiveness of the proposed method is the magnitude of the probability of error in determining the process with a signal from the taken samples. In accordance with this criterion, the error probabilities were calculated when processing the same processes without mutually suppressing the processes, suppressing a weak signal by a strong one (according to the method of [1]) and after smoothing the signal on the signal correlation interval according to the proposed method. The results are shown in the table.

The following notation is used in the table: a = A / σ — ratio of the signal amplitude to the effective value of noise σ at the output; and ^2/2 is the ratio of signal power to input noise power; R _osh , P $\genfrac{}{}{0ex}{}{\text{(1)}}{\text{osh}}$ , P $\genfrac{}{}{0ex}{}{\text{(2)}}{\text{osh}}$ - the probability of erroneous determination of the SP with the signal without suppressing one of the processes, after suppressing one of the random processes (according to the method [1]) and after smoothing the process on the signal correlation interval using the inventive method, respectively.

A graphical representation of the dependence of _Osh (P $\genfrac{}{}{0ex}{}{\text{(1)}}{\text{osh}}$ , P $\genfrac{}{}{0ex}{}{\text{(2)}}{\text{osh}}$ ) from the values of the ratio of the amplitude of the signal to the effective value of the noise at the output is shown in FIG. 4.

Literature
1. A. S. on the application 7771894, class. H 04 B 7/00 dated 06/30/78. The method of incoherent reception of discrete signals. A.I. Danilenko, A.Yu. Levashev.

 2. Korzhik V. I. Calculation of noise immunity of discrete message transmission systems. Directory. -M .: Radio and communications, 1981, pp. 106-128.

 3. Kalinin I.I., Cherenkova E.L. Radio wave propagation and radio link operation. -M .: Communication, 1979, 293 p.

 4. Chelyshev V.D. Reception radio centers. -M.: Communication, 1975, 197 p.

 5. Search, detection and measurement of signal parameters in radio navigation systems. / Ed. Yu.M. Kazarinova. -M .: Soviet radio. 1975, p. 85-127.

 6. Gonorovsky V.G. Radio circuits and signals. -M .: Radio and communications, 1981, 347 p.

Claims (1)

 The method of incoherent reception of discrete signals, which consists in comparing the input processes in two reception channels, calculating a single step function, forming the transfer function and the transfer function of the complement and multiplying them respectively with the input processes in the first and second reception channels, characterized in that the preliminary input processes in the reception channels are smoothed in a predetermined time interval corresponding to the noise correlation interval, based on them, pulse functions are formed in each reception channel They integrate them with a time constant corresponding to the correlation interval of the signal, after which the amplitudes of the integrated pulse functions of the first and second reception channels are compared, according to the results of which a single step function is calculated, from which the transfer function and the transfer function of the addition are formed, which are multiplied with the input processes of the channels receptions having correspondingly larger and smaller values of the amplitudes of the integrated pulse functions, and the processes obtained after multiplication are not edayut tract in further processing.

Images