The is a powerful decoding method for . It uses forward and backward recursions to calculate probabilities of states and transitions in a , enabling accurate soft-decision decoding of received sequences.
This algorithm plays a crucial role in turbo codes and iterative decoding. By computing log-likelihood ratios for each bit, it provides soft information that can be used in iterative decoding schemes, improving overall error correction performance.
Recursion and Metrics
Forward and Backward Recursion
Top images from around the web for Forward and Backward Recursion
calculates the probability of reaching each state in the trellis at time k based on the probabilities of the states at time k−1 and the branch metrics connecting them
Denoted as αk(s), where s is the state at time k
calculates the probability of reaching each state in the trellis at time k based on the probabilities of the states at time k+1 and the branch metrics connecting them
Denoted as βk(s), where s is the state at time k
Recursions are initialized with known probabilities at the start and end states of the trellis (α0(s0)=1, βN(sN)=1)
Branch and State Metrics
Branch metrics represent the probability of transitioning from one state to another along a specific branch in the trellis
Calculated using the channel output and the expected output for each possible (Hamming distance or Euclidean distance)
State metrics combine the branch metrics and the probabilities from the forward and backward recursions to determine the likelihood of being in a particular state at time k
Denoted as γk(s′,s)=p(yk,sk=s∣sk−1=s′), where s′ is the previous state and s is the current state
Probabilities and Ratios
A Posteriori Probabilities (APP)
APP represents the probability of a particular bit being 0 or 1 given the entire received sequence
Calculated by summing the state metrics for all states at time k where the bit is 0 or 1
APP(uk=0)=∑(s′,s):uk=0αk−1(s′)γk(s′,s)βk(s)
APP(uk=1)=∑(s′,s):uk=1αk−1(s′)γk(s′,s)βk(s)
Log-Likelihood Ratios (LLRs) and Max-Log-MAP Approximation
LLRs represent the logarithm of the ratio of the APP for a bit being 1 to the APP for a bit being 0
LLR(uk)=logAPP(uk=0)APP(uk=1)
Max-log-MAP approximation simplifies the LLR calculation by using the maximum state metrics instead of summing over all states