>I'll start to record few voice file to be more involved in voxforge, the main

>opensource software.

Thanks for helping out.  VoxForge does not provide software, only a speech corpus and acoustic models

>Can someone got links or documentation about the analysis system used

>by voxforge to got voice/wav data --> to text ?

see the VoxForge tutorial

Thanks a lors for your answer :)

Just few questions about the acoustic model

1/ When I record a sentence, all the audio data seems to be recorded, no cut at all, I think that each word are used with speech recognition programm so the work of SPHINX or JULIUS or ... to do thant and not VOXFORGE ?

2/Is there simple sample of HMM algorithms/software not only specific for audio but to compare binary data voice/picture etc...

3/Is there any reference about history of speech recognition ?

4/During years 1950 research lab tried to make understanding pictures from sound but these researchs seems to be totaly disapearred today...

5/Noises : I just talking about noises during voice recording and that expert explain to me that if the learning session was during "noisy" environment, and the "speech recognition" is during an other noisy environment the mysterious HMM can take of that and "understand"... but if there is no more noise during "speech recognition" from "noisy corpus" the "software" will be absolutely confused due to HMM algorythm... Why ? How HMM work with noises ?

PS : If you see CedricVasseurAliasBeepMaster in contributor list... it's my voice ;-)