Acoustic Model Discussions

Flat
Julius VAD decoder and "sp" model in grammar
User: grisza50
Date: 7/20/2013 12:39 pm
Views: 3047
Rating: 10

Hi,

I would like to use Julius with "decoder VAD option". J have compiled all necessary files. My grammar is very simple ( i have words labeled in two "class".

My .grammar file:

S : NS_B SENT NS_E
SENT: CLASSB
SENT: CLASSA
S: NS_B WORD_LOOP NS_E
WORD_LOOP: WORD_LOOP SENT
WORD_LOOP: SENT

My .voc file:

% NS_B
<s>        sil
% NS_E
</s>        sil
% CLASSA

WORD1 pronunciation1

WORD2 pronunciation2

...
% CLASSB

WORD3 pronunciation3

WORD4 pronunciation4

I set the following options:

-spsegment           
-spdur 10           
-pausemodels sp   
-spmargin 40           
-spdelay 4           


But i always get "no pause word in dictionary needed for decoder-based VAD" error. I guess I have to add "sp" model somewhere but how?

I would like to use VAD to check if every word in continuous sequence fit to classA or B

Thank you for your help!!

--- (Edited on 7/20/2013 12:39 pm [GMT-0500] by grisza50) ---

PreviousNext