xvector-cnceleb
所属分类:模式识别(视觉/语音等)
开发工具:Shell
文件大小:0KB
下载次数:0
上传日期:2020-09-22 14:10:25
上 传 者:
sh-1993
说明: 基于kaldi的x向量,在Cn Celeb上训练,
(kaldi based x-vector trained on Cn-Celeb,)
文件列表:
cmd.sh (827, 2020-09-22)
conf/ (0, 2020-09-22)
conf/mfcc.conf (225, 2020-09-22)
conf/vad.conf (110, 2020-09-22)
exp/ (0, 2020-09-22)
exp/.DS_Store (6148, 2020-09-22)
exp/result.sh (269, 2020-09-22)
local/ (0, 2020-09-22)
local/.DS_Store (8196, 2020-09-22)
local/aishell_data_prep_ft.sh (2124, 2020-09-22)
local/make_cnceleb.sh (1874, 2020-09-22)
local/make_musan.py (4103, 2020-09-22)
local/make_musan.sh (1262, 2020-09-22)
local/make_voxceleb1.pl (4644, 2020-09-22)
local/make_voxceleb2.pl (2272, 2020-09-22)
local/nnet3/ (0, 2020-09-22)
local/nnet3/.DS_Store (8196, 2020-09-22)
local/nnet3/xvector/ (0, 2020-09-22)
local/nnet3/xvector/.DS_Store (6148, 2020-09-22)
local/nnet3/xvector/prepare_feats_for_egs.sh (2748, 2020-09-22)
local/nnet3/xvector/run_xvector.sh (6357, 2020-09-22)
local/nnet3/xvector/tuning/ (0, 2020-09-22)
local/nnet3/xvector/tuning/run_xvector_1a.sh (6222, 2020-09-22)
local/prepare_for_eer.py (587, 2020-09-22)
path.sh (343, 2020-09-22)
run_cnceleb.sh (6566, 2020-09-22)
sid/ (0, 2020-09-22)
sid/.DS_Store (6148, 2020-09-22)
sid/compute_min_dcf.py (6158, 2020-09-22)
sid/compute_vad_decision.sh (2240, 2020-09-22)
sid/compute_vad_decision_gmm.sh (5213, 2020-09-22)
sid/extract_ivectors.sh (4135, 2020-09-22)
sid/extract_ivectors_dnn.sh (4371, 2020-09-22)
sid/gender_id.sh (6626, 2020-09-22)
sid/init_full_ubm_from_dnn.sh (4337, 2020-09-22)
sid/music_id.sh (4815, 2020-09-22)
sid/nnet2/ (0, 2020-09-22)
sid/nnet2/get_egs2.sh (16285, 2020-09-22)
sid/nnet2/get_lda.sh (7891, 2020-09-22)
... ...
# xvector-cnceleb
kaldi based x-vector trained on Cn-Celeb database.
# Kaldi configuration
export KALDI_ROOT=/kaldi/dir
# Usage
bash run_cnceleb.sh
# Result
- x-vector(cn-celeb) + PLDA(cn-celeb)
- CN-Celeb Eval Core:
- EER: 16.71%
- minDCF(p-target=0.01): 0.7657
- minDCF(p-target=0.001): 0.8823
- x-vector(voxceleb) + PLDA(cn-celeb)
- CN-Celeb Eval Core:
- EER: 12.43%
- minDCF(p-target=0.01): 0.6064
- minDCF(p-target=0.001): 0.7381
# References:
- https://www.danielpovey.com/files/2018_icassp_xvectors.pdf(X-VECTORS)
- https://arxiv.org/pdf/1911.01799.pdf(Cn-Celeb)
- Voxceleb database:http://openslr.org/49/
- 开源voxceleb模型:https://kaldi-asr.org/models/m7
- CN-Celeb database:http://openslr.org/82/
近期下载者:
相关文件:
收藏者: