Re: [linux-audio-dev] Speech analysis

I should have clarified that the voice samples will be submitted by court reporters who are trained to leave short pauses between most words to achieve maximum clarity for use with speech recognition. And the markers need not all place themselves accurately. If we get a 90% +/ accuracy rate, I would be very pleased. We would check and edit all markers by listening. For our purpose, we may also concatenate frequently-used phrases of very short words, such as "in a" Susan Jean-Marc Valin wrote: Le lundi 06 juin 2005 à 09:52 -0400, Susan Cragin a écrit : Hi. I'm a lurker from the Open-Source Speech Recognition Initiative. We are starting to collect voice samples and need an audio program that can segment speech by word, or attempt to, by recognizinig all the silences and placing markers at each location. Any ideas? If your speech is really clean, then maybe a simple energy criterion (E lower than threshold for at least T time) may work. If not, I don't think anything simple will. Besides, you jut can't make the difference between two words separated by silence and a longer word. Jean-Marc