RU2197790C2 - Method for blind determination of data burst transmission speed - Google Patents

Abstract

FIELD: radio communication facilities; cellular radio communication systems such as code-division ones. SUBSTANCE: proposed method that may be found useful for forward and backward channels of SD MA 2000 system involves step-by-step decoding of respective speeds, and termination of decoding when CRC cyclic check sums coincide in reception and condition of symbolic error probability below SER <T threshold is satisfied. Main goal of subsequent conclusions is finding speed decoding procedure in which mean computational complexity is minimal. Speed decoding procedure may be chosen by various methods such as sequential checking of hypotheses starting from minimum and up to maximum. As an alternative of choosing decoding procedure hypotheses may be checked using current data burst and starting from decision taken on speed of preceding data burst. Hypotheses may be checked using procedure ensuring minimal sum of products of transmission speed by respective a posterior probability. EFFECT: facilitated procedure, reduced processing time and power requirement at high precision of speed identification. 3 cl, 3 dwg

Description

Technical field
The invention relates to radio communications, and in particular to devices for determining the transmission speed of data packets in cellular radio communication systems, for example, code-division multiplexing, and can find application both in the forward and reverse channels of the CDMA 2000 system.

State of the art
As you know, like the IS-95 forward channel, the main CDMA 2000 forward and reverse channels provide four data rates from two sets:
first set (Set1): 9600, 4800, 2400, 1200 bit / s;
second set (Set2): 14400, 7200, 3600, 1800 bit / s.

 In the main channel, convolutional coding of various redundancies is used in all ranges of transmission rates. Moreover, the code length in all cases is 9.

 The repetition of bits for speed correction is performed both in the forward and reverse channels at data transfer rates lower than the maximum. At the same time, when the data transfer rate is lower than the maximum used, the signal power decreases. This leads to a significant reduction in interference of multiple access and, consequently, to increase the throughput of the communication system. However, at the receiving end, the data rate is unknown, therefore, it is necessary to use a device to determine it.

 A known method for blindly determining a data rate using correlation between repeating bits at data rates below the maximum described in E. Cohen and Lou "Multirate Detection for the IS-95 CDMA forward traffic channels", IEEE GLOBECOM 95, November 1995 [1] and E. Cohen and H. Lou "Multirate Detection for the IS-95A CDMA forward traffic channels using 13 kbps speech coder", IEEE GLOBECOM 96, November 1996 [2].

 The term "blind" is a translation from English of the generally accepted concept of "Blind Data Rate Detection", which means a method for determining the speed without using any information, indicator or feature specially designed for this. Such a multi-speed detector uses an algorithm for calculating three critical values, consistent with possible speeds, and then comparing with thresholds.

 In this case, the method of calculating the critical values is simple in terms of the volume of calculations - without multiplication, but using only the sum of the absolute value, however, the method of comparing the critical variables with the threshold system is very complicated and not optimal, which also reduces the reliability of decision making.

 A known method described in patent WO 9508888 "Multirate serial decoding method for code division multiple access mobile", QUALCOMM Inc. [3].

 This method serves to provide automatic decoding of each frame with any unknown information transfer rate.

 A serial decoder (Viterbi decoder) performs automatic decoding for each of the many modes by performing multiple passes to decode each frame with each of all possible specified information rates and provides an initial information rate.

 This method allows you to accurately determine the speed of the transmitted data, however, four times (in the case of CDMA 2000) the use of a Viterbi decoder significantly complicates the device for implementing this method.

Closest to the technical nature of the proposed is a method for determining the transmission rate of information described in patent US 5774496 "Method and apparatus for determining data rate of transmitted variable rate in a communications receiver", QUALCOMM Inc., Jun. 30, 1998 [4]. The method for determining the data rate of the received signal described in the said patent is also based on four times decoding and analysis of the cyclic checksum at the reception. The term “cyclic redundancy check” is a sequence of bits delivered to a transmitted data packet at the transmitting end of a communication line and designed to verify the quality of received data at the receiving end. This sequence is obtained on the basis of all transmitted information bits of the packet. At the receiving end of the communication, the calculation of this sequence is also performed, but based on the received information bits. The discrepancy between the transmitted checksum and the newly calculated at the receiving end of the communication line indicates that either the information bits or the checksum were received incorrectly. The method described in [4] is as follows:
- decode and re-encode at a first data rate a packet of the specified received signal to form a first estimate of the received signal and the first cyclic checksum (CRC) at the reception;
- compare the first estimate of the received signal with the received signal and calculate the first number of errors (SER), while an error occurs when the specified received signal does not match the specified first signal estimate, the specified number of errors (SER) and the first quality indicator - CRC at reception determine the first error metric;
- process the specified received signal to generate a second received signal representing the second data rate;
- decode and re-encode at a second data rate of the specified received signal to form a second estimate of the received signal and the second CRC at the reception;
- comparing the second estimate of the received signal with the received signal and counting the second SER, an error occurs when the specified received signal does not match the specified second signal estimate, the specified number of SER and the second quality indicator determine the second error metric;
- process the specified received signal to generate a third received signal representing a third data rate;
- decode and re-encode at a third data rate of the specified received signal to form a third estimate of the received signal and the third CRC at the reception;
- comparing the third estimate of the received signal with the received signal and counting the third quantity of SER, an error occurs when the specified received signal does not match the specified third signal estimate, the specified number of SER and the third CRC at the reception determine the third error metric;
- process the specified received signal to generate a fourth received signal representing the fourth data rate;
- decode and re-encode at the fourth data rate of the specified received signal to form the fourth assessment of the received signal and the fourth CRC at the reception;
- compare the fourth estimate of the received signal with the received signal and calculate the fourth number of errors (SER), while an error occurs when the specified received signal does not match the specified fourth signal estimate, the specified number of errors (SER) and the fourth quality indicator is CRC at reception determining a fourth error metric;
- predict the data rate of the specified received signal based on a comparison of each of the indicated error metrics;
- when processing the received signal in the repetition mode (forward channel IS-95), 2, 4 or 8 samples are summed, in DBR mode (reverse channel IS-95) certain samples are selected.

 The disadvantage of this method is the use of four-fold Viterbi decoding, which complicates the device that implements it, increases the processing time and power consumption.

SUMMARY OF THE INVENTION
The task that the claimed invention solves is to simplify the method for blindly determining the transmission speed of a data packet at the receiving end of a communication line (hereinafter “reception”), reducing the average processing time and power consumption while maintaining high accuracy of speed discrimination.

To solve this problem, a method for blindly determining the data transfer rate, which consists in decoding the received signal in accordance with the hypothesis of the data transfer rate, compute a cyclic checksum at the reception, and perform reverse encoding (in this case, “reverse encoding” assumes that all decoded information bits, as well as bits of the cyclic checksum, are again encoded in order to obtain an estimate of the received code symbols) and form an estimate of the probability of a symbol error, they reduce the probabilities of a symbolic error with a threshold, additionally introduce a new sequence of operations: decoding is stopped and a decision is made on the data transfer rate if, when testing another hypothesis, the probability of a symbolic error is less than a threshold and the cyclic checksum at the reception coincides, and the order of testing hypotheses is chosen in this way to minimize the average computational cost, if after checking all hypotheses the speed is not determined, then it is determined from the conditions:
- if the cyclic checksum at the reception matches only one hypothesis, then the decision on the data transfer rate is made in favor of this hypothesis;
- if the cyclic checksums at the reception coincide for several hypotheses at the same time, then the decision on the data transfer rate is made in favor of the one that provides the minimum probability of a symbolic error;
- if the cyclic checksum in the example does not coincide for any hypothesis, then the decision on the data transfer rate is made in favor of the hypothesis that provides the minimum probability of a symbolic error.

 In addition, hypothesis testing is carried out sequentially, starting from the minimum to the maximum.

 Hypothesis testing on the current package can begin with a decision about the speed of the previous package.

 Another option for testing hypotheses is that hypothesis testing is performed in such a way that the minimum sum of the products of the transmission rate and the corresponding posterior probability is provided.

The proposed method is as follows:
1) decode the received signal one by one in accordance with the hypothesis of the data rate until the speed is correctly determined;
2) calculate the cyclic checksum (CRC) at the reception;
3) perform reverse coding and form an estimate of the probability of a symbolic error (SER);
4) compare the probability of a symbolic error with a threshold;
5) taking into account the verification of the cyclic checksum (CRC) at the reception, the data rate is determined;
6) the decoding is stopped and a decision is made on the data transfer rate if, when testing another hypothesis, the probability of a symbolic error is less than a threshold and the cyclic checksum at the reception coincides;
7) the procedure for testing hypotheses is chosen in such a way as to minimize the average computational cost, if after checking all the hypotheses the speed is not determined, then it is determined from the conditions:
- if the cyclic checksum at the reception matches only one hypothesis, then the decision on the data transfer rate is made in favor of this hypothesis;
- if the cyclic checksums at the reception coincide for several hypotheses at the same time, then the decision on the data transfer rate is made in favor of the one that provides the minimum probability of a symbolic error;
- if the cyclic checksum at the reception does not coincide for any hypothesis, then the decision on the data transfer rate is made in favor of the hypothesis that provides the minimum probability of a symbolic error.

Hypothesis testing can be carried out in various ways:
- hypothesis testing is carried out sequentially, starting from the minimum to the maximum;
- hypothesis testing on the current package begins with the decision on the speed of the previous package;
- alternatively, hypothesis testing is performed in such a manner that provides a minimum sum of products of the transmission rate by the corresponding posterior probability.

List of drawings
Graphic materials illustrating the claimed invention:
FIG. 1 is a block diagram for implementing the proposed method for determining the data transfer rate;
figure 2 is an example of a block for determining posterior probability;
FIG. 3 - dependence of the probability of packet error on the value of the ratio of bit energy to spectral noise density;
4 is the computational complexity of decoding.

Preferred Embodiment
To implement the proposed method uses the device shown in figure 2, where it is indicated:
1 - memory block,
2 - averaging unit,
3 - Viterbi decoder,
4 - block calculation of a cyclic checksum (CRC) at the reception,
5 - encoder
6 - comparison unit,
7 - counter
8 - decision block,
9 is a block for determining posterior probability,
10 - control unit.

 The device contains a series-connected memory block 1, the input of which is the input of the device, an averaging block 2, a Viterbi decoder 3, an encoder 5, a comparison block 6, a counter 7, a decision block 8, the output of which is the output of the device, a priori probability determination block 9, block control 10. The first output of the control unit 10 is connected to the second input of the Viterbi decoder 3, the second output of the control unit 10 is connected to the second input of the encoder 5 and the third output of the control unit 10 is connected to the second input of the CRC calculation unit 4 at the reception. The output of the averaging unit 2 is also connected to the second input of the comparison unit 6. The output of the Viterbi decoder 3 is connected to the input of the CRC calculation unit 4 at the reception, the output of which is connected to the second input of the decision unit 8.

 The device operates as follows.

 The samples of the received signal are stored in the memory unit 1. Since the information transfer rate for each received signal is different, the decoder input signal is prepared by averaging one, two, four, and eight samples in accordance with the hypothesis about the speed of the data packet using the averaging unit 2. Decode the samples of the received signal in the Viterbi decoder 3 and for each received signal, calculate the cyclic checksum at the reception in block 4 of the CRC calculation. The signal after decoding is backcoded in encoder 5 to form an estimate of the received signal. In comparison block 6, the received signal is compared with a threshold (the threshold being determined by the type of encoder) and then counter 7 calculates the number of errors, forming the probability of a symbol error (SER).

 The comparison results are transmitted to decision block 8, where they determine the speed of the received information by comparing the probability of a symbolic error with a threshold and taking into account the results of checking the cyclic checksum at the reception. For each stage of determining the speed, the decision block 8 considers the following decision options. If the probability of a symbolic error is less than a threshold and the cyclic checksum at the reception coincides, then decoding is stopped and this hypothesis is considered true. In this case, the procedure for testing hypotheses is chosen in such a way as to minimize the average computational cost.

 If, after checking all the hypotheses, the speed is not determined, then it is determined from the following conditions. If the reception CRC matches only one hypothesis, then the decision about the data packet transmission rate is made in favor of this hypothesis. If the reception CRCs coincide for several hypotheses at the same time, then the decision on the data packet transmission rate is made in favor of the one that provides the minimum probability of a symbolic error. If the CRC at the reception does not coincide for any hypothesis, then the decision on the transmission rate of the data packet is selected from the hypothesis that provides the minimum probability of a symbolic error.

 To determine the decoding order, a posterior probability determination block 9 is used, an example of which is shown in FIG. 2. At the initial time, all the counters 13 are set to state 1. When 1 arrives at the logic device 11 from the decision block 8, the logic device 11 supplies 1 to the input of the channel corresponding to a certain speed, and to all other channels - 0. The result of the counter accumulation 13 is proportional to the posterior probability of speed. In this case, when a signal arrives at the upper input of the counter 13, the accumulation of the counter increases by 1, when a signal arrives at the lower input of the counter 13, it decreases by 1.

 The control unit 10 controls the operation of the device.

 We prove the achievement of the goal.

In the framework of the generally accepted approach, Viterbi decoding is performed four times. If we take the computational complexity of decoding a data packet of maximum speed as 1, then, given the decrease in packet length at lower speeds, the complexity of the conventional blind speed detector is
1 + 0.5 + 0.25 + 0.125 = 1.875, (1)
where cost = [1, 0.5, 0.25, 0.125] is the computational complexity of decoding the corresponding speeds.

Вероятность того, что декодирование дойдет до соответствующего шага, равна:
Ps₁= 1, Ps₂= 1-Pb_orderi,1, Ps₃=(1-Pb_orderi,1)•(1-Pb_orderi,2), Ps₄=(1-Pb_orderi,1)•(1-Pb_orderi,2)•(1-Pb_orderi,3).It is proposed to perform decoding of the corresponding speeds in turn, and if the CRC coincides and the SER <T condition is satisfied, stop decoding. The main goal of further conclusions is to find such a decoding order of speeds at which the average computational complexity will be minimal. When using a four-speed vocoder, 4! = 24 different orders are possible (ie, order ₁ = [1 2 3 4], order ₂ = [1 2 4 3], etc.). In other words, it is necessary to find the optimal procedure for testing speed hypotheses, and it should be noted that the ith hypothesis has a certain probability p _i and its verification requires computational costs cost _i . Let p ₁ , p ₂ , p ₃ and p ₄ be the probabilities of the corresponding rates and order _{i the} ith decoding order (i∈ [1, 24]), then the probability that decoding will be stopped at the jth step (if decoding reached the j-th step) is equal to

The probability that decoding will reach the corresponding step is:
Ps ₁ = 1, Ps ₂ = 1-Pb _{orderi, 1} , Ps ₃ = (1-Pb _{orderi, 1} ) • (1-Pb _{orderi, 2} ), Ps ₄ = (1-Pb _{orderi, 1} ) • (1 -Pb _{orderi, 2} ) • (1-Pb _{orderi, 3} ).

Оптимальная очередность декодирования будет соответствовать порядку с минимальной средней стоимостью. Для примера рассмотрим случай равновероятных гипотез, т. е. p₁= p₂= p₃=p₄=0.25. Минимальная стоимость, соответствующая, очевидно, порядку декодирования [4 3 2 1], равна С=0.8124. Если же вероятности гипотез равны р=[0.4 0.1 0. 0.4], то оптимальным является порядок декодирования [4 1 3 2] со стоимостью С=0.88.The average computational cost of decoding in accordance with the i-th order is

The optimal decoding order will correspond to the order with the lowest average cost. As an example, we consider the case of equally probable hypotheses, i.e., p ₁ = p ₂ = p ₃ = p ₄ = 0.25. The minimum cost, obviously corresponding to the decoding order [4 3 2 1], is C = 0.8124. If the probabilities of the hypotheses are equal to p = [0.4 0.1 0. 0.4], then the decoding order [4 1 3 2] with a cost of C = 0.88 is optimal.

 It should be noted that in the above reasoning, the result of testing hypotheses was assumed to be unmistakable. In real conditions, the average cost will be slightly higher.

 The simulation results of the decoder with the proposed determinant of speed in the channel with white Gaussian noise are shown in figure 3 and figure 4. Parameters for an a priori known data rate are also given there for comparison.

 As follows from the simulation results, the decoding characteristics in both cases are almost the same, while at the level of packet error of interest 1%, the computational complexity of the proposed method, on average, is 0.9, which is 2.0 times lower than the complexity of the generally accepted approach .

 A comparative analysis of the method for blindly determining the transmission speed of a data packet with a prototype shows that the present invention is significantly different from the prototype, since the successive Viterbi decoding can significantly reduce processing time and power consumption while maintaining high accuracy of speed discrimination.

Claims (3)

 1. The method of blindly determining the data transfer rate, which consists in decoding the received signal in accordance with the hypothesis of the data transfer speed, calculating the cyclic checksum at the reception, performing reverse encoding and estimating the probability of a symbol error, comparing the probabilities of a symbol error with a threshold characterized in that the decoding is stopped and a decision is made on the data transfer rate if, when testing the next hypothesis, the probability of a symbolic error is less than a threshold and the cyclic checksum at the reception coincides, while the procedure for testing hypotheses is chosen so as to minimize the average computational cost, if after checking all the hypotheses the speed is not determined, then it is determined from the conditions: if the cyclic checksum at the reception matches only one hypothesis, then the decision on the data transfer rate is made in favor of this hypothesis, if the cyclic checksums on reception coincide for several hypotheses at the same time, then the decision on the data transfer rate is made t in favor of the one that provides the minimum probability of symbolic error, if the cyclic checksum at the reception does not coincide for any hypothesis, then the decision on the data transfer rate is made in favor of the hypothesis that provides the minimum probability of symbolic error.  2. The method according to p. 1, characterized in that the hypothesis test is carried out sequentially, starting from the minimum to the maximum.  3. The method according to p. 1, characterized in that the hypothesis testing on the current package begins with the decision on the speed of the previous package.

Images