Operating System: CentOS7.6.1810_x64
gcc Version: 4.8.5
Brief introduction
Kaldi was born in 2009 JohnsHopkins University, the beginning of the project focused on the subspace Gaussian model (SGMM) modeling and vocabulary learning sample survey, the code of conduct is based on the development of HTK, is now C ++ as the primary language. But with the addition of more participants, especially the depth of support for the neural network (DNN), so that the development of Kaldi more than several other well-known open source projects. More importantly, Kaldi maintained and updated very timely, basically every day new progress report, and follow-up in terms of academic research of new algorithms more quickly. Many foreign companies and research institutions are using Kaldi platform, of course, more domestic companies, in fact, is also based on this platform to do to improve, especially in recent years, emerging companies and the Enterprise Institute.
Official Website: http://kaldi-asr.org/
github Address: https://github.com/kaldi-asr/kaldi
Compile
Get the source:
git clone https://github.com/kaldi-asr/kaldi
Review the installation steps:
[root@local src]# cat kaldi/INSTALL This is the official Kaldi INSTALL. Look also at INSTALL.md for the git mirror installation. [for native Windows install, see windows/INSTALL] (1) go to tools/ and follow INSTALL instructions there. (2) go to src/ and follow INSTALL instructions there. [root@local src]#
That is:
1, into the tools directory of the installation instructions INSTALL;
Implementation Extras / check_dependencies SH check dependencies, the installation according to the prompt; Then performs the make -j 8 commands make (where 8 is the number of concurrent during make, may be modified according to the number of CPU);
2. After installation tools directory contents into the src directory, follow the instructions to install the INSTALL;
./configure --shared make depend -j 8 make -j 8
Where 8 is the number of concurrent during make, it may be modified according to the number of cpu.
Examples of Use
Contents: kaldi / egs / yesno
Training File: http://www.openslr.org/resources/1/waves_yesno.tar.gz
Run the script:
source path.sh export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/root/src/kaldi/tools/openfst/lib ./run.sh
Normal end:
... local/score.sh --cmd utils/run.pl data/test_yesno exp/mono0a/graph_tgpr exp/mono0a/decode_test_yesno local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0 %WER 0.00 [ 0 / 232, 0 ins, 0 del, 0 sub ] exp/mono0a/decode_test_yesno/wer_10_0.0
If you download the source file is too slow or training, I can get from sharing links.
This article addresses corresponding resource: https://pan.baidu.com/s/1ygrlvjgAcw0cXpQpIjNQIQ&shfl=sharepset
After the public may be concerned about micro-channel number 19,101,901 get reply extraction code.
This article github Address: