RU2251704C2 - Method of detection of periodic pulse sequences and estimation of period sequences - Google Patents

Abstract

FIELD: measurement technology; computer technology.

SUBSTANCE: method can be used in devices for detecting and evaluating parameters of random pulse flows with discrete time. Method is based upon measurement of time intervals between moments of arrival adjacent pulses of input flow and storage of primary sample (mass) of N values of time intervals. Innovation concludes in calculation of first, second,…, (N+7)-th secondary samples of values of time intervals taking values of the primary sample of time intervals among moments of arrival of the first, second, …, (N+7)-th (reference) pulse correspondingly before moments of arrival of all other pulses of initial realization of the flow. Statistical estimations of greatest common divisor (GCD) are calculated within limits of separate secondary samples of two and more time intervals. Particular rows of estimation distributions of GCD of separate secondary samples of time intervals are formed. Particular rows of distribution of estimations of GCD are integrated where grouping of their values is revealed. Decision on detection of periodical pulse sequences is made on the base of facts of grouping of GCD estimations within integrated row of distribution of their values. Average values of revealed groups of GCD estimations are taken as estimations of periods of detected pulse sequences.

EFFECT: reduced time of detection; improved precision of estimation.

8 dwg

Description

The invention relates to measuring and computing equipment and can be used in devices for the joint detection and estimation of parameters of random pulse flows with discrete time observed in the paths of the post-detector signal processing at an unknown period of their succession, for example, during radio reconnaissance of radio emission sources with programmed tuning of the operating frequency (MHF) , as well as in monitoring the radio frequency broadcast in radio astronomy, radiophysics, radio communications and radar.

A known method of detecting a periodic sequence of pulsed signals, including recording using a thin electron beam the implementation of the signal stream on an analog medium (magnetic drum, potentioscope), the rotation speed of which (drum, beam) is coordinated in time with individual implementations of the signal stream coming from the output of the receiving device , accumulation on the recording surface of an analog signal storage device and the decision to detect a periodic sequence of pulse signals the catch for exceeding the accumulated signal of a certain threshold [1. Finkelstein M.I. Basics of radar. - M .: Owls. Radio, 1973, p. 254].

Note that the number of analog pulse storage devices includes a large group of pulse storage devices on the delay lines with positive feedback [2. Lezin Yu.S. Optimum filters and drives of pulse signals. - M .: Owls. radio, 1969].

A disadvantage of the known method based on analog signal accumulation is the limited number of effectively processed pulses and the inability to detect a periodic pulse sequence with an unknown repetition period.

There is also a method of detecting a periodic sequence of pulsed signals, including double (in time and amplitude) discrete transformation of the input signal shape, delaying the converted signal stream in the form of a sequence of sendings “0” and “1” to N _s ≥ 2 discrete quantities τ ₃ (n) = nТ, n = 0, 1, 2, ..., N _s -1 and checking the condition for the coincidence of direct pulses (n = 0) and N _s -1 of the delayed sequences “0” and “1” in time, the decision to detect a periodic sequence of pulsed signals taken upon coincidence to units (“1”) of N _s sequences (2≤ k≤ N _s ) [1. Finkelstein M.I. Basics of radar. - M .: Owls. Radio, 1973, p.257].

The method based on the method of matching pulse signals has no restrictions on the number of effectively processed pulses.

A disadvantage of the known method for detecting a periodic pulse sequence based on the matching method is the inability to detect a periodic pulse sequence with an unknown repetition period.

Closest to the technical nature of the proposed method for detecting periodic pulse sequences is a method for detecting periodic pulse sequences and evaluating their period, implemented in the device [3. RF patent №2003988, IPC G 01 R 23/10, 1993] and including the measurement of time intervals (VI) between adjacent pulses of the input stream implementation, the accumulation of a sample (array) of N values of VI, determination of the maximum value of VI (t _{i-1, i} ) Suite in the sample, modular transformation of the VI in the form of:

counting the number N _{j of the} same residue values Δ t _ij for each value of the trial period T _j (conversion module) and conducting threshold tests N _j ><N _pop to detect a periodic pulse sequence with an estimate of its period in the form of the value of T _j at which _Exceeded threshold N _pop .

The disadvantages of this method are the large time required to detect randomly thinned out periodic pulse sequences and the low accuracy of the estimation of their period. As follows from (1), a large detection time is due to inevitable errors (“misses”) in the choice of the initial value of T _{j at the beginning of the} trial period due to possible cases when the maximum VI (t _{i-1, i} ) _max exceeds the true period of the decimated pulse sequence (see figure 2, the interval between the second and third pulses τ ₂₃ ). The low accuracy of the period estimation is associated with the use of the minimal modulus method in the known method, which is not optimal when processing VI for real normally distributed errors of observation of the moments of arrival of pulses of periodic sequences [4. Gilbo E.P., Chelpanov I.B. Signal processing based on ordered selection. - M .: Owls. Radio, 1975, p. 68]. For normal VI observation errors, the least squares method is optimal [5. Linnik Yu.V. Least squares method and the basics of observation processing theory - M .: Fizmatgiz, 1962, p.123].

The objective of the invention is to reduce the detection time of randomly thinned periodic pulse sequences and to improve the accuracy of estimating their period in the presence of interfering pulses.

The problem is solved due to the fact that in the known method, including measuring the VI between the moments of arrival of adjacent pulses of the input stream implementation, the accumulation of the primary sample (array) of N VI values, the first, second, ..., ( The N + 1) -th secondary samples of the values of the VI between the moment of arrival of the first, second, ..., (N + 1) -th (reference) pulses, respectively, until the moments of arrival of all other pulses of realization of the input stream, are calculated within separate secondary samples of the VI statistical estimates the largest common divisor (GCD) for possible combinations of two or more VIs, form private series of distribution of GCD estimates for individual secondary samples of SI, combine the private series of distribution of GCD estimates in which a grouping of their values is revealed, decide on the detection of periodic impulse sequences according to the facts grouping the estimates of GCD in the combined series of the distribution of their values, and for the estimates of the periods of detected pulse sequences take the average values of the identified groups of GCD estimates .

For the statistical estimation of the GCD of the values of the VI between the pulses of the input stream, the maximum coefficient of direct regression passing through the beginning of rectangular coordinates with the least sum of squares of deviations from the nodal points of the semilattice (τ, m], where τ is the ordinate axis of continuous values of the VI; m = 1, 2 , ... is the abscissa axis of the integer multiplicities of the VI in units of the unknown period of the detected pulse sequence [6. Anishin AS, Baturin Yu.O. “Radio Engineering” Journal, 2002, No. 10, pp. 73-77].

The fact of grouping the estimates of the GCD in the distribution of their values is revealed by exceeding the specified threshold by the number of GCD estimates falling within the identity interval (proximity) of their values specified in magnitude.

In the proposed method, the problem of joint detection and estimation of the pulse sequence period is solved by taking into account differences in the properties of random flows of pulses (events) with continuous and discrete time.

A discrete scale of moments t _n = nT, n = 1, 2, ..., of the possible occurrence of true pulses of a periodic sequence is detected by the results of multiple statistical estimates of GCD for various combinations of two or more VI between pulses of the input stream implementation. Due to the fact that the probability of anomalous errors and the conditional variance of the GCD estimates obtained from combinations of discrete VIs generated by pulses of a periodic sequence are significantly lower than the similar quality indicators of the GCD estimates for combinations including random VIs with a continuous distribution, the decision rule for detecting a periodic pulse sequence is based on the identification of the facts of grouping estimates of GCD in the combined series of the distribution of their values. At the same time, the average values of the revealed groups of GCD estimates are taken as estimates of the periods of detected periodic pulse sequences.

In more detail, the essence of the proposed method for the joint detection of a periodic pulse sequence and the evaluation of their period is as follows.

Let there be a realization of a random stream of standard pulses in shape (FIG. 2), including some pulses No. = 2, 3, 5, 6, 8 of a randomly thinned periodic sequence and interfering pulses No. = 1, 4, 7 of a random stream with continuous time .

First, we make the assumption that the errors in observing the moments of arrival of pulses of a periodic sequence are equal to zero.

при τ _ij означают номера импульсов, которыми определяются соответствующие ВИ первичной и вторичных выборок.We write the initial sample of N = 7 VI between adjacent pulses of the input stream in the zero row of the table shown in figure 5. Based on the values of the SI of the primary sample, we calculate the first, second, .... eighth secondary samples of the SI, counted in two directions (to the “past” and “future”) from the moment of the appearance of the first, second, etc. (reference) pulses, respectively, until the appearance of each pulse of the input implementation of a random stream. The values of the VI of the first, second, etc. We write the secondary samples in the corresponding rows of the table. Indices i and

when τ _ij mean the numbers of pulses that determine the corresponding VI of the primary and secondary samples.

Analysis of the contents of the table allows us to draw the following conclusions. The time intervals recorded in the first, fourth and seventh row of the table are continuous random variables, since they are formed by pairs of pulses, of which at least one (reference) is a pulse of a random stream with continuous time. In other rows of the table there are random VIs with both continuous and discrete distribution. In particular, random VIs with a discrete distribution in the table are marked with hatching.

Within each secondary selection of VIs, we organize a search of possible combinations of two or more VIs and for each combination of VIs using the operator of statistical estimation of GCD, we determine the most plausible assessment of GCD.

Let us consider the features (character) of the distribution of estimates of GCD obtained in the framework of separate secondary samples of VIs on the subsets of various combinations of their VIs. It is known [6] that the distribution of GCD estimates using the GCD operator of two or more discrete random variables with the structure τ _k = m _k T, k = 1, 2, ..., N; where m _k are random natural numbers (the so-called multiplicities of VI in units of 7), is discrete with the following possible values of the estimates: GCD = γ T, γ = 1, 2, ... In this case, the probability of occurrence of the GCD = T estimate with increasing size sample N≥ 2 increases by law

and for values of N≥ 2 it is much higher than the sum of the probabilities of the appearance of the remaining values of the estimates of GCD = 2T, 3T ....

On the contrary, when processing the implementation of random variables with a continuous or mixed distribution, the operator of a statistical estimate of the GCD gives estimates of the GCD with a continuous distribution.

Taking into account the noted selective properties of the GCD operator (statistics) in the processing of discrete random variables, coincidence of estimates is possible, and in the processing of VI, among which there is at least one random VI with a continuous distribution, such coincidence of estimates is theoretically excluded.

Obviously, taking into account the observation errors of discrete HIs, the noted property of coincidence of the GCD estimates does not disappear without a trace, but is transformed into the property of grouping these estimates in the vicinity of T, 2T, 3T, .... The zone of the most likely grouping of GCD estimates in accordance with (2) is the vicinity of the desired parameter T. As an illustration, Figures 3 and 4 show the experimental distributions of the estimates HÔD / T in the absence (σ Δ = 0) and the presence (σ Δ = 0.05 T) of observation errors N = 3 discrete random VIs distributed over geometric law with param rum p = 0.125, respectively, obtained by statistical evaluation simulation operator GCD.

) частном ряду распределения выявляют по превышению заданного порога N₀₁(N₀₁<N) количеством оценок, попадающих в определенный по величине интервал Δ T₁ близости (идентичности) их значений.Within the framework of separate secondary samples, VIs form private series of distribution of estimates of GCD of various combinations of two or more VIs. The fact of grouping the estimates of GCD in the jth (

) a private row of the distribution is revealed by exceeding a predetermined threshold N ₀₁ (N ₀₁ <N) by the number of ratings falling within a certain interval Δ T _{1 of a} proximity (identity) of their values determined by the magnitude.

The private series of the distribution of estimates of GCD, in which the grouping of their values is revealed, are summarized in the combined series of distribution. The latter is subjected to a final test to identify the facts of grouping the estimates of the GCD with its parameter values for the decision rule: N ₀₂ > N and Δ T ₂ = Δ T ₁ . When identifying the grouping of estimates of GCD in the combined row, a decision is made to detect a periodic impulse sequence with an estimate of its period equal to the average value of the estimates of GCD in the group.

Thus, the introduction into the inventive method of the operator of a statistical estimate of the GCD of the values of SI realized by the least squares method allowed to reduce the time (due to the reliability) of detecting thinned periodic pulse sequences and to improve the accuracy of estimating their period.

The proposed method is new, because from the generally accepted information there is no known way to jointly detect a periodic pulse sequence and evaluate its period under conditions characterized by random skipping of pulses in the presence of interfering pulses, including the measurement and accumulation of N selective VIs between adjacent pulses of the input stream, calculation of N + 1 secondary waveform samples, counted from each (reference) pulse to other pulses of a fixed flow realization, statistical estimation of GCD values VI in various combinations, the formation of a combined series of distribution of GCD estimates and the adoption of a positive decision on the detection of a periodic impulse sequence by grouping the estimates of GCD in the combined series of distribution of their values, while the average value of the identified group of GCD estimates is taken as an estimate of the period of the detected pulse sequence.

The proposed method has an inventive step, since it does not explicitly follow from published scientific data and known technical solutions that the claimed sequence of operations for processing the implementation of a pulse signal stream using the GCD operator will allow detecting randomly thinned periodic pulse sequences with an estimate of their period.

The proposed method is industrially applicable, as for its technical implementation can be used typical logical and structural elements of discrete and computer technology.

Figure 1 shows the structural diagram of a possible variant of the technical implementation of the method, figure 2 the implementation of the input stream φ (t) in the form of a randomly thinned periodic sequence of pulses (No. 2, 3, 5, 6, 8) in the presence of interfering pulses (No. 1, 4, 7), the ⊗ sign on the time axis t indicates the moments of discrete time t _n = nT, n = 1, 2, ..., figure 3 shows the characteristic of the GCD operator in the form of the experimental distribution of GCD estimates (parameter T) in the absence of observation errors of discrete random VI, in Fig.4 - characteristic operator and GCD in the presence of observation errors of discrete random VIs, in Fig. 5 is a table of primary (rows No. 0) and secondary (rows No. = 1 ... 8) VI samples generated when each input pulse implementation is used as a reference for each pulse; Fig.6 shows an experimental private series of distribution of estimates of GCD, in which their grouping is revealed, Fig.7 - experimental private series of distribution of estimates of GCD, in which there is no grouping, Fig.8 - a combined series of distribution of estimates of GCD, in which their grouping.

, запоминающее устройство (ЗУ) 2 на N двоичных чисел, вычислитель 3 вторичных выборок ВИ, группу 4 из N управляемых многоразрядных ключей, N-разрядный двоичный счетчик 5, вычислитель 6 оценок НОД различных сочетаний из двух и более ВИ τ _ij,

, i≠ j и анализатор 7 распределения оценок НОД, при этом вход измерителя ВИ 1, объединенный со входом записи ЗУ 2, является входом последовательности импульсов (фиг.2), выход измерителя ВИ 1 соединен с информационным входом ЗУ 2, N многоразрядных выходов которого соединены с соответствующими входами вычислителя 3 вторичных выборок ВИ, N выходов которого соединены с входами соответствующих ключей группы 4, выход старшего разряда двоичного счетчика 5 соединен с управляющим входом вычислителя 3, а разрядные выходы счетчика 5 - с управляющими входами соответствующих ключей группы 4, выходы которых соединены с входами вычислителя 6 оценок НОД, выход которого соединен с информационным входом анализатора 7 распределения оценок НОД. Анализатор 7 имеет вход 1 задания дискретного интервала Δ Т распределения оценок НОД (интервала их идентичности) и входы 2 и 3 задания двух порогов N_{0 1} и N_{0 2} в виде числа оценок НОД на интервале Δ T для выявлении фактов группирования оценок НОД в частных и объединенном рядах распределения соответственно, а также информационные выходы (вых. 1 и вых. 2) факта обнаружения и значения оценки периода

обнаруженной периодической импульсной последовательности соответственно.A device that implements the claimed method (figure 1) contains a measuring device VI 1, τ _{i, i + 1} ,

, a storage device (memory) 2 for N binary numbers, a calculator 3 of the secondary samples of the VI, a group of 4 N-controlled multi-bit keys, an N-bit binary counter 5, a calculator of 6 estimates of the GCD of various combinations of two or more VI τ _ij ,

, i ≠ j and the analyzer 7 of the distribution of estimates of the GCD, while the input of the VI meter 1, combined with the recording input of the memory 2, is the input of the pulse sequence (figure 2), the output of the VI meter 1 is connected to the information input of the memory 2, N multi-bit outputs of which connected to the corresponding inputs of the calculator 3 secondary samples VI, N outputs of which are connected to the inputs of the corresponding keys of group 4, the high-order output of the binary counter 5 is connected to the control input of the calculator 3, and the bit outputs of the counter 5 are connected to the control inputs The appropriate key group 4, whose outputs are connected to inputs of the calculator 6 estimates the GCD, the output of which is connected to the data input of the analyzer 7 GCD distribution estimates. The analyzer 7 has an input 1 of the job of the discrete interval Δ T of the distribution of GCD estimates (the interval of their identity) and inputs 2 and 3 of the job of two thresholds N _{0 1} and N _{0 2} in the form of the number of GCD estimates in the interval Δ T to identify the facts of grouping the estimates of GCD in private and the combined distribution series, respectively, as well as information outputs (output 1 and output 2) of the fact of detection and the value of the evaluation period

detected periodic pulse sequence, respectively.

A device that implements this method works as follows. The measuring instrument VI 1 determines the values of N VI between the moments of arrival of adjacent pulses of the input stream, which are accumulated in the memory 2 (see the zero line of the table in figure 5). Then, the calculator 3 by the zero state of the counter 5 calculates the first secondary sample of N VI (row No. 1 of the table in figure 5) when using as the reference point the moment of occurrence of the first (reference) pulse of the initial implementation of the stream.

As part of the first secondary selection of VIs using a group of 4 managed keys and an N-bit binary counter 5 passing 2 ^N eigenstates, possible combinations of two or more non-zero values of the VIs go to the calculator 6 of the statistical estimates of the GCD. The values of the GCD estimates from the output of the calculator 6 go to the input of the analyzer 7, which accumulates them in the framework of the first private row of the distribution (histogram) with an interval step Δ T ₁ = 0.001T. Upon the fact that threshold N _{01 is} exceeded by the number of GCD estimates on an arbitrary interval Δ T of the private distribution series, a group of GCD estimates (Fig. 6) or its absence (Fig. 7) is revealed.

When revealing the grouping of estimates of GCD in a private row (Fig.6), this distribution row is summed up in a combined row (Fig.8), otherwise, the private row of the distribution of estimates of GCD (Fig.7) is erased. In the future, the second secondary sample of the VI is calculated (row No. 2 of the table) and the operation of the device is repeated.

After the processing of the (N + 1) -th secondary sample of the IW upon the fact that the threshold N _{02 is} exceeded by the number of GCD estimates in an arbitrary interval Δ T ₂ = 0.001T of the combined distribution series (Fig. 8), they decide to detect a periodic pulse sequence (output 1 analyzer 7) and for estimating the period of the detected periodic pulse sequence, take the average value of the GCD estimates in the identified group (output 2 of analyzer 7).

The operation of the device that implements the claimed method is shown on the example of processing the implementation of the input pulse stream (figure 2), containing pulses of one thinned periodic sequence. The claimed method allows to detect two or more thinned out periodic pulse sequences with an estimate of their period.

A series 1 meter from N VI can be built on the basis of the well-known VI meter [Mirsky G.Ya. Radio-electronic measurements. - M .: Energy, 1975, p.175]. The storage device 2 can be performed on a multi-bit shift register for N binary numbers. Calculator 3 is an adder of binary numbers and can be performed on the basis of, for example, K 555 IM6 modules [Shilo V.L. Popular digital circuits. Directory. - M .: Radio and communications, 1987]. Binary counter 5 and group 4 of multi-digit keys are typical structural elements of computer technology. The calculator 6 of the statistical estimates of the GCD can be implemented on a microprocessor according to the well-known algorithm described in [6]. As an analyzer 7 of the distribution of estimates of GCD, devices for estimating the law of distribution of random variables can be used, supplemented by simple means of automatic analysis of the distribution of estimates of GCD [Smooth B.C. Probabilistic Computing Models. - M .: Nauka, 1973, p.143].

Thus, in the proposed method, instead of accumulating the number of identical residues in the prototype method, the number of similar in value estimates of GCD obtained on the set of possible combinations of VI between pulses of the input stream implementation is accumulated.

The claimed method has the following advantages compared with the prototype:

- it is characterized by increased reliability of detection of randomly thinned periodic pulse sequences due to the introduction of a new procedure for processing VIs in the form of GCD statistics of their values, which allowed to expand the scope of its application due to the possibility of joint detection and evaluation of non-energy parameters of frequency hopping signals with single-frequency radio surveillance in the presence of interfering signals;

- has increased accuracy and reliability of the estimation of the period T due to the statistical processing of VI using the least squares method and the introduction of the procedure for accumulating estimates, which reduces the level of anomalous errors;

- does not use a priori information about the search area of the period values of the detected pulse sequences.

Claims (1)

A method for detecting periodic pulse sequences and estimating their period, including measuring time intervals between the arrival times of adjacent pulses of the input stream implementation, accumulating a primary sample (array) of TV values of time intervals, characterized in that the first and second are calculated from the values of the primary sample of time intervals ..., (N + 1) -th secondary samples of the values of the time intervals between the moment of arrival of the first, second, ..., (N + 1) -th reference pulses, respectively, up to the moments of arrival of all other At the same time, in the framework of separate secondary samples of time intervals, calculate statistical estimates of the largest common divisor (GCD) of possible combinations of two or more time intervals, form private series of distribution of GCD estimates for individual secondary samples of time intervals, combine private series of distribution of GCD estimates , in which the grouping of their values is revealed, decide on the detection of periodic impulse sequences based on the facts of grouping estimates of GCD in the combined series of the distribution of their values, and for the estimates of the periods of the detected pulse sequences, the average values of the identified groups of GCD estimates are taken.

Images