International Journal of Natural Computing Research

Volume 10 • Issue 1 • January-March 2021

Decoding Algorithm by Cooperation

Between Hartmann Rudolph Algorithm and
a Decoder Based on Syndrome and Hash
My Seddiq El Kasmi Alaoui, LIMSAD Laboratory, Faculty of Sciences Ain Chock, Hassan II University, Casablanca,
Morocco
Said Nouh, TIM Lab, Faculty of Sciences Ben M’sik, Hassan II University, Casablanca, Morocco

ABSTRACT

In this paper, the authors present a concatenation of Hartmann and Rudolph (HR) partially exploited
and a decoder based on hash techniques and syndrome calculation to decode linear block codes.
This work consists firstly to use the HR with a reduced number of codewords of the dual code then
the HWDec which exploits the output of the HR partially exploited. Researchers have applied the
proposed decoder to decode some Bose, Chaudhuri, and Hocquenghem (BCH) and quadratic residue
(QR) codes. The simulation and comparison results show that the proposed decoder guarantees very
good performances, compared to several competitors, with a much-reduced number of codewords
of the dual code. For example, for the BCH(31, 16, 7) code, the good results found are based only
on 3.66% of the all codewords of the dual code space, for the same code the reduction rate of the
run time varies between 78% and 90% comparing to the use of Hartmann and Rudolph alone. This
shows the efficiency, the rapidity, and the reduction of the memory space necessary for the proposed
concatenation.

Keywo
BCH Codes, Decoding Algorithm, Error Correcting Codes, Hartmann Rudolph (HR), HWDec, Linear Codes,
PHR-HWDec, QR Codes

1. INTR DUCTI N

Given that digital communication has become increasingly important, the volume of data exchanged
is growing continuously and the communication channels are not entirely reliable so the need to
ensure the reliability of the transmitted data to the reception is a pivotal process; hence channel
coding has become an indispensable means for detecting and correcting errors on computer networks,
telecommunication systems and data storage. A variety of error correcting codes are implemented in
diverse devices such as Smartphones, compact discs (CDs), digital versatile discs (DVDs), hard disks
or packets transferred over Interconnected Network (Internet) or over mobile networks. Given the
difficulty of the problem, several linear code decoding algorithms are used to improve the measured
performance as a function of bit error rate (BER).
Decoding algorithms employed in communication systems are classified into two principal
categories: soft decision and hard decision algorithms. Soft decision algorithms exploit directly
the received symbols and they employ principally the Euclidian distance as a measure to minimize
the distance. On the other hand, hard decision algorithms work on binary inputs that result from

DOI: 10.4018/IJNCR.2021010102

Copyright © 2021, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.


15
International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

thresholding of the transmit channel output. This type of algorithms usually use the Hamming
distance as a metric(Clarck & Cain, 1981). Among the works that are interested in the hard decision
algorithms, we find these based on the genetic algorithms (GA) (Azouaoui et al., 2012), (Gallager,
1962), (Morelos-Zaragoza, 2006), (Nouh, 2012), the algebraic decoder (Berlekamp, 1984), (Massey,
1969), (Chen et al., 2007) of Berlekamp-Massey based on compute of syndromes and it has an
efficient mechanism to localize all corrigible errors. We also find some works based on syndrome
decoding and the hash techniques (Chen et al., 2016), (El Kasmi Alaoui et al., 2019c), for example
in (El Kasmi Alaoui et al., 2019c), the researchers have presented two fast and efficient decoders,
HSDec(hard decision decoder based on Hash and Syndrome Decoding) and HWDec(Hard decision
decoder based on the error Weight Decreasing).
For soft decision decoders, we find an algorithm applicable for BCH codes based on test and
syndrome computing to localize the error positions (Jung et al., 2016). Askali et al. have presented
a version Soft In-Hard Out (SIHO) where they have used the MacWilliams’s permutation decoding
algorithm (McPD) as a hard decoder (Askali, 2012). To facilitate self-synchronization of the digital
communication systems Shim et al. have proposed a forward error correction codes in communication
channels (Shim, 2017). In (Berkani et al., 2017) a compact Genetic Algorithms (cGA) is used to
generate two dual domain Soft Decision decoders. The low complexity of the HSDec algorithm
compared with its competitors encouraged us to combine it with the Chase-2 algorithm (El Kasmi
Alaoui et al., 2017). In (Fossorier & Lin, 1995), Fossorier et al. have presented the OSD (Ordered
Statistics Decoder) algorithm which is based on ordered statistics and it’s applicable on linear block
codes. A decoding algorithm which principle is different to those cited above, is which presented by
Hartmann and Rudolph (HR) in (Hartmann & Rudolph, 1976). The main idea behind this decoder is
the symbol by symbol decoding of the received sequence. In (El Kasmi Alaoui et al., 2020; El Kasmi
Alaoui et al., 2019a; El Kasmi Alaoui et al., 2019b; Nouh, 2018), several decoders developed from
a serial concatenation are presented.
In (El Kasmi Alaoui et al., 2019a), we have presented a serial concatenation between Hartmann
& Rudolph partially exploited and HSDec decoder that is based on hash techniques and syndrome
calculation. The main idea in HSDec decoder is based on a hash function that permits to find the
error pattern directly from the syndrome of the received word. The storage position of each corrigible
error pattern is equal to the decimal value of its syndrome and therefore the time complexity is much
reduced comparing to known low complexity decoders. The main disadvantage of HSDec is its
spatial complexity, because it requires to previously storing all corrigible error patterns in memory.
To solve this problem, we propose to replace HSDec by HWDec decoder that is based also on hash
techniques and syndrome calculation, but it requires storing only the weight of each corrigible error
pattern instead of the error pattern itself.
In this paper, we present a decoding scheme developed from a serial concatenation between the
Hartmann Rudolph decoder and HWDec decoder. The remainder of this paper is structured as follows.
In section 2 we present the proposed serial concatenation between PHR and HWDec. In section 3, we
present experiment results and comparison of the proposed decoder. A complexity study is presented
in section 4. Finally, a conclusion is outlined in section 5.

2. TH PR P S D C NCAT NATI N B T N PHR AND H D C

2.1 Hartmann and Rudolph Decoder

The Hartmann & Rudolph (Hartmann & Rudolph, 1976) algorithm is a symbol by symbol decoder,
it is based on a probabilistic study to determine the value of a symbol rj of a sequence r. The owners
of this decoder have shown that this probability depends on all the codewords of the dual code C ⊥
this makes its time complexity very high, of exponential order O(n22n-k), and therefore unusable for
codes with reduced coding rate.

16
 International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

The formula used to determine if the mth bits of the decoded word c’ equal to 1 or 0 is as follows:

 2n −k n 
c ⊥ ⊕δ
 jq mq
 '  1 − φq 
cm = 0 if ∑∏   >0
(1)
 
j =1 q =1 
1 + φq 

c ' = 1 else
 m

where:

1, if i = j Pr(rm | 1)
δij =  and φm =
0, else Pr(rm | 0)


The bit c jq⊥ represents the q th bit of the j th codeword of the dual code C ⊥ .

2.2 H Dec
In (El Kasmi Alaoui et al., 2019c), the authors have presented the HWDec decoder (Hard decision
decoder based on decreasing of the error weigh) that works as follows:

Algorithm 1. The HWDec algorithm

Inputs:
b: The binary word to decode of length n.
The parity check matrix H or the generator polynomial h(x) of the
dual code.
TH2: The hash table of 2n-k integers.
t: The error correcting capability
POW2: The vector of n-k columns contains in each cell i the value
of the 2i
Outputs: The corrected word c.
Begin
e←The zero vector of length n
Compute the syndrome S of b.
We1←WeightOfError(S, TH2)
If (We1≤t) then
For i←1 to n do
b[i]←1-b[i]
S←Syndrome of b
We2←WeightOfError(S,TH2)
If (We2=We1-1) then
e[i]←1
End If
End For
End If
c←b⊕e
End

17
International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

The HWDec uses the following function which computes the weight of the error associated with
a given syndrome. The table TH2 associates for each syndrome S the weight of the unique corrigible
error of syndrome equal to S.

Algorithm 2. The WeightOfError function algorithm

Function WeightOfError(S: binary vector of n-k digits, TH2, POW2):

Integer
Position←hash(S, POW2)
w ←TH2[position]
Return w
End Function

2.3 The Proposed Serial Concatenation Between HR

Partially xploited and H Dec Decoder
To reduce the very high time complexity of the Hartmann & Rudolph decoder, the authors of (El
Kasmi Alaoui et al., 2019a) have used a reduced number M of words of the dual code instead of
2n-k proposed by Hartmann & Rudolph; and to guarantee a better result they have completed the
error correction by a second decoder. With this serial concatenation, they have reduced the time
complexity from O(n22n-k) to a quadratic complexity O(Mn2 + C(HSDec)), where C(HSDec) is the
time complexity of the HSDec decoder.
The HSDec decoder requires the storage of all corrigible error patterns in the memory what can
be an obstacle under certain conditions. For this, in this article we propose to replace it by HWDec
decoder which just requires the storage of the weights of the error patterns and not the patterns
themselves. So, for a code C (n, k, d) and due to the use of HWDec instead of HSDec, we reduce n
times the memory space used to store the corrigible error patterns. We call PHR-HWDec the resulting
decoder of the serial concatenation between Hartmann & Rudolph partially exploited and the HWDec
decoder. The concatenation scheme proposed in this work is based on the following two steps:

• The first one consists in correcting the low reliability symbols, precisely those whose value is
below a reliability threshold (RT): partial use of the HR decoder.
• The second step is to definitively decode the sequence prepared in the previous step by the
HWDec decoder.

The Figure 1 schema summarizes the principle of the PHR-HWDec decoding algorithm.

3. XP RIM NT R SULTS AND C MPARIS N

In this section, we give the performances of the PHR-HWDec algorithm for some linear codes and a
comparison with other decoding algorithms over an AWGN binary channel (Additive White Gaussian
Noise) with a BPSK (Binary Phase Shift Keying) modulation is done. We note that the Binary Error
Rate (BER) is calculated based on a minimum number of 200 residual errors and on a minimum
number of 1000 transmitted blocks. We also mention that without coding/decoding techniques the
BER reaches 10-5 for a value of SNR equal to 9.6 dB over an AWGN binary channel using a BPSK
modulation.

3.1 xperiment Results of PHR-H Dec

Figures 2 and 3 present respectively the obtained performances of the proposed decoder for some
BCH and Quadratic Residue (QR) codes.

18
 International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

Figure 1. The proposed concatenation decoding scheme

Based on these figures, we deduce that the gain of coding measured for a BER value equal to
10-5 is about 3 dB for the BCH(31, 16, 7) code, about 2.8 dB for the BCH(31, 21, 5) code, about 2.8
dB for the BCH(63, 51, 5) code, 4.2 dB for the QR(31, 16, 7) code and about 3.9 dB for the QR(23,
12, 7) code.

4. C MPARIS N B T N TH P RF RMANC S GUARANT D

B PHR-HSD C AND TH S F PHR-H D C

In order to show the efficiency of the proposed decoding scheme, we have applied it to calculate the
errors correction performances of several BCH and QR codes. In figures 4 and 5, we drew comparisons
of correction performances guaranteed by the PHR-HSDec and PHR-HWDec decoders respectively
for BCH(31, 16, 7), QR (31, 16, 7) and BCH (63, 51, 5) codes.
From Figure 4(a), we can deduce that the two decoders guarantee the same performance in
terms of the BER for the studied code. Figure 4(b) shows that from an SNR value equal to 3 dB, the
performance guaranteed by PHR-HWDec relatively exceeds that of PHR-HSDec for the QR(31, 16,
7) code.

Figure 2. Performances of PHR-HWDec decoder for some BCH codes of length (a) 31; (b) 63

19
International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

Figure 3. Performances of PHR-HWDec decoder for some QR codes

Figure 4. Comparison of the performances of PHR-HSDec and PHR-HWDec decoders for (a) BCH(31, 16, 7) code; QR(31, 16, 7) code

Figure 5, shows that the performance guaranteed by PHR-HWDec relatively exceeds that of
PHR-HSDec for the BCH(63, 51, 5) code.

20
 International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

Figure 5. Comparison of the performances of PHR-HSDec and PHR-HWDec decoders for BCH(63, 51, 5) code

5. C MPARIS N B T N TH P RF RMANC S
GUARANT D B PHR-H D C AND C MP TIT RS

In order to situate the performances guaranteed by our decoder compared to those ensured by the
competitors, we subjected it to a succession of comparison. In figure 6(a), we present a comparison
of BER performances between PHR-HWDec, PHR-Chase(El Kasmi Alaoui et al., 2020), Chase-
HSDec(El Kasmi Alaoui et al., 2017), SDHT(El Kasmi Alaoui et al., 2019b) and PHR-HSDec(El
Kasmi Alaoui et al., 2019a) for the QR(31, 16, 7) code. From figure 6(a), we can deduce that the
performances of our PHR-HWDec decoder relatively pass those guaranteed by PHR-HSDec, SDHT
and PHR-Chase and greatly exceed those of Chase-HSDec.
In Figure 6(b), we present a comparison of BER performances between PHR-HWDec, PHR-
Chase (El Kasmi Alaoui et al., 2020), Chase-HSDec (El Kasmi Alaoui et al., 2017) and SDHT (El
Kasmi Alaoui et al., 2019b) for the BCH(31, 16, 7) code. From figure 6(b), we can deduce that the
performances guaranteed by our PHR-HWDec decoder are same to those of SDHT and PHR-Chase
and far exceed those of Chase-HSDec.
In Figure 7(a), we plot the error correction performances of PHR-HWDec, PHR-HSDec (El Kasmi
Alaoui et al., 2019a), cGAD(Berkani et al., 2017), Chase-HSDec(El Kasmi Alaoui et al., 2017) and
PHR-SPDA (Nouh, 2018) for the BCH(63, 45, 7) code. From this figure we deduce that the correction
performances of PHR-HWDec exceed those of cGAD, PHR-SPDA and Chase-HSDecfor this code.
This figure also shows that PHR-HSDec remains the best decoder for this code.
The performance comparison results of the PHR-HWDec decoder with those of PHR-HSDec,
Chase-HSDec, PHR-ARDecGA, PHR-SPDA and ocGAD for BCH (63, 51, 5) code have been plotted
in Figure 7(b); this figure shows that the correction performances provided by PHR-HWDec are the

21
International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

Figure 6. Comparison of the performances of PHR-HWDec and several concurrent for (a) QR(31, 16, 7) code; BCH(31, 16, 7) code

best compared to those of all competitors for the studied code. For example for a BER=10-5, PHR-
HWDec provides a coding gain of about 0.4 dB over the PHR-SPDA decoder.

6. IMPACT F TH C NCAT NATI N SCH M N

TH P RF RMANC F TH HR D C D R

In order to show the importance of the proposed concatenation, we studied the performances of the
Hartmann & Rudolph decoder without HWDec and those of the proposed concatenation for BCH(31,
21, 5), BCH (31, 16, 7), BCH (63, 51, 5) and QR(31, 16, 7) codes; the results of these comparisons
are shown in Figures 8 and 9 respectively.
From Figure 8, for an SNR value equal to 5.5 and due to the proposed concatenation scheme,
we deduced that the coding gain reaches 0.5 dB for the BCH (31, 21, 5) code and exceeds 1.7 dB
for the BCH(31, 16, 7) code.

Figure 7. Comparison of the performances of PHR-HWDec and several concurrent for (a) BCH(63, 45, 7) code; BCH(63, 51, 5) code

22
 International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

Figure 8. Performances comparison of HR and PHR-HWDec decoders for (a) BCH (31, 21, 5) code; (b) BCH (31, 16, 7) code

From Figure 9, for an SNR value equal to 5 and due to the proposed concatenation scheme, we
deduced that the coding gain reaches 0.2 dB for the BCH (63, 51, 5) code and exceeds 1.4 dB for
the QR(31, 16, 7) code.
These figures clearly show the impact of the concatenation scheme on the error correction
performance guaranteed by the Hartmann & Rudolph decoder for all the codes studied.

7. C MPL XIT STUD

In the previous subsection, we have showed the positive impact of the schema of concatenation on
the error correction performances. In this section, we present a time complexity study to show the
speed of the decoding algorithm proposed.
In the introduction section, we have indicated that our goal in this paper is to decrease the
spatial complexity of the PHR-HSDec decoder. For this, we have proposed to replace HSDec in the
concatenation schema by HWDec given that does not store all corrigible errors pattern but just their

Figure 9. Performances comparison of HR and PHR-HWDec decoders for (a) BCH (63, 51, 5) code; (b) QR (31, 16, 7) code

23
International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

Table 1. Time complexities of the Hartmann and Rudolph decoder without and with concatenation

Time Complexity
Decoder
Cyclic Linear Codes Non-Cyclic Linear Codes
Hartmann and Rudolph O(n 2 )
2 n-k
O(n 2 )
2 n-k

PHR-HSDec O(Mn2+ log(n).log(n-k)+3n-2k) O(Mn2+n2)

PHR-HWDec O(Mn2+ n*(log(n).log(n-k)+3n-2k)) O(Mn2+n3)

weights which allows us to minimize the memory space used in the operation of preparing the hash
table.
In Table 1, we present the time complexities of the Hartmann & Rudolph decoder without and
with concatenation for cyclic and non-cyclic linear codes. We note, that HSDec and HWDec decoders
exploit the polynomial form to considerably reduce the run time of decoding cyclic codes in addition
to the gain of time that they offered by the hash techniques used.
The comparison of the time complexities of Hartmann and Rudolph without and with
concatenation shows that with concatenation we were able to go from exponential complexity to a
quadratic one. Also, this table shows that the time complexity of HWDec exceeds n times that of
HSDec; on the other hand, the spatial complexity of HSDec has been reduced n times in the HWDec
algorithm. So, we have two decoders the first is PHR-HSDec with reduced time complexity but high
spatial and the second decoder PHR-HWDec with a high time complexity compared to PHR-HSDec
but reduced spatial.
We also carried out an experimental study of time complexity to show the impact of concatenation
on the execution time of the Hartmann and Rudolph decoder. In Figure 10, we present the ratio (in
percent) between the execution time required by PHR-HWDec and that required by HR.
From this figure, we can deduce that in the case where the concatenation scheme is used; the
reduction rate of the run time varies between 78% and 90% for BCH(31, 16, 7) code and between
78% and 86% for QR(31, 16, 7) code.
We note that the good results, found in the case of BCH(31, 16, 7) and QR(31, 16, 7) codes, are
obtained just with a reduced number of codewords of the dual code:

Figure 10. Ratio between the run time required by PHR-HWDec and that required by HR (in %)

24
 International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

• 3.66% of the size of the dual space in the case of the BCH(31, 16, 7) code.
• 15.25% of the size of the dual space in the case of the QR(31, 16, 7) code.

From the above, we can conclude that our proposed PHR-HWDec decoder provides very good
performance in terms of the bit error rate and also in terms of complexity.
Due to this concatenation; we were able, in addition to reducing the execution time of HR, to
improve the error correction performances of the HR decoder, which shows the success of the idea
of concatenation.

8. C NCLUSI N

In this paper, we have presented the PHR-HWDec decoder which represents an improved version
in terms of spatial complexity compared to the previous version where HSDec has been used. The
advantage of HWDec is that it stores just the weights of the error patterns and not the patterns
themselves. We have applied the proposed decoder to decode some BCH and QR codes. The simulation
and comparison results show that the proposed decoder guarantees very good performances compared
to studied competitors. The number of codewords used in the decoding process is very small, which
has allowed us to alleviate in a very powerful way the time complexity. For example, by applying our
decoding scheme to the BCH(31, 16, 7) code, we were able to reduce the execution time of the HR
algorithm by a rate that varies between 78% and 90%. The best results of this concatenation schema
will open new ways for the artificial intelligence algorithms in the coding theory field.

25
International Journal of Natural Computing Research
Volume 10 • Issue 1 • January-March 2021

R F R NC S

Askali. (2012). A Soft decision version of the Permutation decoding algorithm. NTCCCS 12 workshop. Oujda.
Azouaoui, A., Chana, I., & Belkasmi, M. (2012). Efficient Information Set Decoding Based on Genetic
Algorithms. International Journal of Communications, Network and Systems Sciences, 05(07), 5. doi:10.4236/
ijcns.2012.57052
Berkani, A., Azouaoui, A., Belkasmi, M., & Aylaj, B. (2017). Improved Decoding of linear Block Codes using
compact Genetic Algorithms with larger tournament size. International Journal of Computer Science Issues, 14.
Berlekamp, E. R. (1984). Algebraic Coding Theory (rev. ed.). Aegean Park Press.
Chen, Y., Huang, C., & Chang, J. (2016). Decoding of binary quadratic residue codes with hash table. IET
Common., 10(1), 122–130. doi:10.1049/iet-com.2015.0546
Chen, Y. H., Truong, T. K., Chang, Y., Lee, C. D., & Chen, S. H. (2007). Algebraic decoding of quadratic residue
codes using Berlekamp-Massey algorithm. Journal of Information Science and Engineering, 23, 127–145.
Clarck, G. C., & Cain, J. B. (1981). Error-Correction Coding for Digital Communication. Plenum.
doi:10.1007/978-1-4899-2174-1
El Kasmi Alaoui, M. S., Nouh, S., & Marzak, A. (2019a). Fast and efficient decoding algorithm developed
from concatenation between a symbol-by-symbol decoder and a decoder based on syndrome computing and
hash techniques. In The 1st International Conference on Embedded Systems and Artificial Intelligence ESAI’19.
FSDM, Usmba Fez Morocco.
El Kasmi Alaoui, M. S., Nouh, S., & Marzak, A. (2019b). High Speed Soft Decision Decoding of Linear Codes
Based on Hash and Syndrome Decoding. International Journal of Intelligent Engineering and Systems, 12(1).
El Kasmi Alaoui, , M.SJoundan, , I.ANouh, , SMarzak, , A. (2020). Improved decoding of linear block codes
by concatenated decoders. International Journal of Recent Technology and Engineering, 8(5).
El Kasmi Alaoui, M. S., Nouh, S., & Marzak, A. (2017). A low complexity soft decision decoder for linear block
codes. In The 1st International Conference On Intelligent Computing in Data Sciences (ICDS). University My
Ismail, superior School of Technology.
El Kasmi Alaoui, M. S., Nouh, S., & Marzak, A. (2019c). Two New Fast and Efficient Hard Decision Decoders
Based on Hash Techniques for Real Time Communication Systems. In Lecture Notes in Real-Time Intelligent
Systems, RTIS 2017. Advances in Intelligent Systems and Computing (Vol. 756). Springer. doi:10.1007/978-3-
319-91337-7_40
Fossorier, M. P. C., & Lin, , S. (1995). Soft decision decoding of linear block codes based on ordered statistics.
IEEE Transactions on Information Theory, 184(5), 1379–1396. doi:10.1109/18.412683
Gallager, R. G. (1962). Low-Density Parity-Check Codes. I.R.E. Transactions on Information Theory, 8(1),
21–28. doi:10.1109/TIT.1962.1057683
Hartmann, C. R., & Rudolph, L. D. (1976). An optimal symbol by symbol decoding rule for linear codes. IEEE
Transactions on Information Theory, 22(5), 514–517. doi:10.1109/TIT.1976.1055617
Jung, B., Kim, T., & Lee, H. (2016). Low-Complexity Non-Iterative Soft-Decision BCH Decoder Architecture
for WBAN Applications. Journal of Semiconductor Technology and Science, 16.
Massey, J. L. (1969). Shift-register synthesis and BCH decoding. In IEEE 1969 Transaction on Information
Theory IT-15, 1, 122–127. doi:10.1109/TIT.1969.1054260
Morelos-Zaragoza, R. H. (2006). The Art of Error Correcting Coding (2nd ed.). John Wiley and Sons.
doi:10.1002/0470035706
Nouh. (2018). Efficient Serial Concatenation of Symbol By Symbol and Word by Word decoders. International
Journal of Innovative Computing, Information, & Control, 14.
Nouh, S. (2012). Majority voting procedure allowing soft decision decoding of linear block codes on binary
channels. International Journal of Communications, Network and Systems Sciences, 5.

26