update:test迁移重命名为digital-human
@@ -0,0 +1,41 @@
|
||||
本文档是开发类文档,如需部署小智服务端,[点击这里查看部署教程](../../README.md#%E9%83%A8%E7%BD%B2%E6%96%87%E6%A1%A3)
|
||||
|
||||
如需查看唤醒词模型下载、运行时配置和详细使用说明,[点击这里查看专题文档](../../docs/digital-human-wakeword.md)
|
||||
|
||||
# 项目介绍
|
||||
|
||||
digital-human 是独立的数字人测试模块,负责提供本地测试页面、前端交互资源、唤醒词运行时和事件桥能力,用于联调整个数字人交互链路。
|
||||
|
||||
# 快速启动
|
||||
|
||||
安装依赖:
|
||||
|
||||
```bash
|
||||
pip install -r wakeword_runtime/requirements.txt
|
||||
```
|
||||
|
||||
启动模块:
|
||||
|
||||
```bash
|
||||
python start.py
|
||||
```
|
||||
|
||||
# 访问地址
|
||||
|
||||
启动后可访问:
|
||||
|
||||
- 页面地址:http://127.0.0.1:8006/index.html
|
||||
- 事件桥地址:ws://127.0.0.1:8006/wakeword-ws
|
||||
- 健康检查:http://127.0.0.1:8006/health
|
||||
|
||||
# 目录说明
|
||||
|
||||
- `start.py`:模块启动入口
|
||||
- `index.html`:数字人测试页面入口
|
||||
- `wakeword_runtime`:本地唤醒词运行时与配置目录
|
||||
- `js`、`css`:页面前端脚本与样式
|
||||
- `images`、`resources`:页面资源文件
|
||||
|
||||
# 相关文档
|
||||
|
||||
唤醒词专题文档:[../../docs/digital-human-wakeword.md](../../docs/digital-human-wakeword.md)
|
||||
|
Before Width: | Height: | Size: 185 B After Width: | Height: | Size: 185 B |
|
Before Width: | Height: | Size: 4.2 KiB After Width: | Height: | Size: 4.2 KiB |
|
Before Width: | Height: | Size: 136 KiB After Width: | Height: | Size: 136 KiB |
|
Before Width: | Height: | Size: 321 KiB After Width: | Height: | Size: 321 KiB |
|
Before Width: | Height: | Size: 290 KiB After Width: | Height: | Size: 290 KiB |
@@ -24,11 +24,11 @@
|
||||
<p>您可以使用nginx映射启动测试页面,也可以请按照以下步骤启动测试运行时:</p>
|
||||
<ol>
|
||||
<li>打开命令行终端</li>
|
||||
<li>命令行进入到 xiaozhi-server/test 目录</li>
|
||||
<li>命令行进入到 digital-human 目录</li>
|
||||
<li>执行以下命令启动测试页面运行时:</li>
|
||||
</ol>
|
||||
<pre>python start_test_runtime.py</pre>
|
||||
<p>然后在浏览器中访问:<strong>http://localhost:8006/test_page.html</strong></p>
|
||||
<pre>python start.py</pre>
|
||||
<p>然后在浏览器中访问:<strong>http://localhost:8006/index.html</strong></p>
|
||||
`;
|
||||
document.body.appendChild(warningDiv);
|
||||
});
|
||||
|
Before Width: | Height: | Size: 1.7 MiB After Width: | Height: | Size: 1.7 MiB |
|
Before Width: | Height: | Size: 2.4 MiB After Width: | Height: | Size: 2.4 MiB |
|
Before Width: | Height: | Size: 6.6 MiB After Width: | Height: | Size: 6.6 MiB |
@@ -25,7 +25,7 @@ class TestRuntimeHttpServer:
|
||||
|
||||
@property
|
||||
def page_url(self) -> str:
|
||||
return f"http://127.0.0.1:{self.port}/test_page.html"
|
||||
return f"http://127.0.0.1:{self.port}/index.html"
|
||||
|
||||
@property
|
||||
def bridge_url(self) -> str:
|
||||
@@ -1,221 +0,0 @@
|
||||
# Test 页面语音唤醒
|
||||
|
||||
## 概述
|
||||
|
||||
Test 页面集成了基于 **Sherpa-ONNX** 的高精度语音唤醒功能,支持自定义唤醒词和实时检测。使用轻量级关键词检测模型,提供毫秒级响应速度。
|
||||
|
||||
## 唤醒词模型
|
||||
|
||||
### 模型下载(必需)
|
||||
|
||||
**重要说明**: 项目不包含模型文件,需要提前下载配置。
|
||||
|
||||
### 官方模型下载地址
|
||||
|
||||
- **官方模型列表**: <https://csukuangfj.github.io/sherpa/onnx/kws/pretrained_models/index.html>
|
||||
- **推荐模型**: `sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01`
|
||||
|
||||
### 下载和配置步骤
|
||||
|
||||
#### 1. 下载模型包
|
||||
|
||||
```bash
|
||||
# 方法1:直接下载(推荐)
|
||||
cd main/xiaozhi-server/test/wakeword_runtime/
|
||||
wget https://github.com/k2-fsa/sherpa-onnx/releases/download/kws-models/sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2
|
||||
|
||||
# 解压
|
||||
tar xvf sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.tar.bz2
|
||||
|
||||
# 方法2:使用ModelScope
|
||||
pip install modelscope
|
||||
python -c "
|
||||
from modelscope import snapshot_download
|
||||
snapshot_download('pkufool/sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01', cache_dir='./models')
|
||||
"
|
||||
```
|
||||
|
||||
#### 2. 配置模型文件
|
||||
|
||||
模型包下载后包含以下文件:
|
||||
|
||||
```
|
||||
sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01/
|
||||
├── encoder-epoch-12-avg-2-chunk-16-left-64.int8.onnx # 速度优先
|
||||
├── encoder-epoch-12-avg-2-chunk-16-left-64.onnx #
|
||||
├── encoder-epoch-99-avg-1-chunk-16-left-64.int8.onnx # 速度优先
|
||||
├── encoder-epoch-99-avg-1-chunk-16-left-64.onnx # 精度优先
|
||||
├── decoder-epoch-12-avg-2-chunk-16-left-64.onnx #
|
||||
├── decoder-epoch-99-avg-1-chunk-16-left-64.onnx # 精度优先
|
||||
├── joiner-epoch-12-avg-2-chunk-16-left-64.int8.onnx # 速度优先
|
||||
├── joiner-epoch-12-avg-2-chunk-16-left-64.onnx #
|
||||
├── joiner-epoch-99-avg-1-chunk-16-left-64.int8.onnx # 速度优先
|
||||
├── joiner-epoch-99-avg-1-chunk-16-left-64.onnx # 精度优先
|
||||
├── tokens.txt # Token映射表(必需)
|
||||
├── keywords_raw.txt # 模型包里可能附带(可选,runtime 不依赖)
|
||||
├── keywords.txt # 现成的
|
||||
├── test_wavs/ # 测试音频(可选)
|
||||
├── configuration.json # 模型元信息(可选)
|
||||
└── README.md # 说明文档(可选)
|
||||
```
|
||||
|
||||
#### 3. 选择配置方案
|
||||
|
||||
**方案一:精度优先(推荐)**
|
||||
|
||||
```bash
|
||||
cd sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01
|
||||
|
||||
# 创建模型目录
|
||||
mkdir -p ../models
|
||||
|
||||
# 复制精度优先的epoch-99 fp32三件套
|
||||
cp encoder-epoch-99-avg-1-chunk-16-left-64.onnx ../models/encoder.onnx
|
||||
cp decoder-epoch-99-avg-1-chunk-16-left-64.onnx ../models/decoder.onnx
|
||||
cp joiner-epoch-99-avg-1-chunk-16-left-64.onnx ../models/joiner.onnx
|
||||
|
||||
# 复制配套文件
|
||||
cp tokens.txt ../models/tokens.txt
|
||||
# keywords_raw.txt 如果模型包里附带,可自行保留;test runtime 不依赖它
|
||||
```
|
||||
|
||||
**方案二:速度优先**
|
||||
|
||||
```bash
|
||||
cd sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01
|
||||
|
||||
# 创建模型目录
|
||||
mkdir -p ../models
|
||||
|
||||
# 复制速度优先的epoch-99 int8三件套
|
||||
cp encoder-epoch-99-avg-1-chunk-16-left-64.int8.onnx ../models/encoder.onnx
|
||||
cp decoder-epoch-99-avg-1-chunk-16-left-64.onnx ../models/decoder.onnx
|
||||
cp joiner-epoch-99-avg-1-chunk-16-left-64.int8.onnx ../models/joiner.onnx
|
||||
|
||||
# 复制配套文件
|
||||
cp tokens.txt ../models/tokens.txt
|
||||
```
|
||||
|
||||
**注意事项**:
|
||||
|
||||
- **不要混用fp32与int8**:三个模型文件必须保持一致的精度
|
||||
- **优先选择epoch-99**:比epoch-12训练更充分,精度更高
|
||||
- **必需文件**:`encoder.onnx` + `decoder.onnx` + `joiner.onnx` + `tokens.txt` + `keywords.txt`
|
||||
|
||||
### 最终模型文件结构
|
||||
|
||||
配置完成后,模型文件应放在 `wakeword_runtime/models/` 目录下(完整路径:`main/xiaozhi-server/test/wakeword_runtime/models/`):
|
||||
|
||||
```
|
||||
wakeword_runtime/models/
|
||||
├── encoder.onnx # 编码器模型(重命名后)
|
||||
├── decoder.onnx # 解码器模型(重命名后)
|
||||
├── joiner.onnx # 连接器模型(重命名后)
|
||||
├── tokens.txt # 拼音Token映射表(228行版本)
|
||||
├── keywords.txt # 关键词配置文件(首次启动自动生成)
|
||||
└── keywords_raw.txt # 可选,runtime 不依赖
|
||||
```
|
||||
|
||||
## 启动方式
|
||||
|
||||
在 `main/xiaozhi-server/test` 目录执行:
|
||||
|
||||
```bash
|
||||
pip install -r wakeword_runtime/requirements.txt
|
||||
python start_test_runtime.py
|
||||
```
|
||||
|
||||
启动后默认地址:
|
||||
|
||||
- 页面地址:`http://127.0.0.1:8006/test_page.html`
|
||||
- 事件桥地址:`ws://127.0.0.1:8006/wakeword-ws`
|
||||
- 健康检查:`http://127.0.0.1:8006/health`
|
||||
|
||||
停止方式:
|
||||
|
||||
- 在运行终端按 `Ctrl+C`
|
||||
- 会同时停止静态页面服务、事件桥和唤醒词检测流程
|
||||
|
||||
|
||||
## 配置文件说明
|
||||
|
||||
配置文件位于 [config.json](./config.json)。
|
||||
|
||||
当前主要配置项:
|
||||
|
||||
```json
|
||||
{
|
||||
"wakeword": {
|
||||
"enabled": true
|
||||
},
|
||||
"model_dir": "models",
|
||||
"audio": {
|
||||
"input_device": null,
|
||||
"sample_rate": 16000,
|
||||
"channels": 1
|
||||
},
|
||||
"detector": {
|
||||
"num_threads": 4,
|
||||
"provider": "cpu",
|
||||
"max_active_paths": 2,
|
||||
"keywords_score": 1.8,
|
||||
"keywords_threshold": 0.1,
|
||||
"num_trailing_blanks": 1,
|
||||
"cooldown_seconds": 1.5
|
||||
},
|
||||
"logging": {
|
||||
"level": "INFO",
|
||||
"dir": "logs",
|
||||
"file": "wakeword-runtime.log"
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
各字段含义:
|
||||
|
||||
| 参数 | 说明 |
|
||||
| --- | --- |
|
||||
| `wakeword.enabled` | 是否启用本地唤醒词检测 |
|
||||
| `model_dir` | 模型和词表所在目录 |
|
||||
| `audio.input_device` | 麦克风输入设备,默认使用系统默认设备 |
|
||||
| `audio.sample_rate` | 采样率,默认 `16000` |
|
||||
| `audio.channels` | 声道数,默认 `1` |
|
||||
| `detector.num_threads` | 检测器线程数 |
|
||||
| `detector.provider` | 推理 provider,当前通常为 `cpu` |
|
||||
| `detector.max_active_paths` | 搜索路径数 |
|
||||
| `detector.keywords_score` | 关键词增强分数 |
|
||||
| `detector.keywords_threshold` | 检测阈值 |
|
||||
| `detector.num_trailing_blanks` | 尾随空白数量 |
|
||||
| `detector.cooldown_seconds` | 连续触发冷却时间 |
|
||||
| `logging.level` | 日志等级 |
|
||||
| `logging.dir` | 日志目录 |
|
||||
| `logging.file` | 日志文件名 |
|
||||
|
||||
## 推荐使用流程
|
||||
|
||||
### 首次使用
|
||||
|
||||
1. 准备 `models/` 目录下的模型文件和 `tokens.txt`
|
||||
2. 确认 `models/keywords.txt` 存在
|
||||
3. 在 `test` 目录运行 `python start_test_runtime.py`
|
||||
4. 浏览器打开 `http://127.0.0.1:8006/test_page.html`
|
||||
5. 进入设置页检查“唤醒词”配置
|
||||
|
||||
### 修改唤醒词
|
||||
|
||||
1. 打开 test 页面设置
|
||||
2. 切到“唤醒词”页签
|
||||
3. 修改启用状态或唤醒词列表
|
||||
4. 点击“应用唤醒词”
|
||||
5. 根据提示决定是否立即重启
|
||||
|
||||
### 禁用唤醒词
|
||||
|
||||
1. 将“启用本地唤醒词”改成禁用
|
||||
2. 点击“应用唤醒词”
|
||||
3. 建议立即重启一次
|
||||
|
||||
禁用后:
|
||||
|
||||
- 页面与事件桥仍然可用
|
||||
- 唤醒词检测不会继续运行
|
||||