Decoding method and device, and decoder
The present disclosure relates to decoding methods and devices. One example method includes receiving N LLRs corresponding to a to-be-decoded signal, where N is a code length, classifying K decoded bits into reliable bits and unreliable bits based on at least one of a prior LLR or a posterior LLR, generating M decoding paths based on the N LLRs and a preset rule, and selecting each stage of target decoding path based on PM values of the M decoding paths to obtain a decoding result of each stage of decoded bit.
1. A method for decoding a to-be-decoded signal executed at a decoder system, wherein the decoder system includes at least one memory and at least one processor, wherein the at least one memory is configured to store a program instruction, and wherein the at least one processor is configured to control program instruction execution to cause the method to be performed, the method comprising:
receiving, at the decoder system, N log-likelihood ratios (LLRs) corresponding to a to-be-decoded signal, wherein N is a code length;
classifying, at the decoder system, K decoded bits into reliable bits and unreliable bits based on at least one of a prior LLR or a posterior LLR, wherein the prior LLR is an average LLR that is of a decoded bit and that is obtained based on channel statistical information, and the posterior LLR is an LLR that is of a decoded bit and that is calculated in real time based on the N LLRs;
generating, at the decoder system, M decoding paths based on the N LLRs and a preset rule, wherein 1≤M≤Q, Q is a preset total quantity of decoding paths, a total quantity of paths at each stage does not exceed L, L is a preset total quantity of decoding paths with a same length, and L≤Q, and wherein the preset rule is that: there is one next-stage path for the reliable bits and two next-stage paths for the unreliable bits, a decoding path with a path length of j consists of decision results of a first to a j th decoded bit, and each decision result corresponds to one stage of the decoding path, K is a final quantity of decoded bits, N, M, Q, L, K, and j are all positive integers, and 1≤j≤N;
selecting, at the decoder system, each stage of target decoding path based on path metric (PM) values of the M decoding paths to obtain a decoding result of each stage of decoded bits;
when a length of a first decoding path reaches N, performing, at the decoder system, cyclic redundancy check (CRC) on K decoded bits in the first decoding path; and at least one of:
returning, at the decoder system, a decoding success result if the CRC succeeds; or
continuing to perform, at the decoder system, CRC on K decoded bits in a next decoding path whose length reaches N, and returning a decoding failure result if a result of performing the CRC on K decoded bits in each decoding path whose length reaches N is that the CRC fails.
2. The method according to claim 1 , wherein during the generating M decoding paths based on the N LLRs and a preset rule, the method further comprises:
selecting l decoding paths with minimum PM values from the M decoding paths, and developing the l decoding paths in parallel, wherein 1≤l≤L.
3. The method according to claim 1 , wherein the classifying K decoded bits into reliable bits and unreliable bits based on at least one of a prior LLR or a posterior LLR comprises:
obtaining the reliable bits and the unreliable bits through classification according to one of the following rules:
{
If
LLR
post
(
j
)
≥
C
·
LLR
prior
(
j
)
or
j
∈
a
reliable
bit
LLR
prior
(
j
)
≥
θ
,
If
LLR
post
(
j
)
<
C
·
LLR
prior
(
j
)
and
j
∈
an
unreliable
bit
LLR
prior
(
j
)
<
θ
,
;
{
If
LLR
prior
(
j
)
≥
θ
,
j
∈
a
reliable
bit
If
LLR
prior
(
j
)
<
θ
,
j
∈
an
unreliable
bit
;
and
{
If
LLR
post
(
j
)
≥
η
,
j
∈
a
reliable
bit
If
LLR
post
(
j
)
<
η
,
j
∈
an
unreliable
bit
,
wherein
LLR prior (j) is the prior LLR, LLR post (j) is the posterior LLR, C, θ, and η are preset parameters, j is the j th bit, and 1≤j≤N.
4. The method according to claim 1 , wherein the generating M decoding paths based on the N LLRs and a preset rule comprises:
calculating an LLR of each stage of decoded bit in each decoding path in real time based on the N LLRs, and calculating a PM value of the corresponding decoding path based on the LLR of each stage of decoded bit; and
generating the decoding path according to the preset rule, selecting M decoding paths with minimum PM values when a total quantity of generated decoding paths is greater than Q, and sorting the M decoding paths in ascending order of PM values.
5. The method according to claim 4 , wherein the calculating an LLR of each stage of decoded bit in each decoding path in real time based on the N LLRs, and calculating a PM value of the corresponding decoding path based on the LLR of each stage of decoded bit comprises:
calculating a PM value of each decoding path according to the following formulas:
PM
l
0
=
0
,
∀
l
PM
l
j
=
PM
l
j
-
1
+
Δ
post
j
+
Δ
prior
j
,
j
∈
{
1
,
…
,
N
}
;
Δ
post
j
=
{
0
,
sign
(
LLR
post
(
j
)
)
=
1
-
2
·
z
(
j
)
LLR
post
(
j
)
,
otherwise
,
wherein z(j) is a decision value of the j th bit, and the value is 0 or 1; and
Δ
prior
j
=
-
∫
0
∞
Δ
post
j
f
(
Δ
post
j
)
d
Δ
post
j
,
wherein ƒ(x) is a probability density function of x; and
PM l j is a PM value of a decoding path corresponding to the j th decoded bit, and Δ prior j is negation of a mean value of Δ post j .
6. The method according to claim 1 , wherein before the selecting each stage of target decoding path based on path metric (PM) values of the M decoding paths, the method further comprises:
sorting the M decoding paths in ascending order of PM values based on a sorting condition, wherein the sorting condition is that: only currently split paths are sorted, and a path that is not split does not participate in the sorting.
7. The method according to claim 1 , wherein when M=Q, the method further comprises:
enabling path management to preferentially develop a shorter path until a difference between a length of a longest decoding path in all decoding paths and a length of a shortest decoding path in all the decoding paths is less than or equal to a preset threshold.
8. The method according to claim 1 , further comprising:
pruning at least one decoding path whose PM value is greater than a preset value in the M decoding paths, wherein the preset value is obtained through offline calculation or online calculation.
9. The method according to claim 8 , wherein the preset value is:
min
l
∈
{
1
,
…
,
Q
}
PM
l
+
ln
(
K
·
L
)
+
B
,
wherein
B is a preset parameter.
10. The method according to claim 1 , wherein the prior LLR is obtained through calculation according to the following formula:
LLR
prior
(
j
)
=
Send
an
all
-
zero
codeword
∫
-
∞
+
∞
LLR
post
(
j
)
f
(
LLR
post
(
j
)
)
dLLR
post
(
j
)
,
wherein
LLR prior (j) is the prior LLR.
11. A decoder, comprising:
a receiver, the receiver configured to receive N log-likelihood ratios (LLRs) corresponding to a to-be-decoded signal, wherein N is a code length;
at least one memory, the at least one memory configured to store a program instruction, an LLR table, a partial sum, a decoding result, a decoding path queue, and a path metric PM value queue, wherein the partial sum is an input of a G function; and
at least one processor, the at least one processor configured to control program instruction execution, wherein the at least one processor is configured to:
classify K decoded bits into reliable bits and unreliable bits based on at least one of a prior LLR or a posterior LLR, wherein the prior LLR is an average LLR that is of a decoded bit and that is obtained based on channel statistical information, and the posterior LLR is an LLR that is of a decoded bit and that is calculated in real time based on the N LLRs;
generate M decoding paths based on the N LLRs and a preset rule, wherein 1≤M≤Q, Q is a preset total quantity of decoding paths, a total quantity of paths at each stage does not exceed L, L is a preset total quantity of decoding paths with a same length, and L≤Q, and wherein the preset rule is that: there is one next-stage path for the reliable bits and two next-stage paths for the unreliable bits, a decoding path with a path length of j consists of decision results of a first to a j th decoded bits, and each decision result corresponds to one stage of the decoding path K is a final quantity of decoded bits N, M, Q, L, K, and j are all positive integers, and 1≤j≤N; and
select each stage of target decoding path based on path metric (PM) values of the M decoding paths to obtain a decoding result of each stage of decoded bit;
when a length of a first decoding path reaches N, perform cyclic redundancy check (CRC) on K decoded bits in the first decoding path; and at least one of:
return a decoding success if the CRC succeeds; or
continue to perform CRC on K decoded bits in a next decoding path whose length reaches N, and return a decoding failure if a result of performing the CRC on K decoded bits in each decoding path whose length reaches N is that the CRC fails.
12. The decoder according to claim 11 , wherein when generating the M decoding paths based on the N LLRs and the preset rule, the at least one processor is further configured to:
select l decoding paths with minimum PM values from the M decoding paths; and
develop the l decoding paths in parallel, wherein 1≤l≤L; and
the at least one memory is further configured to:
store a decoding path selected from the decoding path queue; and
store a decoding path whose PM value is calculated.
13. The decoder according to claim 11 , wherein the at least one processor is configured to:
obtain the reliable bits and the unreliable bits through classification according to one of the following rules:
{
If
LLR
post
(
j
)
≥
C
·
LLR
prior
(
j
)
or
j
∈
a
reliable
bit
LLR
prior
(
j
)
≥
θ
,
If
LLR
post
(
j
)
<
C
·
LLR
prior
(
j
)
and
j
∈
an
unreliable
bit
LLR
prior
(
j
)
<
θ
,
;
{
If
LLR
prior
(
j
)
≥
θ
,
j
∈
a
reliable
bit
If
LLR
prior
(
j
)
<
θ
,
j
∈
an
unreliable
bit
;
and
{
If
LLR
post
(
j
)
≥
η
,
j
∈
a
reliable
bit
If
LLR
post
(
j
)
<
η
,
j
∈
an
unreliable
bit
,
wherein
LLR prior (j) is the prior LLR, LLR post (j) is the posterior LLR, C, θ, and η are preset parameters, j is the j th bit, and 1≤j≤N.
14. The decoder according to claim 11 , wherein the at least one processor is configured to:
calculate an LLR of each stage of decoded bit in each decoding path in real time based on the N LLRs, and calculate a PM value of the corresponding decoding path based on the LLR of each stage of decoded bit; and
generate the decoding path according to the preset rule, select M decoding paths with minimum PM values when a total quantity of generated decoding paths is greater than Q, and sort the M decoding paths in ascending order of PM values.
15. The decoder according to claim 14 , wherein the at least one processor is configured to:
calculate a PM value of each decoding path according to the following formulas:
PM
l
0
=
0
,
∀
l
PM
l
j
=
PM
l
j
-
1
+
Δ
post
j
+
Δ
prior
j
,
j
∈
{
1
,
…
,
N
}
;
Δ
post
j
=
{
0
,
sign
(
LLR
post
(
j
)
)
=
1
-
2
·
z
(
j
)
LLR
post
(
j
)
,
otherwise
,
wherein z(j) is a decision value of the j th bit, and the value is 0 or 1; and
Δ
prior
j
=
-
∫
0
∞
Δ
post
j
f
(
Δ
post
j
)
d
Δ
post
j
,
wherein ƒ(x) is a probability density function of x; and
PM l j is a PM value of a decoding path corresponding to the j th decoded bit, and Δ prior j is negation of a mean value of Δ post j .
16. The decoder according to claim 11 , wherein the at least one processor is further configured to:
before selecting each stage of target decoding path based on the path metric (PM) values of the M decoding paths, sort the M decoding paths in ascending order of PM values based on a sorting condition, wherein the sorting condition is that: only currently split paths are sorted, and a path that is not split does not participate in the sorting.
17. The decoder according to claim 11 , wherein the at least one processor is further configured to:
enable path management when M=Q, to preferentially develop a shorter path until a difference between a length of a longest decoding path in all decoding paths and a length of a shortest decoding path in all the decoding paths is less than or equal to a preset threshold.
18. The decoder according to claim 11 , wherein the at least one processor is further configured to:
prune at least one decoding path whose PM value is greater than a preset value in the M decoding paths, wherein the preset value is obtained through offline calculation or online calculation.
19. The decoder according to claim 18 , wherein the preset value is:
min
l
∈
{
1
,
…
,
Q
}
PM
l
+
ln
(
K
·
L
)
+
B
,
wherein
B is a preset parameter.
20. The decoder according to claim 11 , wherein the prior LLR is obtained through calculation according to the following formula:
LLR
prior
(
j
)
=
Send
an
all
-
zero
codeword
∫
-
∞
+
∞
LLR
post
(
j
)
f
(
LLR
post
(
j
)
)
dLLR
post
(
j
)
,
wherein
LLR prior (j) is the prior LLR.