Re: Turbo , Convolutional Codes

[Mike Miller]

Cc'ing the help list. Please remember to include the mailing list on
all replies, so that others may benefit from the discussion. Also
please try to reply inline on Octave mailing lists instead of replying
at the top without any context.

On Mon, Nov 18, 2013 at 18:34:16 +0100, Jan 'oglop' Gazda wrote:
> I have learned a lot about convolutional codes ... I started from scratch ..
> The final output of my project should be simple example of coding and
> decoding process for convolutional or turbocode .. primary in octave
> If you know about better programming language or so, then tell me. I know
> little of python and C

Octave is certainly suitable. Whatever language you are most familiar
with should be fine.

> Let's say I have a
> word which I want to code: W=[1 0 1 1]
> with my convolutional coder which has Generator vectors: G1 = [1 0 1], G2 =
> [1 1 1]
> So code word can be done like this:
> output1 = mod(conv(W,G1),2),
> output2 = mod(conv(W,G2),2)
>
> output1 =  1   0   0   1   1   1
>
> output2 =   1   1   0   0   0   1
>
> CodeWord = [ 11 01 00 10 10 11] - I wrote simple function to merge the
> outputs

Looks ok so far. Note that by using the full convolution you are
implicitly adding tail bits to flush the encoder. So technically this
is the output for an input sequence of

  w = [1 0 1 1 0 0]

Adding tail bits gives you a known final state for the encoder trellis.

> Now I have to decode the word, but I don't know how to demonstrate it like
> this .. do I have to code full viterbi or hard decoder by myself or are
> there any mathematical methods to decode it.
> I did not get so far with viterbi decoding, I just know how it works and how
> to use the trellis to decode.

For a specific short example like this, you could start by manually
recreating the steps that the Viterbi algorithm would do for each
symbol in the coded sequence.

HTH,

-- 
mike