-
I'm wondering if it's possible to use `OnlineNnetFeaturePipeline` with cmvn on mfcc/plp/fbank features before decoding? `OnlineFeaturePipeline` accepts `cmvn-config` and `global-cmvn-stats` but `Onlin…
-
I Downloaded kaldi-generic-en-tdnn_sp-r20190227 and tried to use it with the kaldi online decoders.
I used the following arguments:
```bash
./online2-tcp-nnet3-decode-faster --feature-type=mfcc -…
-
Use this code, the `uttname` is utterance name with `spk+id`, but `uttline` is `spk` in `rttm` file. It always skips and then goes wrong.
Some Debug info:
- when running code like this, we skip …
-
Hi,
I was trying to use kaldi-websocket-python with a pre-trained German audio model, so I substituted the English model with my own one. I tried to place all the files of the new audio model in th…
-
Hi All,
In Speaker recognition, I am finding log likelihood ratio using ivector-plda-scoring(PLDA). In the paper "Probability Linear Discriminant Analysis" by S.Ioffe, after finding likelihood scor…
-
When I tried to build kaldi, I got the error:
Makefile:136: recipe for target 'cudafeat' failed
make: *** [cudafeat] Error 2
There were also some warnings like:
feature-online-cmvn-cuda.cu…
-
I keep getting a runtime error and the service keeps shutting down. I have described what I did below.
The README said to get my own kaldi model and must have a yaml file that describes the model.…
ghost updated
4 years ago
-
hi,
i have read the paper which the project cited , but in the paper it tells me you trained the model using mfcc features ,you use melspec in the segment.py 's _wav2feats ,so what exactly fe…
-
Hi,
Currently I am trying to compile `online2-tcp-nnet3-decode-faster.cc` and create an static library. I tried following the command `./configure --static` inside `src` folder but it didn't crea…
-
I'm trying to build this test program on my **raspberry pi 3** :
```C++
#include "vosk/vosk_api.h"
using namespace std;
int main(int argc, char const *argv[])
{
VoskModel *model = vosk_mode…