In information theory, how do I calculate the probability of an erroneous transmission? Let's take for instance a binary symmetric channel with an error probability $ 1-d=0.25 $ and send codewords of length 6 coded in a Hamming code able to correct up to 1 error.

  • $\begingroup$ I don't know any Hamming codes of length 6? AFAIK the length of a Hamming code is one less than a power of two. No matter - the question is still answerable, if it is about a single-error-correcting code of length six. $\endgroup$ – Jyrki Lahtonen Aug 31 '12 at 10:20
  • $\begingroup$ @JyrkiLahtonen That's a good remark, addressed by a previous comment unfortunately already deleted. It must be one of the Hamming code's variations. Thats the version I was taught at my University and the fact it is a non-standard version indeed caused me some confusion. I would be thankful if someone told me what are the objectives and the theory behind this modification. $\endgroup$ – infoholic_anonymous Aug 31 '12 at 15:42
  • $\begingroup$ I was the one who deleted my comment asking which shortened Hamming code was under consideration. I have posted a detailed answer about how shortened Hamming codes work in an answer to another question by the OP. I don't think the question is fully answerable without knowing more about the code. $\endgroup$ – Dilip Sarwate Aug 31 '12 at 20:23

We assume independence of bit errors. This is a somewhat dubious assumption, since errors often occur in bursts.

The probability of erroneous interpretation (or inability to decode) of a codeword of length $6$ is the probability that $2$ or more bits are incorrectly transmitted. The probability that $0$ bits are wrong is $(0.75)^6$. The probability that exactly $1$ bit is wrong is $6(0.25)(0.75)^5$. Add these two numbers, subtract the result from $1$.

  • $\begingroup$ Actually, you have found the probability $P(C)$ of correct decoding. For shortened Hamming codes, the complementary probability $1 - P(C)$ is the sum of the error probability $P(E)$ (meaning the output is the wrong codeword) and the failure probability $P(F)$ (meaning the decoder is unable to decode the received word into a valid codeword because a detectable but undecodable error pattern has occurred. My answer to this question describes how such decoder failures occur, and they do depend on how the shortening was done. $\endgroup$ – Dilip Sarwate Aug 31 '12 at 20:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.