# A MATLAB implementation of CHiME4 baseline Beamformit
**Please let me know if there are any bugs.**
[email protected]
## References
- "Acoustic beamforming for speaker diarization of meetings", Xavier Anguera, Chuck Wooters and Javier Hernando, IEEE Transactions on Audio, Speech and Language Processing, September 2007, volume 15, number 7, pp.2011-2023.
- [official beamformit github](https://github.com/xanguera/BeamformIt)
## Requirements
| script | requirement |
|---|---|
| beamformit.m | MATLAB supporting audioread (also you can use on OCTAVE by installing signal package) |
| beamformit_step_by_step.mlx | MATLAB supporting mlx format |
## Implementation detail
See beamformit_step_by_step\*.html (and beamformit_step_by_step.mlx)
## How to run
See beamformit.m
## Result
![](sample11.png)
```sh
## original version
local/chime4_calc_wers.sh exp/tri3b_tr05_multi_noisy beamformit_5mics exp/tri3b_tr05_multi_noisy/graph_tgpr_5k
-------------------
best overall dt05 WER 13.66% (language model weight = 11)
-------------------
dt05_simu WER: 14.34% (Average), 12.82% (BUS), 17.09% (CAFE), 11.90% (PEDESTRIAN), 15.56% (STREET)
-------------------
dt05_real WER: 12.98% (Average), 15.96% (BUS), 12.67% (CAFE), 10.02% (PEDESTRIAN), 13.26% (STREET)
-------------------
et05_simu WER: 21.33% (Average), 15.75% (BUS), 22.97% (CAFE), 22.54% (PEDESTRIAN), 24.06% (STREET)
-------------------
et05_real WER: 21.80% (Average), 30.08% (BUS), 20.62% (CAFE), 19.90% (PEDESTRIAN), 16.62% (STREET)
-------------------
## my version
local/chime4_calc_wers.sh exp/tri3b_tr05_multi_noisy bfit_1026_final exp/tri3b_tr05_multi_noisy/graph_tgpr_5k
compute dt05 WER for each location
-------------------
best overall dt05 WER 13.69% (language model weight = 11)
-------------------
dt05_simu WER: 14.31% (Average), 12.86% (BUS), 17.11% (CAFE), 11.90% (PEDESTRIAN), 15.37% (STREET)
-------------------
dt05_real WER: 13.07% (Average), 16.26% (BUS), 12.74% (CAFE), 9.84% (PEDESTRIAN), 13.45% (STREET)
-------------------
et05_simu WER: 21.85% (Average), 15.93% (BUS), 23.50% (CAFE), 23.44% (PEDESTRIAN), 24.52% (STREET)
-------------------
et05_real WER: 22.16% (Average), 30.72% (BUS), 20.88% (CAFE), 20.20% (PEDESTRIAN), 16.87% (STREET)
-------------------
```
没有合适的资源?快使用搜索试试~ 我知道了~
A MATLAB implementation of CHiME4 baseline Beamformit.zip
共39个文件
wav:21个
html:3个
mlx:3个
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 163 浏览量
2023-07-21
20:09:12
上传
评论
收藏 4.33MB ZIP 举报
温馨提示
A MATLAB implementation of CHiME4 baseline Beamformit.zip
资源推荐
资源详情
资源评论
收起资源包目录
A MATLAB implementation of CHiME4 baseline Beamformit.zip (39个子文件)
beamformit_matlab-master
sample13
M06_442C020F_BUS.CH3.wav 197KB
enhan.wav 197KB
M06_442C020F_BUS.CH4.wav 197KB
.log.swp 64KB
orig_enhan.wav 197KB
M06_442C020F_BUS.CH6.wav 197KB
M06_442C020F_BUS.CH1.wav 197KB
wav.list 175B
M06_442C020F_BUS.CH5.wav 197KB
log 4.82MB
beamformit_step_by_step_sample13_why_xcorr_diff.html 585KB
beamformit_step_by_step_sample12_not_same.mlx 214KB
beamformit.m 23KB
sample11
enhan.wav 115KB
M05_443C0207_PED.CH6.wav 115KB
M05_443C0207_PED.CH3.wav 115KB
M05_443C0207_PED.CH4.wav 115KB
orig_enhan.wav 115KB
M05_443C0207_PED.CH1.wav 115KB
M05_443C0207_PED.CH5.wav 115KB
wav.list 175B
log 202KB
sample12
enhan.wav 54KB
F04_421C0207_BUS.CH6.wav 54KB
F04_421C0207_BUS.CH1.wav 54KB
orig_enhan.wav 54KB
F04_421C0207_BUS.CH5.wav 54KB
F04_421C0207_BUS.CH3.wav 54KB
F04_421C0207_BUS.CH4.wav 54KB
wav.list 175B
log 114KB
sample11.png 22KB
beamformit_step_by_step_sample13_why_xcorr_diff.mlx 235KB
1811xx_한국음성학회_권해용.pdf 626KB
csvimport.m 13KB
beamformit_step_by_step_sample12_not_same.html 534KB
beamformit_step_by_step_sample11_almost_same.html 555KB
README.md 2KB
beamformit_step_by_step_sample11_almost_same.mlx 226KB
新建文件夹
共 39 条
- 1
资源评论
AbelZ_01
- 粉丝: 894
- 资源: 5441
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功