[Lena S]

It was actually just my having put train_subset in one too many places in do_alignment. But I did get some extra understanding out of exploring the other possible errors!

[Lena S]

Thanks! I resolved the error and my script appears to be working now.

[Lena S]

I’m getting the “try a larger beam search thresh for HVite” error mentioned above if I remove even just 1 line of my train.scp file. I have run endpointing without getting any errors. So:

1. How do I tell for sure that endpointing has worked?
2. What causes the do_alignment script to raise this error? I’d like to try working out a solution myself, but it isn’t clear to me from the script. Is it that not enough instances of vowels are found in the data? This would seem to make sense, except that only removing 1 line shouldn’t get rid of enough data to raise this error.

[Lena S]

I’m having this issue with endpointing. I’m trying to do the modification of training on less data and I get another error when I run make_mfccs if I don’t do endpointing (as I have seen in a previous run through the commands).

I’ve run these commands:

for FILE in wav/*.wav; do sox $FILE ../silenceTrimmed/$FILE vad -t 6 -s 0.1 -p 0.1 reverse vad -t 4 -s 0.1 -p 0.3 reverse; done

for FILE in ‘ls wav/*.wav’; do sox $FILE ../silenceTrimmed/$FILE vad -t 6 -s 0.1 -p 0.1 reverse vad -t 4 -s 0.1 -p 0.3 reverse; done

and I get the following error:

/Volumes/Network/courses/ss/festival/festival_mac/multisyn_build/bin/sox: cannot execute binary file

[Lena S]

Hi Davy,

Many of us have been using Appleton during the occupation and have had no issues accessing it. The occupiers do not stop students entering to work on assignments.

[Author]

Posts