mirror of
https://github.com/modelscope/FunASR
synced 2025-09-15 14:48:36 +08:00
atsr
This commit is contained in:
parent
a2f263bd05
commit
1162dee2dd
@ -1,178 +0,0 @@
|
||||
---
|
||||
tasks:
|
||||
- audio-visual-speech-recognition
|
||||
domain:
|
||||
- audio, visual
|
||||
model-type:
|
||||
- Autoregressive
|
||||
frameworks:
|
||||
- pytorch
|
||||
backbone:
|
||||
- transformer/conformer
|
||||
metrics:
|
||||
- WER/B-WER
|
||||
license: Apache License 2.0
|
||||
language:
|
||||
- en
|
||||
tags:
|
||||
- FunASR
|
||||
- Alibaba
|
||||
- ICASSP 2024
|
||||
- Audio-Visual
|
||||
- Hotword
|
||||
- Long-Context Biasing
|
||||
datasets:
|
||||
train:
|
||||
- SlideSpeech corpus
|
||||
test:
|
||||
- dev and test of SlideSpeech corpus
|
||||
indexing:
|
||||
results:
|
||||
- task:
|
||||
name: Audio-Visual Speech Recognition
|
||||
dataset:
|
||||
name: SlideSpeech corpus
|
||||
type: audio # optional
|
||||
args: 16k sampling rate, 5002 bpe units # optional
|
||||
metrics:
|
||||
- type: WER
|
||||
value: 18.8% # float
|
||||
description: beamsearch search, withou lm, avg.
|
||||
args: default
|
||||
|
||||
widgets:
|
||||
- task: audio-visual-speech-recognition
|
||||
inputs:
|
||||
- type: audio
|
||||
name: input
|
||||
title: 音频
|
||||
- type: text
|
||||
name: input
|
||||
title: OCR识别文本
|
||||
finetune-support: True
|
||||
---
|
||||
|
||||
|
||||
# Paraformer-large模型介绍
|
||||
|
||||
## Highlights
|
||||
- 热词版本:[Paraformer-large热词版模型](https://www.modelscope.cn/models/damo/speech_paraformer-large-contextual_asr_nat-zh-cn-16k-common-vocab8404/summary)支持热词定制功能,基于提供的热词列表进行激励增强,提升热词的召回率和准确率。
|
||||
- 长音频版本:[Paraformer-large长音频模型](https://www.modelscope.cn/models/damo/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch/summary),集成VAD、ASR、标点与时间戳功能,可直接对时长为数小时音频进行识别,并输出带标点文字与时间戳。
|
||||
|
||||
## <strong>[FunASR开源项目介绍](https://github.com/alibaba-damo-academy/FunASR)</strong>
|
||||
<strong>[FunASR](https://github.com/alibaba-damo-academy/FunASR)</strong>希望在语音识别的学术研究和工业应用之间架起一座桥梁。通过发布工业级语音识别模型的训练和微调,研究人员和开发人员可以更方便地进行语音识别模型的研究和生产,并推动语音识别生态的发展。让语音识别更有趣!
|
||||
|
||||
[**github仓库**](https://github.com/alibaba-damo-academy/FunASR)
|
||||
| [**最新动态**](https://github.com/alibaba-damo-academy/FunASR#whats-new)
|
||||
| [**环境安装**](https://github.com/alibaba-damo-academy/FunASR#installation)
|
||||
| [**服务部署**](https://www.funasr.com)
|
||||
| [**模型库**](https://github.com/alibaba-damo-academy/FunASR/tree/main/model_zoo)
|
||||
| [**联系我们**](https://github.com/alibaba-damo-academy/FunASR#contact)
|
||||
|
||||
|
||||
## 模型原理介绍
|
||||
|
||||
随着在线会议和课程越来越普遍,如何利用视频幻灯片中丰富的文本信息来改善语音识别(Automatic Speech Recognition, ASR)面临着新的挑战。视频中的幻灯片与语音实时同步,相比于统一的稀有词列表,能够提供更长的上下文相关信息。因此,我们提出了一种创新的长上下文偏置网络(LCB-net),用于音频-视觉语音识别(Audio-Visual Speech Recognition,AVSR),以更好地利用视频中的长时上下文信息。
|
||||
|
||||
<p align="center">
|
||||
<img src="fig/lcbnet1.png" alt="AVSR整体流程框架" width="800" />
|
||||
<p align="center">
|
||||
<img src="fig/lcbnet2.png" alt="LCB-NET模型结构" width="800" />
|
||||
|
||||
|
||||
具体来说,我们首先使用OCR技术来检测和识别幻灯片中的文本内容,其次我们采用关键词提取技术来获取文本内容中的关键词短语。最后,我们将关键词拼接成长上下文文本和音频同时输入到我们的LCB-net模型中进行识别。而LCB-net模型采用了双编码器结构,同时建模音频和长上下文文本信息。此外,我们还引入了一个显式的偏置词预测模块,通过使用二元交叉熵(BCE)损失函数显式预测长上下文文本中在音频中出现的关键偏置词。此外,为增强LCB-net的泛化能力和稳健性,我们还采用了动态的关键词模拟策略。实验证明,我们提出的LCB-net热词模型,不仅能够提升关键词的识别效果,同时也能够提升非关键词的识别效果。具体实验结果如下所示:
|
||||
|
||||
<p align="center">
|
||||
<img src="fig/lcbnet3.png" alt="实验结果" width="500" />
|
||||
|
||||
|
||||
更详细的细节见:
|
||||
- 论文: [LCB-net: Long-Context Biasing for Audio-Visual Speech Recognition](https://arxiv.org/abs/2401.06390)
|
||||
|
||||
|
||||
|
||||
## 基于ModelScope进行推理
|
||||
|
||||
- 推理支持音频格式如下:
|
||||
- wav文件路径,例如:data/test/asr_example.wav
|
||||
- pcm文件路径,例如:data/test/asr_example.pcm
|
||||
- ark文件路径,例如:data/test/data.ark
|
||||
- wav文件url,例如:https://www.modelscope.cn/api/v1/models/iic/LCB-NET/repo?Revision=master&FilePath=example/asr_example.wav
|
||||
- wav二进制数据,格式bytes,例如:用户直接从文件里读出bytes数据或者是麦克风录出bytes数据。
|
||||
- 已解析的audio音频,例如:audio, rate = soundfile.read("asr_example_zh.wav"),类型为numpy.ndarray或者torch.Tensor。
|
||||
- wav.scp文件,需符合如下要求(以下分别为sound和kaldi_ark格式):
|
||||
|
||||
```sh
|
||||
cat wav.scp
|
||||
asr_example1 data/test/asr_example1.wav
|
||||
asr_example2 data/test/asr_example2.wav
|
||||
|
||||
cat wav.scp
|
||||
asr_example1 data/test/data_wav.ark:22
|
||||
asr_example2 data/test/data_wav.ark:90445
|
||||
...
|
||||
```
|
||||
|
||||
- 推理支持OCR预测文本格式如下:
|
||||
- ocr.txt文件,需符合如下要求:
|
||||
```sh
|
||||
cat ocr.txt
|
||||
asr_example1 ANIMAL <blank> RIGHTS <blank> MANAGER <blank> PLOEG
|
||||
asr_example2 UNIVERSITY <blank> CAMPUS <blank> DEANO
|
||||
...
|
||||
```
|
||||
|
||||
- 若输入格式wav文件和ocr文件均为url,api调用方式可参考如下范例:
|
||||
|
||||
```python
|
||||
from funasr import AutoModel
|
||||
|
||||
model = AutoModel(model="iic/LCB-NET",
|
||||
model_revision="v2.0.0")
|
||||
res = model.generate(input=("https://www.modelscope.cn/api/v1/models/iic/LCB-NET/repo?Revision=master&FilePath=example/asr_example.wav","https://www.modelscope.cn/api/v1/models/iic/LCB-NET/repo?Revision=master&FilePath=example/ocr.txt"),data_type=("sound", "text"))
|
||||
```
|
||||
|
||||
|
||||
## 复现论文中的结果
|
||||
```python
|
||||
python -m funasr.bin.inference \
|
||||
--config-path=${file_dir} \
|
||||
--config-name="config.yaml" \
|
||||
++init_param=${file_dir}/model.pt \
|
||||
++tokenizer_conf.token_list=${file_dir}/tokens.txt \
|
||||
++input=[${_logdir}/wav.scp,${_logdir}/ocr.txt] \
|
||||
+data_type='["kaldi_ark", "text"]' \
|
||||
++tokenizer_conf.bpemodel=${file_dir}/bpe.pt \
|
||||
++output_dir="${inference_dir}/results" \
|
||||
++device="${inference_device}" \
|
||||
++ncpu=1 \
|
||||
++disable_log=true
|
||||
|
||||
```
|
||||
|
||||
|
||||
识别结果输出路径结构如下:
|
||||
|
||||
```sh
|
||||
tree output_dir/
|
||||
output_dir/
|
||||
└── 1best_recog
|
||||
├── text
|
||||
└── token
|
||||
```
|
||||
|
||||
token:语音识别结果文件
|
||||
|
||||
可以使用funasr里面提供的run_bwer_recall.sh计算WER、BWER、UWER和Recall。
|
||||
详细脚本可以参考funasr里面的demo.sh脚本,需要注意的是你需要修改一下iic/LCB-NET/conf.yaml中CMVN(stats_file)的路径和iic/LCB-NET/dev/wav.scp里面ark的路径,修改为你自己本地的路径,然后跑解码。
|
||||
|
||||
## 相关论文以及引用信息
|
||||
|
||||
```BibTeX
|
||||
@inproceedings{yu2024lcbnet,
|
||||
title={LCB-net: Long-Context Biasing for Audio-Visual Speech Recognition},
|
||||
author={Fan Yu, Haoxu Wang, Xian Shi, Shiliang Zhang},
|
||||
booktitle={ICASSP},
|
||||
year={2024}
|
||||
}
|
||||
```
|
||||
Loading…
Reference in New Issue
Block a user