RU2301492C2 - Method and device for transmitting voice information in digital radio communication system - Google Patents

Abstract

FIELD: radio engineering; transferring voice information in digital radio communication systems.

SUBSTANCE: in order to convert information word into code characters of various concatenated cascades, data bursts are formed on sending end, each burst incorporating information character sequence; information character sequence is supplemented in each data burst with sum of cyclic redundancy check CRC; intermediate code sequence is formed for which purpose information sequence of each data burst supplemented with CRC sum is encoded by recursive convolution code incorporating time-dependent rational transfer function without introducing redundancy, and lattice diagram of non-recursive rational transfer function code is formed; assigned code characters are added to intermediate code sequence formed earlier and encoded by recursive convolution code so as to ensure that the only final node were known at desired depth.

EFFECT: enhanced radio communication quality.

2 cl, 12 dwg

Description

The invention relates to the field of radio engineering, in particular to a method and device for transmitting voice data in a digital radio communication system.

Converting a voice message in a digital radio communication system includes the following steps:

- analog-to-digital conversion of a speech signal,

- channel coding,

- transmission over the air of the code word,

- decode input signal

- digital-to-analog conversion of a digital sequence into a speech signal

Analog-to-digital conversion and digital-to-analog conversion are performed by a vocoder known as a “waveform coder” (see L. R. Rabiner R. W. Schafer, Digital processing of speech signal. Prentice Hall. Englewood Cliffs New Jersey 1978 [1]). Currently, hybrid vocoders allow you to convert data packets containing errors into a voice signal, if the number of erroneous characters does not exceed the set limit, then the original signal can be restored with losses, subjectively subtle, to the subscriber. Thus, the urgent task is to improve the quality of communication when transmitting voice data in a digital radio communication system by reducing the symbolic error of error-correcting coding.

It should be noted that a code detecting errors is also used to detect an erroneous voice data packet. The most effective error detection technology is the cyclic checksum, which is added to the transmitted packet. A full description of the CRC cyclic checksum encoding method is given, for example, in the book Bleichut. Theory and practice of error control codes. - M: 1986 [2]. The cyclic checksum is an almost reliable way to detect errors. Using a checksum not only to detect errors, but also to correct them, can reduce the symbolic error of error-correcting coding.

Convolutional coding is known from Theoretical fundamentals of communications / D.Wozencraft, I.Jeckobs. - M .: Mir, 1969. - p. 640., Elias P. Coding for Noisy Channel / P. Elias // IRE Convention Record. - 1955. - Pt. 4. - P.37-46. [3].

The sequence of transmitted information symbols can be represented as a polynomial s (x) = s ₀ + s ₁ x + s ₂ x ² , where s ₀ , s ₁ , s ₂ are binary symbols located in the order of their transmission.

Then, the sequence of symbols at the output of the convolutional encoder shown in FIG. 1 (a) can be considered as a convolution of the impulse response of the encoder with the input sequence. In a compact form, linear convolution is written as the product of polynomials:

c (x) = s (x) g (x)

Coefficients with _i , of this polynomial are determined by the equalities

Convolution should be understood as a convolution in the Galois field. That is, for binary messages, the sum must be understood as a sum modulo 2, and multiplication by x as a delay operation. Thus, s _i is the input sequence; with _i - counting of the output sequence; g (x) is the generator polynomial, ag _k are its coefficients. To introduce redundancy using another generator polynomial g '(x), an output sample c' (x) is obtained. As a result of this, the input symbol s _i corresponds to two output c (x) and c '(x). The generatrix polynomial can also be represented as a binary number, where the kth digit is equal to the binary coefficient g _k . So for the encoder shown in Figure 1 (a), the polynomials forming are g = 5 ₈ and g '= 7 ₈ , where the subscript denotes the base of the number system. Physically, the encoder is implemented using a delay line. The encoder speed is the ratio of the number of input characters to the number of output characters, and the length of the code constraint is the length of the delay line, which is equal to the degree of the generating polynomial. The coding rate of this encoder is R = 1/2. The length of the code constraint is defined as the minimum length of the delay line at which the code can be implemented, and is equal to K = 3.

Another type of representation of the error-correcting code is the representation using the generating matrix, so in this representation the encoding operation can be written as the multiplication of the vector of information symbols by the generating matrix, the result of which will be a vector of code symbols: SG = C.

. Выходная последовательность, соответствующая произвольной входной последовательности, получается сложением по модулю 2 сдвигов указанной последовательности. Поэтому порождающая матрица может быть представлена в виде:The form of the generating matrix of the convolutional code is determined by the response to a single effect on the encoder shown in figure 1 (a). Let symbol 1 enter the encoder, followed by the characters 0. In this case, the response to such a single action, i.e., the corresponding output sequence of the encoder has the form

. The output sequence corresponding to an arbitrary input sequence is obtained by adding modulo 2 shifts of the specified sequence. Therefore, the generating matrix can be represented as:

As can be seen from the method of constructing the matrix, it has an infinite size, but, nevertheless, the encoder is quite simply implemented on the delay line.

The convolutional code can be represented by generating submatrices, each of which corresponds to a delayed one impulse response, as shown in FIG. 2 (see R. Johannesson and K. S. Zigangirov, Fundamentals of Convolutional Coding. New York, NY, USA: IEEE Press , 1999 [4]). The size of the submatrix corresponds to the number of generating polynomials G _k = {g _{k, 1} , g _{k, 2} } and determines the coding rate R. In this case, the generating matrix for the convolutional code can be represented as:

For the code shown in FIG. 1 (a), the generating submatrices are equal

Along with the term “generating polynomial”, the term “rational transfer function” from the delay operator D is used:

g _k (D) = g _k0 + g _k1 D + g _k2 D ² + g _k3 D ³ + ... + g _km D ^m

In this case, it is said that the information sequence is multiplied by the generating polynomial of the code or by the rational transfer function g _k (D), where D is the delay operator or D-transformation (see [4] p.31).

Coding redundancy is achieved by introducing a plurality of coding stages of the convolutional code g _k (D), g ₁ (D), ..., g _N (D). In each of the coding stages, one code symbol is generated. All code symbols generated in coding stages correspond to one information symbol received on the delay line. Coding operations in each cascade are carried out in parallel.

A recursive convolutional code or a convolutional code with feedback is known (see GDFomey, Jr., "Convolutional codes I: Algebraic structure," IEEE Trans. Inform. Theory, vol.IT-16, p.720-738, Nov. 1970 [5]). Feedback in the encoder is implemented by dividing by polynomial. Such an encoder is also characterized by a rational transmission function. A necessary condition for the feasibility of such a scheme is the equality to unity of an element with zero degree D in the denominator polynomial.

Figure 3 presents two options for implementing a recursive convolutional encoder and the corresponding rational transfer functions:

a) the rational transmission function of the convolutional encoder (option a)

b) the rational transfer function of the convolutional encoder (option b)

Convolutional codes allow a relatively simple soft decoding algorithm, known as the Viterbi algorithm (Viterbi A. Error limits for convolutional codes and the asymptotic optimal decoding algorithm. - see A. Viterbi. Some questions of coding theory. - M., 1970. - P.142 -165 [6]). Convolutional code decoding is based on comparing two code sequences by determining the "distance" between them in the space of valid code words or determining the "weight" of one code word relative to the other. For the case of sequences of binary symbols, the simplest metric that determines the distance between codewords is the Hamming metric, which is equal to the number of distinct characters in the sequences in question. A metric is also defined in soft solutions, which is equal to the sum of geometric distances in the space of signals between the corresponding code symbols in both sequences [2].

The process of decoding a convolutional code is a search for the path with the smallest weight in the tree, the branches of which are valid code combinations. Discarding deliberately wrong paths will avoid an exponential increase in complexity depending on the length of the code block. Given the fact that the encoder is an automaton with a finite number of states, it is convenient to graphically depict the code as a trellis diagram. For the encoder shown in FIG. 1 (a), the trellis diagram is shown in FIG. 1 (b), where the nodes correspond to the states of the encoder defined by the first two bits from the delay line, and the edges correspond to the permissible transitions and are marked with the corresponding values of the output bits. At the initial time t _{0, the} bits recorded in the delay line determine the initialization of the encoder, usually the encoder is initialized with zeros. Each column is the state of the encoder at a discrete time instant t _i . In the decoder, each possible path accumulates its weight by comparing it with a truly accepted sequence, using the Hamming metric or the metric in soft solutions. And thus, it moves into the depth of the trellis diagram, where the depth of the trellis diagram of the convolutional code q in the decoder corresponds to discrete time samples t _i in the encoding device. Note that two paths also come to each state, one of which with the worst metric is discarded.

The modified Viterbi decoding algorithm given in [4] is as follows:

- in the construction of an array of metrics equal to the number of internal states of the encoder, which is considered as an automaton with a finite number of states,

- initialization of the metric array with a priori metric values,

- calculation of path metrics that are equal to the initial state metric plus the admissible transition metric (ADD),

- updating state metrics by selecting the smallest metric of the allowed path to this state (COMPARE),

- remembering the allowed transition with the least metric (SELECT),

- repeating the described procedure until the end of the code word is reached,

- choosing the final state, tracking the main surviving path along the remembered allowed transitions in the opposite direction.

Viterbi's algorithm is optimal in its class of codes by the criterion of the minimum probability of error in receiving a code word, which, however, does not guarantee a minimum number of erroneous characters.

The MAP BCJR algorithm is known (see Optimal decoding of linear codes for minimizing symbol error rate / LR Bahl, J. Cocke, F. Jelinek, J. Raviv // IEEE Transaction on Information Theory. - 1974. - Vol.20, No. 3 . - P.284-287. [7]), which is optimal by the criterion of minimum symbolic error, but it requires complex implementation, large computational resources, and introduces processing delay. The algorithm consists in passing the code word in the forward and backward directions, saving the intermediate results and the subsequent passage, where the intermediate results combine and make soft decisions on the accepted characters.

Compared to the Viterbi algorithm, the MAP decoding process can only be started after receiving the codeword as a whole, which increases the decoding delay and requires complex computational operations.

The Viterbi algorithm allows you to decode a code word as code symbols arrive, the delay is introduced only by the decision making process on decoded symbols or the process of tracking surviving paths, however this delay is not significant and is determined by the maximum depth of the trellis diagram of the convolutional code, usually equal to 5-6 code restriction lengths .

There are known systematic and non-systematic types of coding, which differ in that in the case of systematic coding, the initial information message is an explicit part of the code word, that is, the code word can be divided into information and verification components.

For codes decoded by the Viterbi algorithm, the unsystematic type of encoding is optimal, however, it is known that any unsystematic convolutional code can be represented in a systematic form, and the possibility of its decoding by a standard decoder will be preserved (see [5], as well as Min-Goo Kirn , On Systematic Punctured Convolutional Codes, IEEE Transactions on Communications, Vol. 45 No.2, P.133-9, February 1997 [8]). This is achieved by preliminary convolutional recursive coding, without introducing redundancy, an intermediate code sequence can be formed in such a way that, being encoded by the standard code of the convolutional code, the resulting code word will contain the original information sequence as an integral part, thus, information symbols can be mapped to assigned code characters. The resulting code can be decoded with a standard Viterbi decoder, the original information sequence can be obtained both by reverse coding and by direct conversion of the intermediate code sequence into the original information message made directly in the decoder, with minor modifications of the latter. It should be noted that in this case, precoding is performed with feedback by a recursive convolutional encoder.

However, if you display the information sequence in only one of the coding stages, for example, when the encoding speed is R = 1/2, the effect of reducing the bit error cannot be achieved, due to the code dependence of the bits in this stage, so an error in one of the code symbols , with a high probability leads to an error in the next symbol of this cascade. However, the literature shows the effectiveness of this precoding method for feedback speed R = 1/2 in the channel with fading (see Channel coding techniques for Adaptive Multi Rate Speech Transmission / T. Hindelang, J. Hagenauer, M.Schmauts, Wen Xu [9]). The gain in the symbolic error is obtained due to the fact that the Viterbi algorithm is optimal only by the criterion of the minimum packet error, respectively, the gain is that the number of erroneous characters in the erroneous packets will be reduced due to preliminary coding with feedback and subsequent restoration of the information sequence from the intermediate code sequence.

A well-known way to increase the decoding efficiency is list decoding or list decoding (see Elias P. Error-Correcting Codes for List Decoders / P. Elias // IEEE Transactions on Information Theory. - 1991. - Vol.37, No. 1. - P .5-13. [10]), which consists in the fact that the decoder returns a list of possible codewords. The choice of the correct informational message can be made by checking the cyclic checksum. For example, in the Viterbi algorithm, sheet decoding can be implemented by adding to the list some paths that were previously discarded.

The following implementations of the Viterbi Sheet Algorithm are known:

- N. Seshadri, C.E.W. Sundberg "List Viterbi decoding and its application" // IEEE Trans on Communication, Vol.COM-42, No. 2-4, 1994, P.313-323 [11],

- Martin Röder, Raouf Hamzaoui "Fast List Viterbi Decoding and Application for Source-Channel Coding of Images" http://citeseer.ist.psu.edu/689661.html [12].

In fact, this algorithm is characterized by searching through the tree of saved transitions in the trellis diagram of the Viterbi decoder, where the search is carried out several levels in depth, changing the decisions made by the Viterbi decoder.

The Viterbi sheet algorithm described in [11] is characterized by the following sequence of actions:

1) choose the main surviving path, performing Viterbi decoding;

2) form the first list of alternative paths, tracking the main surviving path in the opposite direction, starting from the step with the highest number, for which at each step add alternative paths to the first list that were merged with the main surviving path;

3) for each path in the first list, the second list is formed, adding alternative paths to it that were merged with this path from the first list;

4) calculate the path metrics from the second list, if the path metric from the second list is less than the metric of any path from the first list, then the path in the first list is changed to the path from the second path list;

5) repeat the steps recursively until a step with a zero number is reached;

6) check the CRC for all paths from the first list;

The disadvantage of this algorithm is its technical complexity, the algorithm can be simplified by performing only steps 1, 2, 6, but in this case its effectiveness drops sharply, for example, in a channel with fading.

Closest to the claimed invention, the technical solution - the prototype is described in US patent No. 6112326, Int.Cl. H03M / 13, Preceding Technique to Lower the Bit Error Rate (BER) of Punctured Convolutional Codes / Ali S. Khayrallah; Ericsson Inc. [13], where precoding is performed by multiplying by a non-systematic matrix, consistent with the puncturing matrix for a number of certain convolutional codes with a code constraint of 5. It should be noted that in the prototype device, precoding is performed without feedback.

The method is characterized by the following sequence of actions:

1) on the transmitting side, the information signal is multiplied by the precoding matrix, obtaining a precoded information signal;

2) multiply the generating matrix having a first convolutional code rate by a precoded information signal, obtaining an encoded information signal;

3) multiply the puncturing matrix having a second speed of the convolutional code, which is higher than the first speed of the convolutional code, by the encoded information signal, receiving a punctured information signal;

4) transmit the punctured information signal on the channel;

5) on the receiving side receive a punctured information signal;

6) the received punctured information signal is multiplied by a matrix inverse to the puncturing matrix, receiving a punctured information signal;

7) decode a non-punctured information signal, obtaining the maximum estimated code sequence;

8) multiply the maximum estimated code sequence by the matrix inverse to the generating matrix, receiving an uncoded signal;

9) multiply the uncoded information signal by the matrix inverse to the precoding matrix, obtaining the transmitted information signal, where the precoding matrix is matched with the puncturing matrix, so as to reduce the rate of occurrence of errors in the transmitted punctured information signal.

The described prototype method is carried out on a device, two implementation options of which are given in [13]. The closest technical solution is a prototype of the claimed device is a device according to the second embodiment. The block diagram of the prototype device is made in figure 4 (transmitting device) and figure 5 (receiving device).

The prototype device contains a transmitting and receiving device connected via channel 9, while the transmitting device 1 (figure 4), contains the first 2, second 4, third 6 and fourth 8 switches, L precoding blocks without feedback 3 ₁ -3 _L , coding block 5 and L puncturing blocks of code symbols 7 ₁ -7 _L wherein the input of the first switch 2 is the input of the transmitting device, L outputs of the first switch 2 are connected to their corresponding inputs L pre-coding blocks without feedback 3 ₁ -3 _L outputs L locks precoding without feedback are connected to the L inputs of the second switch 4, the output of the second switch 4 is connected to the input of a coding block 5, whose output is connected to the input of the third switch 6, L outputs of which are connected with the corresponding inputs of L blocks puncturing code symbols January ₇ - 7 _L L outputs blocks puncturing code symbols July ₁ -7 _L L are connected to the fourth input of the switch 8, the output of which is the output of the transmitter and connected to the inlet channel 9 whose output is connected to the input riemnogo device 10.

The receiving device 10 (figure 5) contains the first 11, second 13, third 15 and fourth 17 switches, L blocks matching the number of code symbols 12 ₁ -12 _L decoding unit 14 and L decoding blocks of the preliminary code 16 ₁ -16 _L with the input the first switch 11 is the input of the receiving device, the L outputs of the first switch 11 are connected respectively to the inputs L of the blocks matching the number of code symbols 12 ₁ -12 _L , the outputs of which are connected to the L inputs of the second switch 13, the output of which is connected to the input of the decoding unit 14, the output which is connected to the input of the third switch 15, the L outputs of which are respectively connected to the inputs L of the decoding blocks of the preliminary code 16 ₁ -16 _L , the outputs of which are connected to the L inputs of the fourth switch 17, the output of which is the output of the device.

Carry out the prototype method on the device (Fig.4 and 5) as follows.

On the transmitting device (Fig. 4), an information signal is supplied to the input of the first switch 2, which it feeds to the input of one of the L precoding blocks without feedback 3.

In the open-loop precoding unit 3, the information signal is encoded by multiplying the information signal by the precoding matrix, obtaining a precoded information signal that is supplied to one of the L inputs of the second switch 4.

The second switch 4 supplies a precoded information signal to the input of the encoding unit 5, where it is encoded, receiving an encoded information signal, which is fed to the input of the third switch 6.

The third switch 6 supplies the encoded information signal to the input of one of the L code symbol puncturing units 7, which punctures the code symbols in the encoded information signal by multiplying the puncturing matrix having a second convolutional code rate that is higher than the first convolutional code rate by the encoded information signal, receiving a punctured encoded information signal.

The output signal from the puncturing unit of code symbols 7 is supplied to one of the L inputs of the fourth switch 8. The punctured encoded information signal is supplied from the output of the fourth switch 8 to the input of channel 9, through which it is transmitted to the receiving device 10.

At the receiving device 10 (Fig. 5), a punctured information signal is received on channel 9, a received punctured information signal is switched on the first switch 11 and transmitted to the input of one of the L blocks for matching the number of code symbols 12. In the block for matching the number of code symbols 12, the received punctured the information signal is multiplied by a matrix inverse to the puncturing matrix, receiving a non-punctured information signal, which from the output of block 10 is fed to one of the L inputs of the second switch 13.

From the output of the second switch 13, a non-punctured information signal is fed to the input of the decoding unit 14. In block 14, a non-punctured information signal is decoded to obtain an uncoded information signal (maximum estimated code sequence), which is transmitted from the output of block 14 to the input of the third switch 15.

From one of the L outputs of the third switch 15 to the input of the corresponding decoding unit of the preliminary code 16 receives an uncoded information signal.

In block 16, an uncoded information signal is decoded by multiplying the uncoded information signal by the matrix inverse to the precoding matrix, obtaining the transmitted information signal, the precoding matrix being matched to the puncturing matrix so as to reduce the rate of occurrence of errors in the transmitted punctured information signal. The transmitted information signal from the output of one of the L blocks 16 is fed to one of the L inputs of the fourth switch 17.

From the output of the fourth switch 17, the transmitted information signal is transmitted to the output of the receiving device 10.

This well-known solution is aimed at reducing symbolic errors in a data transmission system using a punctured convolutional code with a slight increase in complexity (the entire system and / or decoder), as well as providing uneven error protection in a system with punctured convolutional code. However, it still does not provide a reduction in the number of symbolic errors in codes with a coding rate greater than or equal to R = 1/2.

The method and prototype device described above require puncturing code symbols and are defined only for code with a code restriction length of K = 5, i.e. code that is weak. Therefore, the prototype method and device for its implementation cannot provide high quality radio communications.

It should be noted that the mentioned prototype [13] includes a number of feedback-free precoding devices specified by the corresponding set of matrices, a convolutional code encoding device specified by the generating matrix with the first convolutional code speed, and a puncturing unit containing a number of puncturing nodes specified by the corresponding set puncturing matrices, to form a punctured information signal transmitted over the channel, where the selected matrix of the precoding device without fraternal communication is consistent with the selected puncturing matrix in order to reduce BER (probability of error) in the transmitted information signal punctured, the block matching the number of code symbols, which are multiplied by a matrix inverse to the original puncturing matrix, the transmitted punctured information signal to form a non-punctured information signal, Viterbi decoder displaying a non-punctured information signal into a code sequence character estimation signal according to the maximum likelihood rule, a block is formed information sequences along the paths of the trellis diagram of the Viterbi decoder (uncoder), which multiplies the matrix inverse to the generating matrix by the code sequence symbol estimation signal for generating an uncoded information signal, and the block inverse to the applied pre-coding block, multiplying the matrix inverse to the selected precoding matrix, by uncoded information signal for generating a transmitted information signal.

The technical difficulty in implementing the prototype [13] lies in the difficulty of finding precoding matrices consistent with the generating matrix of the code and with the puncturing matrix for codes with a large code restriction.

The problem to which the invention is directed is to improve the quality of radio communications, while the technical result is achieved by reducing symbolic error when decoding a convolutional code according to the Viterbi algorithm.

In particular, to achieve a technical result, the inventive method for transmitting voice data in a digital radio communication system is developed, namely, that

on the transmitting side form data packets, each of which contains an information sequence of characters,

in each data packet complement the information sequence of characters with a cyclic CRC checksum,

, зависящей от времени t, где t - целое число

, g₀(D), g₁(D) - рациональные функции передачи сверточного нерекурсивного кода, используемого для помехоустойчивого кодирования, а х_t - двоичный символ псевдослучайной последовательности;an intermediate code sequence is formed, for which purpose the information sequence of each data packet is encoded with a convolutional recursive code without introducing redundancy with a rational transfer function

depending on time t, where t is an integer

, g ₀ (D), g ₁ (D) are rational transmission functions of a convolutional non-recursive code used for error-correcting coding, and x _t is a binary symbol of a pseudo-random sequence;

form a trellis diagram of the convolutional code,

add the assigned code symbols to the generated intermediate code sequence so that for the generated trellis diagram of the convolutional code the final node is known at a given depth q = t _max ,

encode the generated intermediate code sequence with a non-recursive convolutional code with rational transfer functions g ₀ (D), g ₁ (D), obtaining a code word with displayed information symbols;

transmit the code word on the channel;

on the receiving side, demodulation of the input signal is performed, obtaining a code word in soft decisions;

decode the code word in soft decisions by means of a Viterbi decoder for which:

initialize the Viterbi decoder metric array with a priori metric values, assuming that all paths exit from one node of the trellis diagram of the convolutional code at the initial time, put t = 0;

form a trellis diagram of the convolutional code, where - for each node at a depth of q = t + 1 find the predecessor nodes at a depth of q = t, with which it is connected by an allowed transition;

path metrics are computed for each predecessor node, adding up the metric of the source predecessor node and the metric of the allowed transition,

update the state metric by selecting the smallest metric of the allowed path to this state,

remember the allowed transition with the least metric,

repeat the calculation of path metrics, updating the state metric and remembering the allowed transition with the smallest metric until the specified depth q = t _max is reached,

choose a known final node, at a depth q = t _max ,

tracking the main surviving path, performing memorized transitions along the trellis diagram in the opposite direction, thus obtaining an intermediate code sequence from the main surviving path;

encode the intermediate code sequence with a non-recursive convolutional code with rational transfer functions g ₀ (D), g ₁ (D), obtaining the reconstructed code word;

, a x_t - двоичный символ псевдослучайной последовательности, синхронизированный с двоичным символом псевдослучайной последовательности на передающей стороне, получая восстановленную информационную последовательность, и проверяют циклическую контрольную сумму CRC, в случае ее совпадения заканчивают декодирование;carry out puncturing in the reconstructed codeword, where from each pair of code symbols corresponding to the transition in the trellis diagram of the convolutional code, a code symbol defined by the rational transfer function g _y (D) is left at a depth q = t, where y∈ {0,1} and

, ax _t is the binary symbol of the pseudo-random sequence synchronized with the binary symbol of the pseudo-random sequence on the transmitting side, obtaining the reconstructed information sequence, and the cyclic CRC checksum is checked; if it matches, the decoding is completed;

if the cyclic CRC checksum does not match, a list of alternative paths is formed, for which:

perform memorized transitions from the final node in the opposite direction until the next transition corresponds to code symbols, in one of which the i-th information symbol is displayed,

perform a transition along a different permitted path in the trellis diagram other than the transition that was memorized,

from the current node carry out memorized transitions along the trellis diagram in the opposite direction until reaching the beginning of the trellis diagram, forming an alternative path corresponding to the i-th information symbol,

for each information symbol form their own alternative path,

for each path from the list of alternative paths:

receive an intermediate code sequence, encode the intermediate code sequence with a non-recursive convolutional code with rational transfer functions g ₀ (D), g ₁ (D),

, а х_{t -} двоичный символ псевдослучайной последовательности, синхронизированный с двоичным символом псевдослучайной последовательности на передающей стороне, получая восстановленную информационную последовательность символов, и проверяют циклическую контрольную сумму CRC, в случае ее совпадения заканчивают декодирование,carry out puncturing in the reconstructed codeword, where from each pair of code symbols corresponding to the transition in the trellis diagram of the convolutional code, a code symbol defined by the rational transfer function g _y (D) is left at a depth q = t, where y∈ {0,1} and

, and x _{t is the} binary symbol of the pseudo-random sequence synchronized with the binary symbol of the pseudo-random sequence on the transmitting side, receiving the reconstructed information sequence of the symbols, and the cyclic CRC checksum is checked, if it matches, finish decoding,

return a character information sequence corresponding to the main surviving path if no CRC cyclic checksum check matches.

The technical result is also achieved through the development of the inventive device for transmitting voice data in a digital radio communication system containing a transmitting and receiving device connected by a channel providing data transmission, the transmitting device comprising a CRC cyclic checksum generation unit, an adder, an encoding unit, the first and second open-loop precoding blocks, multiplexer, psp all-random sequence generator and key, while the input of the block of cyclic checksum CRC is the first input of the device - the input of an information signal, the output of the cyclic checksum generating unit CRC is connected to the first input of the adder, the output of which is connected to the inputs of the coding block and the first and second precoding blocks without feedback, the first and second outputs of the block encoding connected respectively to the first and second inputs of the multiplexer, the output of which is the output of the transmitting device and connected to the input of the channel, the outputs of the first and the second pre-coding blocks without feedback are connected respectively to the first and second inputs of the key, the third input of which is connected to the output of the PSP generator, the input of which is the second input of the transmitting device - the input of the synchronization control signal, the output of the key is connected to the second input of the adder; the receiving device comprises a decoder sheet, N reverse encoding blocks, N keys, N cyclic checksum check blocks and a selection block, the input of the decoder sheet being the first input of the receiving device being an information signal input, and connected to the channel output, N outputs of the sheet decoders are connected respectively to the inputs of N reverse coding blocks, the first and second outputs of N reverse coding blocks are respectively connected to the first and second inputs of N keys, the third inputs of which are connected to the outputs of the PS generator P, the input of which is the second input of the receiving device - the input of the synchronization control signal, the outputs of the N keys are connected respectively to the inputs of the N blocks of the cyclic checksum check, the outputs of which are connected respectively to the N inputs of the selection block, the output of which is the output of the receiving device.

Distinctive features of the proposed method for transmitting voice data in a digital radio communication system relative to the prototype are the following:

on the transmitting side:

encode each packet of information symbols with a CRC code,

display information symbols in code symbols of a convolutional code with a coding rate of R = 1/2, wherein the encoding stage is selected according to a pseudo-random law,

forming a code symbol stream by multiplexing the code symbols into a serial stream without puncturing the code symbols,

on the receiving side:

decode the received sequence with a simplified sheet by the Viterbi decoder, forming a list of intermediate code sequences, and each information symbol has its own alternative path. The simplification is that the search is carried out only one level in depth,

in each intermediate code sequence, reverse coding is performed, restoring the sequence of code symbols,

in each restored sequence of code symbols, a restored information sequence is allocated according to a pseudo-random law, similar to that which was applied on the transmitting side to select a code symbol for display,

in each recovered information sequence, the cyclic CRC checksum is checked; if the CRC matches at least one recovered sequence, then it is declared decoded correctly.

The distinctive features of the claimed device for transmitting voice data in a digital radio communication system relative to the prototype are the following:

in the transmitting device

a block has been introduced for generating a cyclic CRC checksum, which complements the information sequence in each data packet, thus encoding the information sequence according to a certain rule;

an adder has been introduced that performs a symbolic summation modulo 2 of the information sequence with the signal from the key output by the augmented cyclic checksum, forming an intermediate code sequence;

a key has been entered that outputs a signal from one of the two precoding units without feedback according to a given rule;

the first and second blocks of pre-coding without feedback have a significant simplification compared to the prototype, which uses L blocks of pre-coding, in addition, they function differently in different ways:

in the first precoding block without reverse

the links encode the generated intermediate code

sequence with non-recursive convolutional code with

;rational transfer function

;

;in the second non-feedback precoding unit, the generated intermediate code sequence is encoded with a non-recursive convolutional code with a rational transfer function

;

, зависящей от времени t, где t - целое число и

, g₀(D), g₁(D) - рациональные функции передачи сверточного нерекурсивного кода, используемого для помехоустойчивого кодирования, а х_t - двоичный символ псевдослучайной последовательности;performing modulo 2 summation on adder 19, the information sequence of each packet is encoded with a convolutional recursive code without introducing redundancy with a rational transfer function

depending on time t, where t is an integer and

, g ₀ (D), g ₁ (D) are rational transmission functions of a convolutional non-recursive code used for error-correcting coding, and x _t is a binary symbol of a pseudo-random sequence;

a multiplexer is introduced, which of two parallel signals at the outputs of the coding unit forms a serial signal, which is a code word that is transmitted along the channel, and the puncturing operation necessary in the prototype is not required;

an SRP generator is introduced, which, by means of a control signal, provides a choice of a code symbol into which the information symbol will be displayed;

to the receiver

a decoder sheet was introduced, which generates the restored intermediate code sequences of the main and alternative paths, and the Viterbi sheet decoding algorithm is taken as the basis, the distinguishing part of the claimed invention from the well-known Viterbi sheet decoding are the following features, ensuring the achievement of the technical effect in a simpler way :

perform memorized transitions from the final node in the opposite direction until the next transition corresponds to code symbols, in one of which the i-th information symbol is displayed,

perform a transition along a different permitted path in the trellis diagram other than the transition that was memorized,

from the current node carry out memorized transitions along the trellis diagram in the opposite direction until reaching the beginning of the trellis diagram, forming an alternative path corresponding to the i-th information symbol,

for each information symbol form their own alternative path,

N reverse coding blocks are introduced, which are functionally similar to coding block 5 on the transmitting device, and N keys, which provide (by the synchronization control signal from the PSP generator 26) a simplified (compared with the prototype) decoding of the preliminary code, as a result, the signal outputs the keys which corresponds to a sequence of code symbols determined by a rational transfer function, thereby obtaining a reconstructed information symbol sequence,

N CRC checksum check blocks were introduced, in which the received recovered information sequences are checked by adding check characters to them - “zero” if the check was successful, or “one” if the cyclic checksum did not match;

a selection block has been introduced in which decoding is completed if the cyclic CRC checksum of any of the restored and verified information sequences coincides.

Thus, the claimed method of transmitting voice data in a digital radio communication system and a device for its implementation can improve the quality of radio communication by reducing symbolic error when decoding a convolutional code according to the Viterbi algorithm, the technical effect is achieved by mapping the information word into code symbols of various parallel convolutional coding coding stages , providing decoding efficiency with a simplified sheet decoder. In addition, the claimed invention has a significant simplification in implementation compared to the prototype, because eliminates the need for the formation of many matrices by enumerating them.

The following is a description of the claimed invention is illustrated by examples and drawings.

Figure 6 is a structural diagram of a transmitting device 1 according to the claimed invention.

7 is a structural diagram of a receiving device 10 according to the claimed invention.

On Fig shows the structure of the block 3 precoding without feedback, block 5 encoding and their rational transfer functions.

Figure 9 is a trellis diagram for a sheet decoder 23 based on the Viterbi algorithm.

10, a reverse coding unit 24, a key 25, and a cyclic checksum verification unit 27 are shown.

Figure 11 shows the simulation results of the proposed algorithm.

The inventive device for transmitting voice data in a digital radio communication system (Fig.6 and 7) contains a transmitting device 1 and a receiving device 10 connected by a channel 9, providing data transmission,

moreover, the transmitting device 1 (Fig.6) contains a CRC cyclic checksum generating unit 18, an adder 19, an encoding unit 5, a first block 3 ₁ and a second non-feedback precoding unit 3 ₂ , a multiplexer 20, an SRP generator 21, a key 22, the input of the CRC cyclic checksum generation unit 18 is the first input of the device, an information signal input, the output of the CRC cyclic checksum generation unit 18 is connected to the first input of the adder 19, the output of which is connected to the inputs of the encoding unit 5, the first feedback block 3 ₁ and second feedback block 3 ₂ without feedback, the first and second outputs of the encoding block 5 are connected to the first and second inputs of the multiplexer 20, the output of which is the output of the transmitting device and connected to the input of the channel 9, the outputs of the first block 3 ₁ and the second block 3 ₂ precoding without feedback are connected respectively to the first and second inputs of the key, the third input of which is connected to the output of the generator PSP 21, the input of which is the second input of the transmitting device - control input the synchronization signal, the output of the key 22 is connected to the second input of the adder 19;

the receiving device 10 (Fig.7) contains a decoder sheet 23, N blocks 24 ₁ -24 _N reverse encoding, N keys 25 ₁ -25 _N , N blocks 27 ₁ -27 _N check cyclic checksum and block 28 selection,

moreover, the input of the sheet decoder 23 is the first input of the receiving device - the input of the information signal, and is connected to the output of the channel 9, N outputs of the sheet decoder 23 are connected respectively to the inputs of N blocks 24 ₁ -24 _N reverse coding, the first and second outputs of N blocks 24 ₁ -24 _N reverse coding, connected to the first and second inputs of the corresponding N keys 25 ₁ -25 _N , the third inputs of which are connected to the outputs of the generator PSP 26, the input of which is the second input of the device - the input of the control synchronization signal, the outputs of N keys 25 ₁ -25 _N s are connected respectively to the inputs of N blocks 27 ₁ -27 _N of the cyclic checksum check, the outputs of which are connected respectively to the N inputs of the selection block 28, the output of which is the output of the receiving device.

The inventive method of transmitting voice data in a digital radio communication system is carried out on a device whose structural diagram is made in Fig.6 and 7.

On the transmitting device (Fig. 6), data packets containing an information sequence are received at the input of the CRC checksum generation unit 18, where in each data packet the information sequence is supplemented with a cyclic CRC checksum, for example, encoding the information sequence according to the rule defined in Physical Layer Standard for cdma2000 Spread. Spectrum Systems, Revision D. / CS0002-D. Version 1.0. Date: February 13, 2004. - www.3gpp2.org/Public_tml/ specs / C.S0002-D_vl.0_021704.pdf- p.2-89 [14], using the generating polynomial for the code that detects errors and adds 12 bits to the input data packet, which is g (x) = x ¹² + x ¹¹ + x ¹⁰ + x ⁹ + x ⁸ + x ⁴ + x + 1.

An informational sequence is formed from the information sequence supplemented by a cyclic checksum without introducing code redundancy. The output signal from block 18 is fed to the first input of adder 19. In adder 19, a symbolic summation is performed modulo 2, supplemented by a cyclic checksum of the information sequence with a signal fed to the second input of the adder from the output of the key 22. In this case, the first and second inputs of the key 22 respectively, the signals from the outputs of the first block 3 ₁ and the second block 3 ₂ precoding without feedback are received, and the synchronization control signal from the generator 21 of the SRP is fed to the third input of the key 22. The key 22 provides the output signal of one of the precoding blocks without feedback 3 ₁ or 3 ₂ according to the rule defined by the formula

where g _{0, t} , g _{1, t} are the samples of the signal from the outputs of the precoding blocks without feedback 3 ₁ and 3 ₂ ,

x _t is the reference pseudo-random sequence generated by the method indicated, for example, in [14] with the generating polynomial h (D) = D ¹⁷ + D ¹⁴ +1.

x _t acts as a control signal synchronization key 22.

The SRP generator 21 is triggered by a synchronization control signal, thereby providing symbol-by-symbol synchronization at the receiving and transmitting sides.

An intermediate code sequence from the output of the adder 19 is fed to the input of the coding block 5, the first block 3 ₁ and the second precoding block 3 ₂ without feedback.

In the coding unit 5, a convolutional code trellis diagram is generated, the assigned code symbols are added to the generated intermediate code sequence so that the final node at a given depth q = t _max is known for the convolutional code trellis diagram, and the generated intermediate code sequence is encoded with a non-recursive convolutional code with rational transfer functions g ₀ (D), g ₁ (D), receiving a code word with displayed information symbols. As an example, we consider the generating polynomials defined in [14] for a non-recursive convolutional code with rate R = 1/2, where g ₀ = 561 ₈ , ag ₁ = 753 ₈ .

In the first precoding block 3 _1, a convolutional code trellis diagram is generated without feedback, the assigned code symbols are added to the generated intermediate code sequence so that for the generated convolutional trellis diagram the final node is known at a given depth q = t _max , the generated intermediate code is encoded code sequence with non-recursive convolutional code with rational transfer function

In the second feedback precoding block 3 _2, a convolutional code trellis diagram is formed, the assigned code symbols are added to the generated intermediate code sequence so that for the generated convolutional trellis diagram the final node is known at a given depth q = t _max , the generated intermediate code is encoded code sequence with non-recursive convolutional code with rational transfer function

, зависящей от времени t, где t - целое число и

, g₀(D), g_g1(D) - рациональные функции передачи сверточного нерекурсивного кода, используемого для помехоустойчивого кодирования, а х_t - двоичный символ псевдослучайной последовательности.Summing the signals on the summing unit 19, the information sequence of each packet is encoded with a convolutional recursive code without introducing redundancy with a rational transfer function

depending on time t, where t is an integer and

, g ₀ (D), g _g1 (D) are rational transmission functions of a convolutional non-recursive code used for error-correcting coding, and x _t is a binary symbol of a pseudo-random sequence.

In block 5 encoding carry out encoding with a convolutional code with a coding rate of R = 1/2. Of the two parallel signals at the outputs of the encoding unit 5, by means of the multiplexer 20, a serial signal is formed, which is a code word, which is transmitted through channel 9.

Channel 9 is a collection of transceivers, as well as a radio channel through which the code word is transmitted.

At the receiving device (Fig. 7), an input signal is received over channel 9 and demodulated, receiving a code word in soft decisions. The sequence of soft decisions is fed to the input of the decoder sheet 23, which generates the restored intermediate code sequences of the main and alternative paths as follows:

initialize the Viterbi decoder metric array with a priori metric values, assuming that all paths exit from one node of the trellis diagram of the convolutional code at the initial time, put t = 0;

form a trellis diagram of the convolutional code, where for each node at a depth of q = t + 1 find the predecessor nodes at a depth of q = t, with which it is connected by an allowed transition;

path metrics are calculated for each predecessor node, adding up the metric of the source predecessor node and the metric of the allowable transition;

update the state metric by selecting the smallest metric of the allowed path to this state,

remember the allowed transition with the least metric;

repeat the calculation of path metrics, updating the state metric and remembering the allowed transition with the smallest metric until the specified depth q = t _max is reached;

choose a known final node at a depth q = t _max ;

tracking the main surviving path, performing memorized transitions along the trellis diagram in the opposite direction, thus obtaining an intermediate code sequence from the main surviving path;

performing memorized transitions from the final node in the opposite direction until the next transition corresponds to code symbols, in one of which the i-th information symbol is displayed;

perform the transition along another permitted path in the trellis diagram, different from the transition that was memorized;

from the current node, memorized transitions are performed along the trellis diagram in the opposite direction until the beginning of the trellis diagram is reached, forming an alternative path corresponding to the ith information symbol;

for each information symbol form their own alternative path;

for the main path and each path from the list of alternative paths, an intermediate code sequence is obtained, and intermediate code sequences can be processed both in parallel and sequentially until the cyclic CRC checksum matches in any restored information sequence. In the sequential case, the processing begins with the restored information sequence corresponding to the main surviving path. As an example of execution, a parallel processing variant is given.

, а х_t - двоичный символ псевдослучайной последовательности, синхронизированный с двоичным символом псевдослучайной последовательности на передающей стороне, получая, таким образом, восстановленную информационную последовательность символов, которая поступает с выхода ключей 25₁-25_N на входы блоков 27₁-27_N проверки циклической контрольной суммы CRC.The recovered intermediate code sequences from the outputs of the sheet decoder 23 are supplied to the inputs of N blocks 24 ₁ -24 _N reverse coding, which are functionally similar to coding block 5. The signals from the outputs of N blocks 24 ₁ -24 _N reverse coding are fed to the first and second inputs of N keys 25 ₁ -25 _N , the third inputs of which receive a control synchronization signal from the generator 26 PSP, which is functionally similar to the generator 21 PSP on the transmitting side, which trigger control signal synchronization, providing character-by-character synchronization on the receiving and transmitting sides. By output signals from the generator 26 to the SRP key outputs January ₂₅ -25 _N enters the signal which corresponds to the sequence of code symbols defined by rational transfer function g _y (D), wherein u∈ {0.1} and

, and x _t is the binary symbol of the pseudo-random sequence synchronized with the binary symbol of the pseudo-random sequence on the transmitting side, thereby obtaining a reconstructed informational sequence of symbols that comes from the output of the keys 25 ₁ -25 _N to the inputs of the cyclic check blocks 27 ₁ -27 _N CRC checksum.

Check the recovered information sequences obtained in the cyclic CRC checksum check blocks 27 ₁ -27 _N , for example, adding a check character to the input sequence, which is zero if the check was successful, and one if the cyclic CRC checksum did not match. The recovered and verified information sequences come from the outputs of blocks 27 ₁ -27 _N to the inputs of the selection block 28, in which decoding is completed if the cyclic CRC checksum of any of these sequences coincides. The restored verified information sequence is supplied to the output of block 28, which is the output signal of the receiving device.

If none of the cyclic checksums during the verification match, then the restored information sequence is returned from the output of block 28.

For a better understanding of the claimed invention, let us consider in more detail the operation of some blocks.

или

. Однако предварительное кодирование с обратной связью выполняют путем суммирования по псевдослучайному закону одного из выходов блоков предварительного кодирования без обратной связи, формируя промежуточную кодовую последовательность. Рекурсивное предварительное кодирование с обратной связью характеризуется своей функцией рациональной функцией передачи

, зависящей от времени t, где t - целое g(t, D) число и

, g₀(D), g₁(D) - рациональные функции передачи сверточного нерекурсивного кода, используемого для помехоустойчивого кодирования, а х_t - двоичный символ псевдослучайной последовательности. В конце каждого пакета в линию задержки поступают предопределенные символы, обычной практикой является обнулять линию задержки нулями.On Fig shows the block 3 precoding without feedback and unsystematic encoder 5 convolutional code and the adder 19. The essence of the procedure is as follows. All non-feedback precoding blocks 3 are cascades of convolutional non-recursive coding performed on one delay line. Each of the non-recursive precoding blocks without feedback is characterized by a rational transfer function, which are shown in the figure:

or

. However, feedback precoding is performed by pseudo-random summation of one of the outputs of the feedback-free precoding blocks, forming an intermediate code sequence. Feedback recursive precoding is characterized by its function rational transfer function

depending on time t, where t is an integer g (t, D) number and

, g ₀ (D), g ₁ (D) are rational transmission functions of a convolutional non-recursive code used for error-correcting coding, and x _t is a binary symbol of a pseudo-random sequence. At the end of each packet, predefined characters enter the delay line; it is common practice to zero out the delay line.

Figure 9 presents a simplified sheet of the decoder to the claimed invention, based on the allowed transitions on the trellis diagram of the convolutional code shown in figure 2. The arrows indicate the permissible transitions in the trellis diagram. The paths along the diagram are tracked in the opposite direction opposite the arrows, remembering in each node of the trellis diagram one of the possible transitions that led to this node in accordance with the Viterbi algorithm. However, another transition can be easily reconstructed from the structure of the trellis diagram. Performing this transition instead of the memorized transition, an alternative path is formed. The remaining transitions in this path correspond to the stored paths in the Viterbi decoding process. The bold line shows the main surviving path, the bold dotted line shows alternative paths corresponding to each information symbol. Alternative paths are not shown to the end, the remaining transitions in each of the alternative paths correspond to the stored paths in the Viterbi decoding process.

10, a reverse coding unit 24 is implemented, where it is shown that the information sequence is extracted from the recovered codeword by means of a key, and then the cyclic CRC checksum is checked in the recovered codeword. The reverse encoding unit 24 is identical to the encoding unit 5.

The effectiveness of the claimed invention is confirmed by the results of computer simulations, which are shown in Fig. 11, where information packets with a length of 192 bits are encoded with a convolutional code with a code restriction length of K = 9 and a coding rate of R = 1/2, and they also use a CRC cyclic checksum in 12 bit Shows the energy gain of coding compared to the well-known Viterbi algorithm in a multipath channel with fading defined as Case II, 30 km / h in "Spatial Channel Model AHG (Combined ad-hoc from 3GPP & 3GPP2)" (SCM Text V6.0) April 22 , 2003 // www.3gpp2.org [15]. Unfortunately, it is not possible to make a comparison with the prototype, since no precoding matrices without feedback for such a code are defined. However, it is clear from the prior art that the energy gain of coding a code with a code restriction length of K = 9 compared to a code with K = 5 is 1-2 dB. A comparison with the well-known Viterbi algorithm suggests that the probability of a symbolic error is reduced by an average of 30%.

A comparative analysis confirms the advantage of the claimed invention, which consists in improving the quality of communication.

The claimed invention can be used in digital radio communication systems of the 3rd generation, using convolutional coding and coding according to the Viterbi algorithm.

Claims (2)

1. A method of transmitting voice data in a digital radio communication system, namely, that data packets are formed on the transmitting side, each of which contains an informational sequence of characters, in each data packet, the informational sequence of characters is supplemented with a cyclic CRC checksum, an intermediate code sequence is formed, for what is the encoding of the information sequence of each data packet with the convolutional recursive code b performed by the augmented cyclic checksum? of introducing redundancy with a rational transfer function

depending on time t, where t is an integer and

, g ₀ (D), g ₁ (D) - rational transfer functions of the convolutional non-recursive code used for error-correcting coding, ax _t - binary symbol of the pseudo-random sequence, form a trellis diagram of the non-recursive convolutional code with rational transfer functions g ₀ (D), g ₁ (D), which shows all permissible paths corresponding coded signals are added to the generated intermediate code sequence encoded recursive convolutional code assigned to the code symbols so that for sform Rowan trellis diagram nonrecursive convolutional code with rational transfer functions g ₀ (D), g ₁ (D), which shows all permissible paths corresponding coded signals has been known only final assembly at a predetermined depth q = t _max, encode the generated intermediate code sequence non-recursive convolutional code with rational transfer functions g ₀ (D), g ₁ (D), receiving the code word with the displayed information symbols, transmit the code word on the channel, the input demodulation is performed of one signal, receiving a code word in soft solutions, decode the code word in soft solutions using a Viterbi decoder, for which an array of Viterbi decoder metrics is initialized with a priori metric values, assuming that all paths exit from one node of the trellis diagram of the convolutional code at the initial time, t = 0, form a trellis diagram of the convolutional code, where for each node at a depth q = t + 1 find the predecessor nodes at a depth q = t, with which it is connected by an allowed transition, for each node They calculate path metrics by adding up the metric of the source predecessor node and the metric of the admissible transition, update the state metric by choosing the smallest metric of the allowed path to this state, remember the allowed transition with the smallest metric, repeat the calculation of the path metrics, updating the state metric and remembering the allowed transition with the smallest metric until until it reaches a predetermined depth q = t _max, selected known final node at the depth q = t _max, main track survivor path, performing zap Ratings transitions of trellis diagram in the reverse direction, thus obtaining an intermediate code sequence of primary survivor path encode intermediate code sequence nonrecursive convolutional code with rational functions transfer g ₀ (D), g ₁ (D), yielding the reconstructed codeword carried puncturing in the reconstructed codeword, where from each pair of code symbols corresponding to the transition in the trellis diagram of the convolutional code, a code symbol is left at a depth q = t, determined th rational transfer function g _y (D), wherein u∈ {0.1} and

, and x _{t is the} binary symbol of the pseudo-random sequence synchronized with the binary symbol of the pseudo-random sequence on the transmitting side, receiving the restored information sequence, and the cyclic CRC checksum is checked, if it matches, the decoding is completed, if the cyclic CRC checksum does not match, a list of alternative paths is generated, for which carry out the memorized transitions from the final node in the opposite direction until the next transition matches Create code symbols, in one of which the ith information symbol is displayed, perform a transition along a different permitted path in the trellis diagram, different from the transition that was memorized; from the current node, remember the transitions along the trellis diagram in the opposite direction until the trellis begins diagrams, forming an alternative path corresponding to the i-th information symbol, form an alternative path for each information symbol, for each path from the list of alternative paths the field chayut intermediate code sequence encoded intermediate code sequence nonrecursive convolutional code with rational functions transfer g ₀ (D), g ₁ (D), is carried puncturing in the reconstructed codeword, wherein each pair of code symbols corresponding to the transition in the trellis diagram of the convolutional code, at a depth q = t, a code symbol is determined by the rational transfer function g _y (D), where y∈ {0,1} and

, and x _{t is the} binary symbol of the pseudo-random sequence synchronized with the binary symbol of the pseudo-random sequence on the transmitting side, receiving the reconstructed information sequence of symbols, and the cyclic CRC checksum is checked, if it matches, the decoding is completed, the information sequence of symbols corresponding to the main surviving path is returned if no CRC cyclic checksum check matched. 2. A device for transmitting voice data in a digital radio communication system, comprising a transmitting and receiving device connected by a channel providing data transmission, the transmitting device comprising a cyclic checksum generating unit CRC, an adder, an encoding unit, a first and second non-feedback precoding unit , a multiplexer, a pseudo-random sequence generator PSP and a key, while the input of the cyclic checksum generating unit CRC is the first input of the device tva - by the input of an information signal, the output of the cyclic checksum generating unit CRC is connected to the first input of the adder, the output of which is connected to the inputs of the encoding unit and the first and second precoding units without feedback, the first and second outputs of the encoding unit are connected respectively to the first and second inputs multiplexer, the output of which is the output of the transmitting device and is connected to the channel input, the outputs of the first and second precoding blocks without feedback are connected are connected respectively with the first and second inputs of the key, the third input of which is connected to the output of the PSP generator, the input of which is the second input of the transmitting device - the input of the control synchronization signal, the output of the key is connected to the second input of the adder, the receiving device contains a decoder sheet based on the decoder Viterbi and the list of the most likely variants of the information sequence, supplemented by a cyclic checksum, N blocks of reverse coding, N keys, N blocks of checking cycles a checksum and a selection block, the input of the sheet decoder being the first input of the receiving device as the input of the information signal and connected to the output of the channel, N outputs of the sheet decoder are connected respectively to the inputs of N blocks of reverse coding, the first and second outputs of N blocks of reverse coding respectively, connected to the first and second inputs of N keys, the third inputs of which are connected to the outputs of the PSP generator, the input of which is the second input of the receiving device - the input of the synchronization control signal, the outputs of the N keys are connected respectively to the inputs of the N blocks of cyclic checksum verification, the outputs of which are connected respectively to the N inputs of the selection block, the output of which is the output of the receiving device.

Images