ROS kinetic语音识别(转)
1.安装依赖
1.1安装ros-kinetic-audio-common
1
|
sudo apt- get install ros-kinetic-audio-common |
1.2 安装libasound2
sudo apt-get install libasound2
1.3 安装libgstreamer0.10
sudo apt-get install gstreamer0.10-*
1.4安装python-gst0.10
sudo apt-get install python-gst0.10
2.安装 libsphinxbase1
下载地址:https://packages.debian.org/jessie/libsphinxbase1
sudo dpkg -i libsphinxbase1_0.8-6_amd64.deb
3. 安装libpocketsphinx1
下载地址:https://packages.debian.org/jessie/libpocketsphinx1
sudo dpkg -i libpocketsphinx1_0.8-5_amd64.deb
4.安装gstreamer0.10-pocketsphinx
下载地址:https://packages.debian.org/jessie/gstreamer0.10-pocketsphinx
sudo dpkg -i gstreamer0.10-pocketsphinx_0.8-5_amd64.deb
5.安装pocketsphinx
5.1进入工作空间,git源码
cd ~/dev/catkin_ws/src
git clone https://github.com/mikeferguson/pocketsphinx
5.2下载英文语音包pocketsphinx-hmm-en-tidigits (0.8-5)
下载的网站:https://packages.debian.org/jessie/pocketsphinx-hmm-en-tidigits
sudo dpkg -i pocketsphinx-hmm-en-tidigits_0.8-5_all.deb
会加压安装到/usr/share/目录如下所示
/usr/share/doc/pocketsphinx-hmm-en-tidigits/changelog.Debian.gz /usr/share/doc/pocketsphinx-hmm-en-tidigits/changelog.gz /usr/share/doc/pocketsphinx-hmm-en-tidigits/copyright /usr/share/pocketsphinx/model/hmm/en/tidigits/feat.params /usr/share/pocketsphinx/model/hmm/en/tidigits/mdef /usr/share/pocketsphinx/model/hmm/en/tidigits/means /usr/share/pocketsphinx/model/hmm/en/tidigits/sendump /usr/share/pocketsphinx/model/hmm/en/tidigits/transition_matrices /usr/share/pocketsphinx/model/hmm/en/tidigits/variances /usr/share/pocketsphinx/model/lm/en/tidigits.DMP /usr/share/pocketsphinx/model/lm/en/tidigits.dic /usr/share/pocketsphinx/model/lm/en/tidigits.fsg
在包pocketsphinx下面建一个model目录,存放语音模型文件
cd ~/dev/catkin_ws/src/pocketsphinx mkdir model
将下载好的语音文件,解压后,将其中的model文件下的所有文件拷贝到~/dev/catkin_ws/src/pocketsphinx/model下,也可以不复制。
6 启动pocketsphinx测试
直接运行pocketsphinx启动脚本命令“roslaunch pocketsphinx robocup.launch”会出现如下错误,
为此,首先我对recognizer.py的做了如下修改:
1)注释掉self.asr.set_property('configured', True)
2)添加lm,dict,hmm支持英语识别(如果是其他语言(如普通话)可以改为别的路径)
self.asr.set_property('lm', '/usr/share/pocketsphinx/model/lm/en/tidigits.DMP') self.asr.set_property('dict', '/usr/share/pocketsphinx/model/lm/en/tidigits.dic') self.asr.set_property('hmm', '/usr/share/pocketsphinx/model/hmm/en/tidigits')
这些安装好之后,可以开始测试pocketsphinx的语音识别功能了:
1.终端运行launch启动文件:
roslaunch pocketsphinx robocup.launch
2.尝试说一些简单的语句,(如果使用英文库,说一些数字;如果使用普通话库讲中文),查看识别结果
3.用rostopic echo进行查看识别的输出,即ROS发布的结果消息)
rostopic echo /recognizer/output
7.也可以下载其他包进行测试
中文普通话
pocketsphinx-hmm-zh-tdt https://packages.debian.org/jessie/pocketsphinx-hmm-zh-tdt
pocketsphinx-lm-zh-hans-gigatdt https://packages.debian.org/jessie/pocketsphinx-lm-zh-hans-gigatdt
后期考虑在launch文件中设置参数,在recognizer.py加载参数的读取
参考:http://blog.csdn.net/x_r_su/article/details/53022746?locationNum=1&fps=1
http://www.ncnynl.com/archives/201701/1220.html