>whats the use of viterbi algorithm in phoneme recognition?

you should be able to find some info on this in the HTK book

yeah but we are confused, because we dont know what its real purpose we have read that there is viterbi decoding and viterbi training... we have just started out and we got really confused if this is necessary or not

is viterbi used to determine to corresponding phoneme from a given acooustic model? how about viterbi training? is the training used to determine the corresponding acoustic model or determine the correposding phoneme directly?

> is viterbi used to determine to corresponding phoneme from a given acooustic model?

Yes

> how about viterbi training?

Despite it's also called "viterbi" it's a different algorithm. The only similarity is that it uses viterbi for the state alignment before it updates the acoustic state detector values.

> is the training used to determine the corresponding acoustic model or determine the correposding phoneme directly?

Training is used to update the acoustic model.