Open glynpu opened 6 years ago
Hello,
Thank you for your feedback. I need to replicate your problem with the file mentioned. Could you please give more details about:
For MFS/LFS, we have not yet made the feature extractors open source, but it is possible to implement it using simple linear regression on log filterbank energies.
Thanks, Srikanth
Hi Srikanth,
Thank you for your kind help.
I upload the config and scp files here
https://github.com/glynpu/diarizationToolkitAnalysis
To replicate my problem, just run diartk.sh in the above repository.
Best wishes Liyong Guo
Hi Srikanth,
I am re-implementing your paper 'Two-Pass IB based Speaker Diarization System using Meeting-Specific ANN based Features'. Parameters like beta, NMI threshold, maximum segment duration are set according to the paper. Scp files are converted from AMI labels. 19 MFCCs are extract by HTK as input features.
For some audio files in AMI-1 data-set, this toolkit works pretty well. However, in case of ES2013c, the toolkit crash with the following prompt. Could you give a clue about what mistakes I have made?
Another problem is about the feature used in this paper. Could you recommend a toolkit to extract MFS (Mel Filterbank Slope) and Linear Filterbank Slope (LFS) features?
Thanks for your time and patience. Best wishes,
Liyong Guo