-
I have some issues when compute gop features for only 1 audio based on gop_speechocean762.
Currently, when I compute gop on CPU, a processing time is around 6s, but when I use GPU by setting config u…
-
Hi, dear author, setting grammar in Vosk is very useful, so I copy UpdateGrammarFst() to Kaldi and test it with open source chain model(http://kaldi-asr.org/models/m13), but the result is bad( I set f…
-
i notice that when given short duration utterances, the PLDA score will have smaller aboslute value. Should there be something like score normalization?
-
1. vosk decode:
model config set:
```
--min-active=200
--max-active=7000
--beam=12.0
--lattice-beam=6.0
--acoustic-scale=1.0
--frame-subsampling-factor=3
--endpoint.silence-phones=1:2:3:4:5:6…
-
Hey guys, I notice there‘s on-the-fly feats in `asr_datamodule.py`:
https://github.com/k2-fsa/icefall/blob/32de2766d591d2e1a77c06a40d2861fb1bbcd3ad/egs/wenetspeech/ASR/pruned_transducer_stateless2/as…
-
I intend to put the best and baselines Kaldi results, e.g. from https://github.com/kaldi-asr/kaldi/blob/master/egs/wsj/s5/RESULTS for WSJ, for all datasets. Those numbers are not always supported by p…
-
Hi,
ubuntu 14.04.
use thi command(.setup.py install) to setup kaldi-pthon, I have set $KALDI_ROOT already
the errors are as follows:
/usr/include/python2.7/numpy/npy_1_7_deprecated_api.h:15:2: war…
-
如果我用librosa 进行特征抽取
# --use-energy=false # use average of log energy, not energy.
# --sample-frequency=16000 # Switchboard is sampled at 8kHz
# --num-mel-bins=40 # similar to Google's setup.
…
-
Recipes for recently added lookahead decoding fail because lookahead libraries are built in `/lib/fst` rather then `/lib`
adding
`export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$KALDI_ROOT/tools/open…
-
Use the kaldi recognizer that is in `/scratch/mim/jsalt/code/asr/kaldi-jsalt/egs/jsalt15-ffs/s5` to recognize the enhanced utterances. It was copied from another system, so there might be errors caus…