Blind Identification of Convolutional Encoder Parameters

Abstract

This paper gives a solution to the blind parameter identification of a convolutional encoder. The problem can be addressed in the context of the noncooperative communications or adaptive coding and modulations (ACM) for cognitive radio networks. We consider an intelligent communication receiver which can blindly recognize the coding parameters of the received data stream. The only knowledge is that the stream is encoded using binary convolutional codes, while the coding parameters are unknown. Some previous literatures have significant contributions for the recognition of convolutional encoder parameters in hard-decision situations. However, soft-decision systems are applied more and more as the improvement of signal processing techniques. In this paper we propose a method to utilize the soft information to improve the recognition performances in soft-decision communication systems. Besides, we propose a new recognition method based on correlation attack to meet low signal-to-noise ratio situations. Finally we give the simulation results to show the efficiency of the proposed methods.

1. Introduction

In digital communication systems, error-correction codes are widely used. To meet high quality of services, new coding schemes are being developed ceaselessly. Therefore, for a communication receiver, it is very difficult to remain compatible with all standards used. But if it is an intelligent receiver, which is able to blindly recognize the coding parameters of a specific transmission context, it can adapt itself to the perpetual evolution of digital communications. Furthermore, the blind recognition techniques are also applied in noncooperative communications. In noncooperative communication contexts, a receiver does not know the coding parameters, so it must blindly recover the encoder before decoding. In this paper we focus on the blind recognition of coding parameters of an encoder which uses convolutional codes as error-correction coding and propose a method to take advantage of the soft information in soft-decision situations.

Some previous literatures discussed the problem of blind recognition of convolutional codes. The authors of [1–3] developed recognition methods in noiseless context, including the rate 1/𝑛 and 𝑘/𝑛 codes, where 𝑛 denotes the codeword length and 𝑘 the dimension. These methods are not suitable for noisy environment. For the case of noisy context, some algorithms were proposed in recent years [4–7]. The algorithm proposed in [7], which we call a Gauss-Jordan elimination through pivoting (GJETP) based algorithm in this paper, completely solved the blind parameter recognition of 𝑘/𝑛 convolutional codes with low complexity and excellent recognition performances.

However, the previous works are all discussed in hard-decision situations. In modern communication systems, more and more soft-decision based algorithms are applied to improve the signal processing performances. For example, the soft-decision based decoding methods always have better performances than hard-decision situations [8–12]. Similarly, some soft-decision based blind recognition of block code parameters also outperforms the hard-decision one [13, 14].

In Section 2 of this paper, based on [6, 7] we propose a method to utilize the soft information to improve the recognition performances in soft-decision systems. And we give another two improvements about the recognition algorithm in Sections 3 and 4: (1) because the authors of [6, 7] did not give a normative algorithm to automatically identify the parameter 𝑛 for a computer program, we propose a formalized scheme to optimize the recognition of 𝑛; (2) the authors of [6, 7] did not consider the synchronization of codewords while the synchronization positions are usually unknown in a noncooperative communication context; we propose to modify the algorithm to identify the codeword synchronization.

Besides, in Section 5 we propose a correlation attack method to meet very low signal-to-noise ratio (SNR) situations, which also includes both hard- and soft-decision algorithms. This method might require more computational time, but it can push the SNR limits of the GJETP-based algorithm proposed in [7]. And we propose some strategies to reduce the searching space according to the structural analysis of the dual codes.

Finally we show the efficiency of the proposed algorithm by computer simulations in Section 6 and conclude the paper in Section 7.

2. Utilizing the Soft Information

The details of the GJETP-based recognition of rate 𝑘/𝑛 convolutional encoder parameters are introduced in [7]. The algorithm includes three major sections:(1)identification of code length 𝑛;(2)identification of a dual code basis;(3)identification of the generator matrix.

In the first procedure, the authors of [7] propose to recognize parameter 𝑛 using the following steps.

Step 1. Set the minimum and maximum value of 𝑙, that is, 𝑙_min and 𝑙_max. Initialize 𝑙 to be 𝑙_min.

Step 2. According to 𝑙 and a sequence of received symbols, we generate an observed data matrix 𝐑_𝑙 as shown in Figure 1, the numbers in which denote the arriving order of the bits in the received data stream.

Figure 1 

Generation of the observed matrix 𝐑𝑙.

Step 3. Transform the matrix 𝐑_𝑙 to a lower triangular matrix 𝐆_𝑙 by GJETP (see [15] for details): 𝐆_𝑙=𝐀_𝑙𝐑_𝑙𝐁_𝑙. (1)

Step 4. Obtain the set 𝐙_𝑙 as follows: 𝐙_𝑙=Card𝑖∈{1,…,𝑙}∣𝑁_𝑙(𝑖)≤𝐿−𝑙2𝛾_opt, (2) where 𝑁_𝑙(𝑖) is the number of “1” in the lower part of the 𝑖th column in the matrix 𝐆_𝑙,𝛾_opt is an optimal threshold [6, 7], and Card{𝑥} denotes the cardinal of 𝑥. An 𝑁_𝑙(𝑖) smaller than (𝐿−𝑙)𝛾_opt/2 indicates that the 𝑖th column of 𝐑_𝑙 has a high probability to be dependent on the other columns.

Step 5. If 𝑙=𝑙_max, execute Step 6. If 𝑙<𝑙_max, let 𝑙=𝑙+1 and go back to Step 2.

Step 6. Output the gap between two consecutive nonzero cardinals, 𝐙_𝑙, as the estimated codeword size 𝑛. The principle of this estimation is that when 𝑙=𝑚𝑛(𝑚∈𝐙⁺) and 𝑙 is larger than a parameter 𝑛_𝑎 (see [7] for details), some columns of 𝐂_𝑙 can be determined by other columns, where 𝐂_𝑙=𝐑_𝑙+𝐄_𝑙 and 𝐄_𝑙 is the error pattern corresponding to 𝐑_𝑙.

According to the GJETP algorithm, the reliability of upper part of 𝐑_𝑙 has larger influence on the successful detection of dependent columns in 𝐑_𝑙. So if we can make the upper part of 𝐑_𝑙 have lower errors, the algorithm can be improved. In hard-decision situations, we cannot determine which rows in 𝐑_𝑙 have lower probabilities to have error decision bits. But in soft-decision situations, we can examine the reliabilities of the rows in 𝐑_𝑙 according to the soft-decisions bits. Therefore, here we propose the following processing procedure inserted into the steps (between Step 2 and Step 3) mentioned above.(1)Fill the observed matrix 𝐑_𝑙 with the soft-decision bits. For each row in 𝐑_𝑙 we find the decision bit that has the lowest reliability and record its reliability value as the reliability of the row.(2)Arrange the rows of 𝐑_𝑙 according to each row’s lowest reliability value to make the first row of 𝐑_𝑙 have the highest reliability, the second row have the second highest reliability, and so on.(3)Obtain the hard decisions of the bits in the rearranged 𝐑_𝑙 and continue Step 3 mentioned above.

After this processing, the upper rows of the rearranged 𝐑_𝑙 have higher reliabilities and therefore have lower probabilities to include error decision bits. So the probability of successful detection of dependent columns can be improved. And as the value of 𝐿 rises, the improvement of the reliabilities of the upper part of 𝐑_𝑙 becomes more effective, so the recognition performance can be improved more obviously than the previous algorithm in [7]. After the estimation of 𝑛, we use Algorithms  1 and   2 in [7] to identify other coding parameters. Note that the parity check vector recognition procedure uses the GJETP algorithm too, so the proposed rearranged steps are also applied before each GJETP processing when recognizing the parity check vectors.

3. Formalizing the Estimation Algorithm of 𝑛

Note that, in Step 6 shown in the previous section, the authors of [7] only give the idea of the estimation of 𝑛. When the noise level is high, there exists a problem that the number of error decision bits is high and some cardinals 𝐙_𝑙 corresponding to the values of 𝑙 that equals 𝑚𝑛 are blank. So not all the gaps between two consecutive nonzero cardinals, 𝐙_𝑙, equal 𝑛. In this case, we need a formalized algorithm to estimate the value of 𝑛 more exactly. This can be done by simply searching all the gaps between two consecutive nonzero cardinals and find out which gap value appears mostly. The detailed algorithm steps are listed below.(1)Let𝐕_𝑙=[𝑙₁,𝑙₂,…,𝑙_𝑠] be the vector that consists of the values of 𝑙 corresponding to the detected nonzero cardinals 𝐙_𝑙 in Step 6.(2)Calculate the vector Δ𝐕_𝑙=[Δ𝑙₁,Δ𝑙₂,…,Δ𝑙_𝑠−1], the elements of which are calculated by Δ𝑙_𝑖(1≤𝑖≤𝑠−1)=𝑙_𝑖+1−𝑙_𝑖. (3)(3)Initialize a vector 𝐐 with length 𝑙_max to be overall zeros and for each value of 𝑖 from 1 to 𝑠−1, we let 𝐐Δ𝑙_𝑖=𝐐Δ𝑙_𝑖+1. (4)(4)Finally we find the maximum element in 𝐐 and output the corresponding gap valueΔ𝑙_𝑖 to be the estimated codeword size 𝑛.

After taking the previously mentioned searching method we can find the most probable gap between two consecutive nonzero cardinals, that is, the estimation of the code length. An example as follows further describes the searching procedure.

Figure 2 shows the recorded vector 𝐍_𝑙 when recognizing a 𝐶(3,2,4) convolutional code based on correct synchronization positions with a low SNR (𝐸_𝑠/𝑁₀=4 dB). From the figure we can see that not all the gaps between two consecutive nonzero elements in 𝐍_𝑙 equal the code length, 3. If we estimate the code length by the gap between the first and the second nonzero elements in 𝐍_𝑙, the estimation is not correct. Implementing the searching steps mentioned above, we can firstly obtain the vector 𝐕_𝑙 as follows: 𝐕_𝑙=21,30,33,39,42,45,48. (5)

Figure 2 

The recorded vector 𝐍𝑙 based on low SNR.

Then we can calculate the vector Δ𝐕_𝑙 according to 𝐕_𝑙 as follows: Δ𝐕_𝑙=936333. (6)

Furthermore, we can calculate the vector 𝐐 by (4). In the vector 𝐐, we have 𝐐(3)=4, 𝐐(6)=𝐐(9)=1, and other elements in 𝐐 equal to zero. 𝐐(3) is the maximum element in the vector 𝐐, so, according to Step 4 mentioned above, we can estimate the code length:𝑛=3.

This example shows an implementation of the formalized algorithm proposed in this section to estimate the code length more exactly in a low SNR situation.

4. Recognition of Synchronization Positions

The GJETP-based dual code method proposed in [7] has good performances on convolutional encoder identifications. But unfortunately, the authors did not consider the codeword synchronization problem. In noncooperative context, the synchronization cannot usually be reached before the coding parameters are correctly estimated. The algorithm is discussed based on correct synchronization. In the case that the codeword synchronization position is unknown, we can but randomly choose a position in the received data stream to be the beginning of a codeword. The randomly chosen synchronization positions have low influence on the recognition of code length, but the recognition of key parameter 𝑛_𝑎 (the minimal length of the rows of matrix 𝐑_𝑙 so that 𝐑_𝑙 includes dependent columns; see [7] for details) and parity check vectors are not correct.

Figure 3 shows the recorded vector 𝐍_𝑙 when recognizing a 𝐶(3,2,4) convolutional code based on correct synchronization positions. From the figure we can estimate the key parameter 𝑛_𝑎=21, which is accordant to the 𝐶(3,2,4) convolutional code. But if we choose an incorrect synchronization position, as shown in Figure 4, the first nonzero element in vector 𝐍_𝑙 appears at the position 𝑙=24, so the key parameter 𝑛_𝑎 is estimated to be 24, which is a fault estimation. In order to obtain correct parity check vectors, we must firstly locate the correct synchronization positions.

Figure 3 

The recorded vector 𝐍𝑙 based on correct synchronization.

Figure 4 

The recorded vector 𝐍𝑙 based on incorrect synchronization.

We assume that when recognizing the code length we fill the matrix 𝐑_𝑙 from the 𝑡₀th position from the received bit stream and the estimated code length and key parameter 𝑛_𝑎 are 𝑛 and 𝑛_𝑎, respectively, and then we propose the following steps  to find the correct codeword synchronization.

Step 1. Let 𝜏=𝑡₀+1, 𝑛_𝑎=𝑛_𝑎, and 𝜏=𝑡₀.

Step 2. Let 𝑙=𝑛_𝑎−𝑛.

Step 3. Fill the matrix 𝐑_𝑙 with row length 𝑙 from the 𝜏th bit in the received data stream.

Step 4. Do the GJETP processing for 𝐑_𝑙 and calculate  𝐍_𝑙, the number of dependent columns.

Step 5. If 𝐍_𝑙>0 and 𝑙<𝑛_𝑎, let 𝑛_𝑎=𝑙, 𝜏=𝜏, and 𝑙=𝑙−𝑛, and go back to Step 3; otherwise, execute Step 6.

Step 6. If 𝜏<𝑡₀+𝑛−1, let 𝜏=𝜏+1 and go back to Step 2.

Step 7. Let 𝑛_𝑎 and 𝜏 be the newly estimations of 𝑛_𝑎 and synchronization position.

The previous steps  corrected the codeword synchronization. For the following procedure of parity check matrix recognition, we use 𝜏 to be the synchronization position to fill the matrix 𝐑_𝑙 and replace 𝑛_𝑎 by 𝑛_𝑎 in the algorithms.

5. Correlation Attack Method

5.1. Hard-Decision Situations

If the polynomial-based generator matrices of a 𝑘/𝑛 convolutional code and the dual code are 𝐆(𝐷) and 𝐇(𝐷), respectively, we have [16] 𝐆(𝐷)×𝐇(𝐷)=0. (7)

According to the analysis of [7], 𝐇(𝐷) has the following style: 𝐇(𝐷)=ℎ_1,1(𝐷)⋯ℎ_1,𝑘(𝐷)ℎ₀(𝐷)ℎ⋮⋱⋮⋱_{𝑛−𝑘,1}(𝐷)⋯ℎ_{𝑛−𝑘,𝑘}(𝐷)ℎ₀(𝐷), (8) where ℎ_𝑗,𝑠(𝐷)(1≤𝑗≤𝑛−𝑘,1≤𝑠≤𝑘) is a polynomial, coefficients of which are all in GF(2): ℎ_𝑗,𝑠(𝐷)=ℎ_𝑗,𝑠(0)+ℎ_𝑗,𝑠(1)𝐷+⋯+ℎ_𝑗,𝑠𝜇^⟂𝐷^𝜇⟂. (9) Parameter 𝜇^⟂ in (7) is the memory of the dual code [6, 7].

ℎ₀(𝐷) is also a polynomial on GF(2) as follows: ℎ₀(𝐷)=ℎ₀(0)+ℎ₀(1)𝐷+⋯+ℎ₀𝜇^⟂𝐷^𝜇⟂. (10) And according to [7], for a realizable convolutional encoder, we have ℎ₀(0)=1. (11)

Based on (8)–(10), we can define the binary form of 𝐇(𝐷) as follows: 𝐇𝐇=^𝜇⟂𝐇_𝜇⟂−1⋯𝐇₀𝐇_𝜇⟂𝐇_𝜇⟂−1⋯𝐇₀𝐇_𝜇⟂𝐇_𝜇⟂−1⋯𝐇₀,⋱⋱⋱⋱ (12) where 𝐇_𝑖=ℎ_1,1(𝑖)⋯ℎ_1,𝑘(𝑖)ℎ₀(𝑖)ℎ⋮⋱⋮⋱_{𝑛−𝑘,1}(𝑖)⋯ℎ_{𝑛−𝑘,𝑘}(𝑖)ℎ₀(𝑖). (13)

It is shown in (12) that each row of the matrix 𝐇 is a shift of the following (𝑛−𝑘)×𝑛(𝜇^⟂+1) matrix 𝐇₀, which we call the basic check matrix of convolutional codes in this paper: 𝐇₀=𝐇_𝜇⟂𝐇_𝜇⟂−1⋯𝐇₀. (14)

So the recognition problem of a convolutional encoder is equivalent to the recognition of 𝐇₀. We call the vectors in 𝐇₀ the basic parity check vectors. According to the dual code principles, we can consider 𝐇₀ to be a parity check matrix for (𝜇^⟂+1) consecutive codewords. That is to say, if a row vector 𝐜 includes (𝜇^⟂+1) consecutive codewords, we have 𝐜×𝐇^𝑇₀=0. (15)

Furthermore, for a matrix 𝐂 consisting of a number of such vectors, we have 𝐂×𝐇^𝑇₀=0. (16)

In this paper we say two vectors 𝐜 and 𝐡 are correlated if they have the same length 𝑙 and ^𝑙_𝑖=1𝑐_𝑖ℎ_𝑖=0, (17) where the sum operator is defined in GF(2) and 𝐜=𝑐₁𝑐₂⋯𝑐_𝑙,𝐡=ℎ₁ℎ₂⋯ℎ_𝑙. (18)

One solution to the recognition of 𝐇₀ is enumerating all possible length values 𝑙 and listing all vectors with length 𝑙 to see which vectors have likelihood to be the basic parity check vectors. This can be implemented by enumerating all possible length values 𝑙, creating the observed data matrix 𝐑_𝑙 with row length 𝑙 (as shown in Figure 1), and trying to find the vectors which are correlated with most rows of 𝐑_𝑙. If we can find such vectors, the corresponding 𝑙 can be considered to be a possible estimation of 𝜔, which is defined to be the row length of the basic check matrix 𝐇₀: 𝜔=𝑛𝜇^⟂+1. (19) Furthermore, we can estimate the code length 𝑛 and dual code memory length 𝜇^⟂ by investigating the factors of the estimated 𝜔.

This scheme takes a very long elapsed time. Here we propose some principles to reduce the searching space.(1)Reduce the candidates of 𝑙.

According to (19), the parameter𝜔is a product of two integers which are all larger than 1. Therefore, we can drop the prime values while searching. So we just need to enumerate the values of 𝑙 from composite numbers.(1)Reduce the candidates of basic parity check vectors.

We can set the first and the last bit of each candidate vector to be 1. Hence, we just need to enumerate the combinations of the middle (𝑙−2) bits for each l-length vector.

In the noisy environment, not all the rows in 𝐑_𝑙 are correlated with the basic parity check vectors. 𝐑_𝑙 can be written as follows: 𝐑_𝑙=𝐂_𝑙+𝐄_𝑙, (20) where 𝐄_𝑙 is the error pattern corresponding to the elements in 𝐑_𝑙 and the elements in 𝐂_𝑙 are the original encoded (noiseless) bits of the received bits in 𝐑_𝑙. If 𝑙=𝑛(𝜇^⟂+1) and no error exists in the received stream, that is, 𝐄_𝑙=0, according to (16), we have 𝐑_𝑙×𝐇^𝑇₀=𝟎. (21)

In most noisy environments, because of the existence of error bits, (21) is not always true even if 𝑙 equals 𝑛(𝜇^⟂+1). So we cannot determine whether a vector 𝐡 is correlated with row vectors of 𝐑_𝑙 by checking the equation 𝐑_𝑙×𝐡^𝑇=𝟎. Instead, we need to compute the likelihood of a vector 𝐡 to be correlated with 𝐑_𝑙 and an appropriate threshold to determine whether the vector 𝐡 can be considered to be a basic check vector. We now deduce the threshold below.

For a given data matrix 𝐑_𝑙 and vector 𝐡, we denote by 𝐪 a vector calculated by 𝐑_𝑙×𝐡^𝑇 as follows: 𝐪=𝐑_𝑙×𝐡^𝑇. (22) Equation (22) is defined in GF(2), so the elements in vector 𝐪 are from the set {0,1}. And we denote by 𝑤𝑡(𝐪) the Hamming weight of the vector 𝐪. In noisy environment, the expectation of 𝑤𝑡(𝐪) can be calculated as follows: 𝐸[𝑤𝑡(𝐪)]=𝐿1−Pr𝐜×𝐡^𝑇=0, (23) where 𝐿 is the number of rows of the observed data matrix 𝐑_𝑙 and 𝐜 is any possible form of the rows of 𝐑_𝑙. Pr(𝐜×𝐡^𝑇=0) is the probability of 𝐜×𝐡^𝑇=0. If the matrix 𝐑_𝑙 is filled with correct parameters, that is, 𝑙=𝑛(𝜇^⟂+1), the first bit filled into 𝐑_𝑙 is the beginning of a codeword, and the vector 𝐡 is a valid check vector, Pr(𝐜×𝐡^𝑇=0) equals the probability that the vector 𝐜 has even error bits. So we can calculate the probability Pr(𝐜×𝐡^𝑇=0) as follows: Pr𝐜×𝐡^𝑇=0=^⌊𝑤/2⌋_𝑖=0𝑤𝜏2𝑖^2𝑖(1−𝜏)^𝑤−2𝑖=1+(1−2𝜏)^𝑤2, (24) where 𝑤 is the Hamming weight of 𝐡 and𝜏 is the channel transition probability. According to (23) and (24), we have 𝐸₁((𝐪=()=𝐸𝑤𝑡))1−1−2𝜏^𝑤2𝐿. (25) According to binomial-distribution theories, we can calculate the variance of 𝑤𝑡(𝐪) as follows: 𝐷₁((𝐪=()=𝐷𝑤𝑡))1−1−2𝜏^2𝑤4𝐿. (26)

If the data matrix 𝐑_𝑙 is based on incorrect parameters or 𝐡 is not a valid check vector, the probability Pr(𝐜×𝐡^𝑇=0) is Pr𝐜×𝐡^𝑇=0=Pr𝐜×𝐡^𝑇=1=12. (27) Therefore, the expectation and variance of 𝑤𝑡(𝐪) are 𝐸₂((𝐪=1=𝐸𝑤𝑡))2𝐷𝐿,₂((𝐪=1=𝐷𝑤𝑡))4𝐿. (28)

So we propose the threshold 𝛿 based on 𝜆-standard deviation principle as follows: 𝐸𝛿=¹+𝜆𝐷₁+𝐸₂−𝜆𝐷₂2=1−(1−2𝜏)^𝑤/2𝐿+𝜆1−(1−2𝜏)^2𝑤𝐿+)/4(1/2−𝜆(1/4)𝐿(2)⁻¹=𝐿2()1−1−2𝜏^𝑤2+𝜆√𝐿41−(1−2𝜏)^2𝑤−1. (29) Experimentally, we suggest choosing the parameter 𝜆 between 6 and 8 to get a good recognition result.

In (29), we must ensure that 𝐸₁+𝜆𝐷₁<𝐸₁−𝜆𝐷₂. (30)

So the number of rows of the observed data matrix 𝐑_𝑙 should meet the following condition: 𝜆𝐿>1−(1−2𝜏)^2𝑤+1(1−2𝜏)^𝑤2. (31)

The threshold 𝛿 and corresponding condition of 𝐿 are based on the prior information about the noise level, that is, the channel transition probability 𝜏. If such prior information is unknown, the calculation of (29) and (31) cannot be done. In this situation, we propose to set the threshold 𝛿 to be 𝐿/10 and let 12𝐿−𝜆14𝐿𝐿>𝛿=.10 (32) Therefore, 𝐿 should meet the following condition: 5𝐿>4𝜆². (33)

The threshold 𝛿=𝐿/10 is just an experimental value. If no parity check vector can be found based on such a threshold, we can increase the value of 𝛿 and redo the searching procedure. If 𝛿=𝜇/𝜃(𝜃>2), we must choose 𝐿 such that 12𝐿−𝜆14𝐿𝐿>𝛿=𝜃; (34) that is, 𝜃𝐿>𝜆𝜃−2². (35)

To implement the algorithm automatically by a computer program, we propose the following procedure to recognize the parameter𝜔defined by (19) and the codeword synchronization position 𝑡₀ as follows.

Step 1. Set the searching range of 𝜔; that is, set 𝜔_min and 𝜔_max.

Step 2. List all the composite numbers between 𝜔_min and 𝜔_max to form a set 𝜔₁𝜔₂⋯𝜔_𝜀, (36) where 𝜔₁<𝜔₂<⋯<𝜔_𝜀,𝜀is the number of composite numbers between 𝜔_min and 𝜔_max.

Step 3. Let𝑖=1.

Step 4. Let𝑙=𝜔_𝑖.

Step 5. Let 𝑡=1; fill the data matrix 𝐑_𝑙 with the received data. The first bit filled into 𝐑_𝑙 is the 𝑡th bit in the received stream.

Step 6. Let 𝑗=1.

Step 7. Create a vector: 𝐡=ℎ₁ℎ₂⋯ℎ_𝑙−1ℎ_𝑙, (37) where ℎ₁=ℎ_𝑙=1 and ℎ₂ℎ₃⋯ℎ_𝑙−1 is the binary form of 𝑗 with length 𝑙−2.

Step 8. Calculate the vector 𝐪 by (22) and denote by 𝑤𝑡(𝐪) the Hamming weight of 𝐪. If 𝑤𝑡(𝐪)<𝛿, stop the searching and output 𝑡 and 𝑙 to be the estimation of 𝑡₀ and 𝜔; that is, let 𝑡₀̂=𝑡 and 𝜔=𝑙. Besides, record 𝐡 as 𝐡_𝑥.

Step 9. If 𝑗<2^𝑙−2−1, let 𝑗=𝑗+1 and go back to Step 7. Otherwise, execute Step 10.

Step 10. If 𝑡<𝑛_max, let 𝑡=𝑡+1 and go back to Step 6. Otherwise, execute Step 9. 𝑛_max is the maximum of possible codeword length. 𝑛_max can be set to the maximum factor of l (except 𝑙 itself).

Step 11. If 𝑖<𝜀, go back to Step 4. Otherwise, execute Step 12.

Step 12. End the searching.

If we can stop the searching from Step 8, we can recognize the codeword length 𝑛 based on 𝑡₀̂, 𝜔, and 𝐡_𝑥. According to (19), the code length 𝑛 is a factor of 𝜔. So we propose to estimate 𝑛 by the following steps.

Step 1. List all the factors of 𝜔 (expect 1 and 𝜔 itself) to form a set 𝑛₁𝑛₂⋯𝑛_𝜎, (38) where 𝑛₁<𝑛₂<⋯<𝑛_𝜎, 𝜎 is the number of the factors of 𝜔 (expect 1 and 𝜔 itself); let𝑙=𝜔.

Step 2. Let 𝑖=1.

Step 3. Let 𝑛=𝑛_𝑖.

Step 4. Fill the data matrix 𝐑_𝑙 as follows: 𝐑_𝑙=𝐜₁𝐜₂⋮𝐜_𝐿, (39) where𝐜₁=𝑐₁⋯𝑐_𝑛∣𝑐_𝑛+1⋯𝑐_2𝑛∣⋯∣𝑐_{𝜇⟂𝑛+1}⋯𝑐_{(𝜇⟂+1)𝑛},𝐜₂=𝑐_𝑛+1⋯𝑐_2𝑛∣𝑐_2𝑛+1⋯𝑐_3𝑛∣⋯∣𝑐_{(𝜇⟂+1)𝑛+1}⋯𝑐_{(𝜇⟂+2)𝑛},𝐜₃=𝑐_2𝑛+1⋯𝑐_3𝑛∣𝑐_3𝑛+1⋯𝑐_4𝑛∣⋯∣𝑐_{(𝜇⟂+2)𝑛+1}⋯𝑐_{(𝜇⟂+3)𝑛},⋮⋮ (40)where 𝑐_𝑖(𝑖≥1) is the (𝑡₀̂+𝑖−1)th bit in the received data stream and 𝜇^⟂=𝜔/𝑛−1.

Step 5. Calculate the vector 𝐪 by 𝐪=𝐑_𝑙×𝐡_𝑥^𝑇 and get 𝑤𝑡(𝐪), the Hamming weight of 𝐪.

Step 6. If 𝑤𝑡(𝐪)<𝛿, stop the searching and output the current 𝑛 to be the estimated codeword length:𝑛=𝑛. Otherwise, execute Step 7.

Step 7. If 𝑖<𝜎, let 𝑖=𝑖+1 and go back to Step 3. Otherwise, execute Step 8.

Step 8. End the searching.

After the estimation of 𝑛, we can simply get the estimation of 𝜇^⟂ by 𝜇^⟂=𝜔𝑛−1. (41)

Finally, we search all the basic check vectors to recover the basic check matrix 𝐇₀. According to (13) and (14), we can estimate the encoder parameter 𝑘 and the𝑛−𝑘parity check vectors by the following steps.

Step 1. Let 𝑘=1.

Step 2. Let 𝑠=1.

Step 3. Create a vector 𝐱_𝑠 as follows: 𝐱_𝑠=𝑐₁⋯𝑐_𝑘𝑐_𝑘+𝑠𝑐_𝑛+1⋯𝑐_𝑛+𝑘𝑐_{𝑛+𝑘+𝑠}𝑐_2𝑛+1⋯, (42) where the definition of𝑐_𝑖(𝑖≥1)is the same as (40).

Step 4. Let 𝑙=(𝑘+1)(𝜇^⟂+1) and fill the elements in 𝐱_𝑠 into the matrix 𝐑_𝑙 with row length 𝑙.

Step 5. Enumerate all vectors 𝐡 with length 𝑙, where the last element of 𝐡 is 1, and calculate 𝑤𝑡(𝐪), where 𝐪=𝐑_𝑙×𝐡_𝑥^𝑇. If any vector 𝐡 can make 𝑤𝑡(𝐪)<𝛿, record 𝐡_𝑠=𝐡 and execute Step 6. Otherwise, execute Step 8.

Step 6. If 𝑠<𝑛−𝑘, let𝑠=𝑠+1 and go back to Step 3. Otherwise, execute Step 7.

Step 7. Stop the searching, and output 𝑘=𝑘 and the recorded 𝑛−𝑘 vectors 𝐡₁,𝐡₂,…,𝐡_𝑘𝑛−.

Step 8. If 𝑘<𝑛−1, let𝑘=𝑘+1 and go back to Step 2. Otherwise, execute Step 9.

Step 9. End the searching.

If such recognition procedure can successfully output𝑛−𝑘vectors𝐡₁,𝐡₂,…,𝐡_𝑘𝑛−, we can finally recover the basic parity check matrix 𝐇₀. We write the vectors 𝐡₁,𝐡₂,…,𝐡_𝑘𝑛− as follows:𝐡₁=ℎ_1,1⋯ℎ_𝑘1,ℎ_1,𝑘+1∣ℎ_1,(𝑘+1)+1⋯ℎ_{𝑘1,(𝑘+1)+}ℎ_1,2(𝑘+1)∣⋯∣ℎ_{𝜇1,⟂(𝑘+1)+1}⋯ℎ_{𝜇1,⟂(𝑘𝑘+1)+}ℎ_{𝜇1,(⟂+1)(𝑘+1)},𝐡₂=ℎ_2,1⋯ℎ_𝑘2,ℎ_2,𝑘+1∣ℎ_2,(𝑘+1)+1⋯ℎ_{2,(𝑘+1)+𝑘}ℎ_2,2(𝑘+1)∣⋯∣ℎ_{2,𝜇⟂(𝑘+1)+1}⋯ℎ_{𝜇2,⟂(𝑘𝑘+1)+}ℎ_{𝜇2,(⟂+1)(𝑘+1)},⋮𝐡_𝑘𝑛−=ℎ_{𝑛−𝑘,1}⋯ℎ_𝑘1,ℎ_1,𝑘+1∣ℎ_{𝑛−𝑘,(𝑘+1)+1}⋯ℎ_{𝑘𝑛−𝑘,(𝑘+1)+}ℎ_{𝑛−𝑘,2(𝑘+1)}∣⋯∣ℎ_{𝜇𝑛−𝑘,⟂(𝑘+1)+1}⋯ℎ_{𝜇𝑛−𝑘,⟂(𝑘𝑘+1)+}ℎ_{𝜇𝑛−𝑘,(⟂+1)(𝑘+1)}, (43)and then the parity check matrix 𝐇₀ can be estimated as follows:𝐇⁰=^{ℎ1,1⋯ℎ1,𝑘ℎ1,𝑘+1ℎ1,𝑘+2⋯ℎ1,2𝑘+1ℎ1,2𝑘+2ℎ1,𝑘+2}_{ℎ⋮⋱⋮⋱⋮⋱⋮⋱⋮𝑛−𝑘,1⋯ℎ𝑛−𝑘,𝑘ℎ𝑛−𝑘,𝑘+1ℎ𝑛−𝑘,𝑘+2⋯ℎ𝑛−𝑘,2𝑘+1ℎ𝑛−𝑘,2𝑘+2ℎ𝑛−𝑘,𝑘+2}^⋯_⋯…. (44)

5.2. Soft-Decision Situations

In hard-decision situations, we calculate 𝑤𝑡(𝐪) by evaluating how many vectors in 𝐑_𝑙 are correlated with a candidate parity check vector 𝐡. In soft decisions, we can utilize the soft output from the receiver to evaluate the reliability of each decision bit. Therefore, we can calculate the probability of each row in 𝐑_𝑙 to be correlated with 𝐡. For example, if a communication system uses BPSK demodulation in an additive-white-Gaussian-noise (AWGN) channel, according to the analysis in [14, 15], we can calculate the probability of 𝐜×𝐡^𝑇=1 and 𝐜×𝐡^𝑇=0 for given vectors 𝐜 and 𝐡 as follows: 𝑃_𝑟𝐜×𝐡^𝑇=1=02+12^𝑤_𝑗=1𝑐tanh^𝑢𝑗𝜎²𝑃_𝑟𝐜×𝐡^𝑇=1=12−12^𝑤_𝑗=1𝑐tanh^𝑢𝑗𝜎², (45) where 𝜎 is the noise variance, 𝑤 is the Hamming weight of 𝐡, 𝑢_𝑗 is the 𝑗th nonzero position in 𝐡, and 𝑐_𝑢𝑗 is the𝑢_𝑗th element in vector c. Based on this, 𝑤𝑡(𝐪) is calculated as follows: 𝑤𝑡(𝐪)=^𝐿_𝑖=1𝑃_𝑟𝐜_𝑖×𝐡^𝑇=1, (46) where 𝐜_𝑖 denotes the𝑖th row in 𝐑_𝑙.

6. Simulation Results

In this section we show the simulation results of the blind recognition of the convolutional coding parameters by utilizing the method introduced in this paper. The simulations include three parts corresponding to our proposed recognition algorithm on different noise level and different observed matrix size 𝐿. In the simulations we assume that the signal is transmitted on a binary symmetry channel (BSC) which is corrupted by AWGN.

For the GJETP-based recognition method, we show the false recognition ratios (FRR) in Figure 5 for several convolutional codes in different channel conditions when the observed window size 𝐿=200. We compare the proposed soft-decision based method with the hard-decision based algorithm proposed in [7]. We can see from the simulation results that when the soft information is introduced into the recognition algorithm, the recognition performances can be improved obviously. And as the SNR rises, the FRR curves descend more rapidly on soft-decision algorithm.

Figure 5 

Recognition performances of GJETP method.

Figure 6 shows the recognition performance of two convolutional codes for different observed matrix size, when the SNR 𝐸_𝑠/𝑁₀=2.36 dB. It shows that the soft information can help to make the FRR descend more rapidly when 𝐿 is rising. That is to say, in soft-decision situations, we can improve the recognition performance by increasing the number of rows in the data matrix of 𝐑_𝑙, while the hard-decision based algorithm cannot.

Figure 6 

Recognition performances on different size of 𝐑𝑙 for soft and hard decisions.

In Figure 7, we compare the performances of convolutional encoder recognition by GJETP method and correlation attack while recognizing the coding parameters of 𝐶(2,1,3). GJETP method is based on Gauss elimination on 𝐑_𝑙, so the influences of error bits are easily diffused during Gauss eliminating. Therefore, the fault-tolerance of GJETP method is limited. Correlation attack method can avoid this problem, so the recognition performance can be improved in low SNR situations.

Figure 7 

Comparison of GJETP method and correlation attack.

7. Conclusions

This paper proposes the methods of utilizing soft information to improve the recognition performance of convolutional encoder parameters. And we propose a formalized estimation of the parameter 𝑛 and synchronization positions. When introducing the soft information the recognition performance can be obviously improved and the simulations show the efficiency of the proposed methods. Besides, we propose a new algorithm to recover the basic parity check matrix by correlation attack. Although this method takes longer elapsed time, it can push the SNR limits of the GJETP method, and some principles are proposed to reduce the searching space. If the channel quality is well, the GJETP method has advantages on short computational delay. For a worse channel quality such that the GJETP does not work, correlation attack method has a significant advantage on its higher fault-tolerance.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.