Hi,

I know that the following line below is the format break up my MFCC files into time segments for my training script:

[code]

filename_00001166_00001570.mfcc=data/filename.mfcc[116,157]
[/code]

This is scripting the feature file for *filename* for the time segment from time=1166ms to 1570.

The transcription is from LDC corpus and annotated such that it's already got time segments marked.

My question is, must all time segments in the training script be contiguous? What if I want to exclude some of the time segments in the transcription?

--- (Edited on 10/5/2010 9:42 am [GMT-0500] by asaaki) ---