idnavid / spkr_diarization Goto Github PK

sgmestim error -- no training data available
sgmestim error -- no training data available
sgmestim error -- no training data available
sgmestim error -- no training data available
/home/manish/spkr_diarization/backup/code/resegment.py:83: RuntimeWarning: invalid value encountered in divide
trans_mat = trans_mat/float(N)
[[ nan nan nan nan]
[ nan nan nan nan]
[ nan nan nan nan]
[ nan nan nan nan]]
gmm_read() -- cannot open file ./out_dir/C3
hmm_read() -- cannot load GMM from file ./out_dir/C3 (state 1, line 2, file .//out_dir//diarizationExample_24502_hmm.txt)
sviterbi error -- cannot read hmm file .//out_dir//diarizationExample_24502_hmm.txt

Hi Navid,

Is it possible to set a minimum duration in the viterbi re-segment step?

In my experiment, the viterbi resegment module produces very short segments as floowing. Is this normal? Could you give me any clue to improve the performance?

C6 10.00000 10.02000 -2.966428e+01
C2 10.02000 10.03000 -1.689653e+00
C6 10.03000 10.04000 -3.208828e+01
C2 10.04000 10.05000 1.641983e-01
C6 10.05000 10.06000 -3.235529e+01
C2 10.06000 10.07000 8.267341e-02
C6 10.07000 10.08000 -3.185583e+01
C2 10.08000 10.09000 -1.047346e+00
C6 10.09000 10.10000 -3.135441e+01

Thank you.

Liyong Guo

It seems that you are not going with the I-Vector approach so can you please link some literature which might help me understand the code better.

Hi Navid
I am a newcomer, and this project is very helpful to me. How can I get the UBM_0 file? Is there an open data set? What is the austalk data mentioned in this spkr_diarization\code\readme.md file?

thanks！