-
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.
…
-
Hi,
Thanks for your quick guide for ivector and xvector extraction. First of all model_3000h is not available on this link "https://pan.baidu.com/s/1kZHcOKQVkv_26emvG69IZg". So I tried to use pret…
-
To prepare for a full provenance chain, it's good to start adding easy-to-obtain prov and timing information to the DANE Results of the ASR worker and the download worker.
Next to the desired provena…
-
https://github.com/kaldi-asr/kaldi/blob/9d235864c3105c3b72feb9f19a219dbae08b3a41/egs/librispeech/s5/local/nnet3/run_ivector_common.sh#L86
According to the comment above this data subset script, 200…
-
It would be nice to have remote-ASR nodes and a plugin that is able to manage several alternative ASR services (such as Google, AWS, Kaldi, ...). Although these would not support grammars, they will b…
-
hello:
Hello, I encountered a problem when running this code. Here is the situation:
(env_nmesc) root@9cbf99057050:/data/Auto-Tuning-Spectral-Clustering# ./run_demo_clustering.sh
=== [INFO] …
-
Dear @JRMeyer ,
Thanks for creating a repo for multitask learning in Kaldi. It is much needed.
I have two questions regarding MTL in Kaldi.
1. I have speech datasets from two different domains …
-
-
Thanks for your work! I'm a big fan of clear notation in comments. You can certainly observe that there is an unsurprising correlation between project size and its comment math intensity. The examples…
-
**Describe your question**
I want to train a standalone transducer (`--asr_task asr_transducer`) with Vocal Tract Length Normalization (VTLN). Unfortunately, I could not find anything about this in t…