COMPUTER MODELLING & NEW TECHNOLOGIES 2014 18(12A) 245-249 Peng Yingying, Ren Xuegang, Hu Defa
245
Multibit-flipping decoding algorithm for low-density parity-check
codes
Yingying Peng
1
, Xuegang Ren
1
, Defa Hu
2
1
Department of Management and Information Engineering, Hunan University of Chinese Medicine, Changsha 410028, Hunan, China
2
School of Computer and Information Engineering, Hunan University of Commerce, Changsha 410205, Hunan, China
Received 1 June 2014, www.cmnt.lv
Abstract
Aiming at the Low-Density Parity-Check Codes, a reliability-based multibit-flipping decoding algorithm is proposed in the paper. The
multibit-flipping criterion is based on the reliable bit position and the threshold in the flipping-decision (number of flipping bits) can
be dynamically adjusted during the decoding process. The proposed algorithm is on the basis of the belief propagation decoding
algorithm, and then can be derived from its theory. Compared with the traditional weighted bit-flipping decoder and the multi-bit
flipping decoder, the proposed decoder can provide a faster converges faster convergent rate and better performances. Simulation
results demonstrate that the proposed algorithm achieves a better balance between performance and complexity.
Keywords: LDPC codes, multibit-flipping algorithm, belief propagation algorithm
1 Introduction
LDPC can adopt many different methods to decode, such
as the soft-decision, the hard-decision and the hybrid
schemes, etc, where the belief-propagation algorithm in
the soft-decision and the minimum-sum algorithm can
achieve its excellent performance but its operating
complexity is rather high. The bit-flipping algorithm in the
hard-decision, such as WBF, MWBF, IMWBF and
IMWBF, etc, can achieve a better balance between
performance and complexity. The general WBF algorithm
and its related algorithms just can flip one bit in each
iteration. If the algorithms can flip multi bits in each
iteration, the above error can be modified and the delaying
of the decoding can be reduced. In the references [8-11],
different types of the MBF decoding methods are
proposed. In the reference [8], each bit can provide a
flipping signal counter and each checking node can count
the reliability in each bit. If a certain bit reaches to the pre-
set threshold, the error probability of the bit is so high that
it must be flipped. The method proposed in the reference
[8] should be improved in the reference [9].
Aiming to the high-information bit, if the bit can reach
to the pre-set threshold and the flipping of the higher
reliable information is delayed, the higher decoding gain
can be obtained. In addition, the reference [10] introduces
a multi-bit algorithm called the Gradient Descent Bit
Flipping method and the algorithm is obtained from the
concept of the gradient descent. Compared with WBF
algorithm, the group shuffled and group replica shuffled
BF (GRSBF) proposed in the reference [10] can
effectively reduce the coding iteration times and have the
excellent decoding performance. Therefore, compared
Corresponding author’s e-mail: 31581677@qq.com
with the signal-bit flipping method, the multi-bit flipping
method can get the faster decoding convergent rate.
Compared with the SP algorithm, the bit flipping
algorithm proposed in the above reference still has the
obvious disparity in the performance aspect. In addition,
the selection of the flipping-decision threshold used in the
MBF algorithm [8-11] can be changed with the signal
noisy ratio and the coding rate. If the selection of the
flipping-decision threshold is not the optimum threshold,
the correction of the performance would be significantly
depredated. In order to improve the above problem, the
paper introduces a reliability-based multibit-flipping
decoding algorithm, and its threshold in the flipping-
decision can be dynamically adjusted during the decoding
process. Additionally, the proposed algorithm is not
related to the communication channel and cannot be
changed with the signal noise. The simulation results show
that the proposed methods can evidently reduce the
average iteration times, achieving the excellent
performance and having a lower operating complexity.
2 Symbol definition
LDPC is defined by the
Parity-
Check Code matrix
,
,
,
where K represents the information length, M represents
the numbers of the parity-check codes, N represents the
length of the codeword’s, d
v
represents the numbers of the
variable node degree, and dc represents the numbers of the
check node degree. The codeword vector
)1,0)(,...,(
21
nN
ccccc
should be input in the BPSK
modulator and output signal vector is
so