RU2653485C1 - Method for adaptive selection of the optimal parameter of the signal correction algorithm - Google Patents

Abstract

FIELD: wireless communication equipment.

SUBSTANCE: invention relates to a communication equipment and can be used in data communication systems with adaptive signal correction for selecting a parameter of the correction algorithm. Using algorithm R {} based on received test signal u₀(t), the impulse response of the correction filter h_cor(t,α) is calculated, with the help of said impulse response, using algorithm R {}, received test signals u₁(t)…u_n(t), delayed by an interval equal to the length of information signal L_i, is corrected, resulting in corrected test signals K₁(t,α)…K_n(t,α), error values are determined, e_j=||K_j(t,α)-K(t)||, j=1…n, dependence of error value e_1…e_n on parameter α is determined by changing the value of this parameter, resulting in an array of parameter values α₁…α_n, which provides the corresponding minimum error value e₁…e_n for each corrected test signal K₁(t,α)…K_n(t,α), then from array α₁…α_n, the final optimal value of the parameter α_opt is selected.

EFFECT: technical result is selecting the optimal parameter of the signal correction algorithm without knowledge of a priori information with increased accuracy of estimate α_opt.

1 cl, 3 dwg

Description

The invention relates to communication technology and can be used in data transmission systems with adaptive signal correction to select a correction algorithm parameter.

In many data transmission systems, adaptive signal correction algorithms are used to compensate for distortions introduced by the communication channel. For this, periodic insertions of a test signal known on the receiving side are carried out in the transmitted signal. This approach is used, for example, in the ARINC-635 standard for airborne data transmission [1].

A large number of different methods, algorithms and their modifications are known that are used for signal correction, for example, the least squares method or the LMS algorithm [2], the RLS algorithm [3], the Tikhonov regularization method [4]. In all these and many other algorithms, to ensure stability and convergence, a certain parameter is introduced, in particular, the regularization parameter (RLS algorithm, Tikhonov regularization method), convergence step size (LMS algorithm). The choice of one or another parameter has a significant effect not only on the stability of the solution, but also on the probability of an error per bit after demodulation of the corrected signal, i.e. noise immunity.

As you know, the task of adaptive correction is reduced to solving two equations, which can be written in the following form:

where K (t) is the transmitted test signal, u (t) is the received test signal, K _m (t) is the transmitted information signal, u _m (t) is the received information signal, h (t) is the channel impulse response, * - convolution operator.

From equation (1) get the approximate impulse response of the channel in the General case in the form:

and the correction result in this case can be written as:

where R {} is a calculation algorithm, α ₁ , α ₂ are the parameters used for the stability of the algorithm. Note that in most practical cases it is acceptable to accept:

then instead of (5) we write:

Various methods are known for choosing the optimal value for this parameter.

, после чего определяют зависимость значения ошибки е от параметра α путем изменения значения этого параметра, в результате чего получают оптимальное значение параметра α_opt, обеспечивающего минимальное значение ошибки е.Closest to the claimed technical solution is the residual method described in [5]. This method is often used to select the regularization parameter in the Tikhonov regularization method. In conditions (3) (6), the residual method consists in the fact that, using the algorithm R {} based on the received test signal u ₀ (t), the impulse response of the channel h (t, α) and the correction filter h _cor (t, α), using which, using some algorithm R {}, the incoming information signal u _m (t) is corrected, as a result of which a corrected information signal K _m (t, α) is obtained, after which the error value e is determined, which serves as difference of the standard deviation of the corrected information signal K _m (t, α), folded with the calculated impulse response of the channel h (t, α), from the received information signal u _m (t) and the dispersion of the noise component Δu, i.e.

, after which the dependence of the error value e on the parameter α is determined by changing the value of this parameter, as a result of which the optimal value of the parameter α _{opt is} obtained _, which ensures the minimum value of the error e.

The disadvantage of the prototype is the need to know certain a priori information, namely the variance of the noise component, the assessment of which is a separate rather complicated task and has a certain error. In addition, when calculating the value of the error e, an additional error is introduced when convolving the approximate (calculated) values of K _m (t, α) and h (t, α).

The aim of the invention is the selection of the optimal parameter of the signal correction algorithm without knowledge of a priori information and without introducing additional error, i.e. increasing the accuracy of the assessment of α _opt .

, j=1…n, после чего определяют зависимость значения ошибки е₁…е_n от параметра α путем изменения значения этого параметра, в результате чего получают массив значений параметров α₁…α_n, обеспечивающих соответствующее минимальное значение ошибки е₁…е_n для каждого откорректированного тестового сигнала K₁(t,α)…K_n(t,α), после чего из массива α₁…α_n осуществляют окончательный выбор оптимального значения параметра α_opt, в качестве которого, в зависимости от конкретного алгоритма R{} и диапазона значений параметров α₁…α_n, берут среднее арифметическое значение или медианное значение из массива α₁…α_n.This goal is achieved by the fact that the method of adaptive selection of the optimal parameter of the signal correction algorithm, which consists in the fact that using the algorithm R {} based on the received test signal u ₀ (t), the impulse response of the correction filter h _cor (t, α) is calculated, characterized in that, using the impulse response of the correction filter h _кop (t, α), using the algorithm R {}, the incoming test signals u ₁ (t) ... u _n (t) are delayed by an interval equal to the length of the information signal L _AND , resulting in receive the corrected test signals K ₁ (t, α) ... K _n (t, α), determine the error values e ₁ ... e _n , which is the standard deviation of the corrected test signal K ₁ (t, α) ... K _n (t , α) from the reference test signal K (t), i.e.

, j = 1 ... n, after which the dependence of the error value e ₁ ... e _n on the parameter α is determined by changing the value of this parameter, resulting in an array of parameter values α ₁ ... α _n providing the corresponding minimum error value e ₁ ... e _n for each corrected test signal K ₁ (t, α) ... K _n (t, α), after which from the array α ₁ ... α _n final selection of the optimal value of the parameter α _{opt is} carried out, for which, depending on the particular algorithm R { } and range of parameter values α ₁ ... α _n, taking the arithmetic mean FRESH value or median value of array α ₁ ... α _n.

In FIG. 1, 2, a structural diagram of a method for adaptively selecting the optimal parameter of a signal correction algorithm is presented.

It contains:

1 - delay line;

2 (1) -2 (n) - processing unit;

3 - parameter selection block.

In turn, each processing unit 2 (1) -2 (n) contains:

2.1 - block calculation of the impulse response;

2.2 - correction filter;

2.3 - error assessment unit;

2.4 - a decisive device.

The method is as follows. The input of the delay line 1 receives a signal containing periodically repeated test and information signals. The structure of such a signal is shown in FIG. 3. The length of each test signal is L _T , the length of each information signal L _AND . The number of taps (outputs) of the delay line 1, equal to n + 1, can be different and is selected based on the specific application. From each of the n + 1 outputs of delay line 1, test signals u ₀ (t) ... u _n (t) of length L _{T are} received, delayed by an interval equal to the length of the information signal L _AND . At the same time, from the 0th output of the delay line 1, the test signal is fed to the first input of each processing unit 2 (1) -2 (n), the second input of which receives test signals from the outputs 1 ... n of the delay line 1, namely, from the 1st output of delay line 1 to the second input of processing unit 2 (1), from the 2nd output of delay line 1 to the second input of processing unit 2 (2), etc.

In each processing unit 2 (1) -2 (n), the following is carried out.

, j=1…n. Полученное на выходе блока оценки ошибки 2.3 значение ошибки е₁…е_n подают на вход решающего устройства 2.4, в котором определяют зависимость значения ошибки е₁…е_n от параметра α путем изменения значения этого параметра на первом выходе решающего устройства 2.4 и передаче его на управляющие входы блока расчета импульсной характеристики 2.1 и корректирующего фильтра 2.2. В результате на втором выходе решающего устройства 2.4, являющегося выходом блока обработки 2(1)-2(n), получают значение параметра α₁…α_n, обеспечивающего минимальное значение ошибки е₁…е_n.The test signal u ₀ received from the first input of processing unit 2 (1) -2 (n) is fed to the input of the impulse response 2.1 calculation unit, from the output of which the impulse response of the correction filter h _kop (t, α) is supplied to the first input of the correction filter 2.2. In this case, the corresponding test signal u ₁ (t) ... u _n (t), received from the second input of the processing unit 2 (1) -2 (t), is fed to the second input of the correction filter 2.2. At the output of the correction filter 2.1 receive the corresponding corrected test signal K ₁ (t, α) ... K _n (t, α), which is fed to the input of the error evaluation unit 2.3. At the same time, in the block for calculating the impulse response 2.1 and the correction filter 2.2, the same algorithm is used, previously indicated as R {}. In the error estimation unit 2.3, an error value is obtained, which is the standard deviation of the corrected test signal K ₁ (t, α) ... K _n (t, α) from the reference test signal K (t), i.e.

, j = 1 ... n. The error value e ₁ ... e _n obtained at the output of the error estimation unit 2.3 is fed to the input of the resolver 2.4, in which the dependence of the error value e ₁ ... e _n on the parameter α is determined by changing the value of this parameter at the first output of the resolver 2.4 and transferring it to control inputs of the block for calculating the impulse response 2.1 and the correction filter 2.2. As a result, at the second output of the resolver 2.4, which is the output of the processing unit 2 (1) -2 (n), the value of the parameter α ₁ ... α _{n is} obtained, which ensures the minimum error value e ₁ ... e _n .

The parameter values obtained at the output of processing blocks 2 (1) -2 (n), which are an array of α ₁ ... α _n , are fed to the corresponding inputs of parameter 3 selection block. In parameter 3 selection block, the final parameter α is finally selected, obtaining the optimal parameter at the output the value of the parameter α _opt . Depending on the specific algorithm R {} and the possible range of values of the parameters α ₁ ... α _n received at the inputs of the parameter 3 selection block, the arithmetic mean value is taken as the optimal value of the parameter α _ort :

or median value corresponding to:

where the array α ₁ '... α _n ' corresponds to the values α ₁ ... α _n sorted in ascending order.

The technical result is the selection of the optimal parameter of the signal correction algorithm with increased accuracy of estimation α _opt .

List of sources

1. ARINC Characteristic 635-4. HF Data Link Protocol. - Dec., 2003.

2. Dzhigan V.I. Adaptive signal filtering: theory and algorithms. - M .: Technosphere, 2013 .-- 528 p.

3. Sayed A.N. Adaptive filters. - New Jersey: Hoboken: John Wiley & Sons, Inc., 2008 .-- 786 c.

4. Tikhonov A.H., Arsenin V.Ya. Methods for solving incorrect tasks / Textbook for universities. - Ed. 3rd fix - M .: Nauka, 1986 .-- 288 p.

5. Verlan A.F., Sizikov B.C. Methods for solving integral equations with computer programs. Reference manual. - Kiev: Naukova Dumka, 1978.- 292 p.

Claims (1)

The method of adaptive selection of the optimal parameter of the signal correction algorithm is that using the algorithm R {} based on the received test signal u ₀ (t), the impulse response of the correction filter h _kor (t, α) is calculated, characterized in that using the impulse response of the correction filter h _cor (t, α), using the algorithm R {}, correct the incoming test signals u ₁ (t) ... u _n (t), delayed by an interval equal to the length of the information signal L _And , as a result, receive the corrected test signal s K ₁ (t, α) ... K _n (t, α), determine the error values e ₁ ... e _n , which is the standard deviation of the corrected test signal K ₁ (t, α) ... K _n (t, α) from the reference test signal K (t), i.e.

, j = 1 ... n, after which the dependence of the error value e ₁ ... e _n on the parameter α is determined by changing the value of this parameter, resulting in an array of parameter values α ₁ ... α _n providing the corresponding minimum error value e ₁ ... e _n for each corrected test signal K ₁ (t, α) ... K _n (t, α), after which from the array α ₁ ... α _n final selection of the optimal value of the parameter α _{opt is} carried out, for which, depending on the particular algorithm R { } and range of parameter values α ₁ ... α _n, taking arithmetic mean cal value or median value of array α ₁ ... α _n.

Images