kaldi data preparation
来源:互联网 发布:windows错误1503 编辑:程序博客网 时间:2024/06/06 16:03
主要两个文件夹data/train和data/lang
train
需要手动创建三个文件
- utt2spk
- text
- wav.scp
以上文件需要提前按照C++方式排序
export LC_ALL=C
然后可以调用steps下的脚本抽特征
steps/make_mfcc.sh --nj 20 --cmd "$train_cmd" data/train exp/make_mfcc/train $mfccdirsteps/compute_cmvn_stats.sh data/train exp/make_mfcc/train $mfccdir
得到两个文件:
- train/feats.scp
- train/cmvn.scp
lang
需要提前准备的文件data/local/dict:
- extra_questions.txt
- lexicon.txt
- nonsilence_phones.txt
- optional_silence.txt
- silence_phones.txt
运行一下脚本生成data/lang
utils/prepare_lang.sh data/local/dict "<UNK>" data/local/lang data/lang
0 0
- kaldi data preparation
- Kaldi nnet3 -------- Data Type
- Preparation
- Library Data Preparation for ICC---1
- 171026 data preparation for cooling-bin
- kaldi
- Kaldi
- Database Testing – Properties of a Good Test Data and Test Data Preparation Techniques
- IBM Certificate Roadmaphttp://www-306.ibm.com/software/data/education/cert/preparation.html
- Effective preparation
- webex preparation
- Problem Preparation
- Problem Preparation
- Problem Preparation
- Preparation for OpenGL programming
- SCEA preparation 1
- Business Objects Certification Preparation
- Preparation and Practice
- Redhat6.5利用yum快速搭建LAMP环境
- 强化学习(Reinforcement Learning, RL)初步介绍
- RMQ算法
- Codeforces Round #297 (Div. 2)E. Anya and Cubes 折半搜索
- PAT甲级1001. A+B Format (20)
- kaldi data preparation
- 大数据江湖之即席查询与分析(下篇)--手把手教你搭建即席查询与分析Demo
- ArcEngine 释放锁文件,彻底移除图层
- 在本地用命令行创建一个git仓库,并推送到远程
- 教你如何搭建一个超完美的服务端渲染开发环境
- 生成式对抗网络GAN汇总
- 层次聚类算法(一)
- Struts2框架
- java 加密之消息摘要算法