使用 sphinx4 识别所有英语单词
我在使用 sphinx4 识别所有英语单词时遇到问题。我将所有以字母“A”开头的英语单词添加到我的java语法文件中并尝试运行它。第一次它给了我java堆空间的错误，我最初保留了256m。 然后我把它改成512m。这次 java 没有出现错误，但是识别器没有识别出任何单词，甚至没有识别出错误的单词。 任何人都可以帮我制作一个包含所有英语单词的语音识别语法文件...
这是我的第一个问题，所以希望能得到一些好的结果。
I have a problem in recognizing all english words using sphinx4. I added all the english words starting with letter 'A' to my grammar file in java and tried to run it. First time it gave me the error of java heap space which i initially kept 256m.
Then i changed it to 512m. this time there were no errors from java, but the recognizer was not recognizing a single word not even wrong words.
Can any one help me in making a speech recognition grammar file with all english words...
This is my first question here so hoping to get the some good results.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
由于您使用的是 SPhinx 4。很多事情取决于您的配置方式。 Sphinx（袖珍 sphinx 和 Sphinx 4）提供转录和即时转换。我的意思是，您可以将声音保存为 wav 文件并执行 wav 到文本的转换，或者您可以简单地说话，保持 sphinx4 应用程序运行，这样它就会转录您所说的每一行。
您必须像使用 2/3/n gram 文件一样配置语法文件。提供基于发音的声学模型，如美国英语的 hub4 声学模型或印度英语的 voxforge 模型。我之前发布过类似的问题
sphinx4 中的语音识别响应很差
您还可以参考此网站http://nsh.nexiwave.com/，它提供了所有必要的信息配置 Sphinx4 的详细信息。基于这些配置我想你应该能够获得有尊严的输出
Since you are using SPhinx 4. A lot of the things depends on the way you have configured. Sphinx (both pocket sphinx and Sphinx 4) offers transcription as well as immediate conversion. I mean, you can save your voice as a wav file and perform a wav to text conversion or you can simply speak keeping the sphinx4 application running so it will transcribe every line that u speak.
You would have to configure the grammar file as in using a 2/3/n gram file. Provide the acoustic model based on the pronunciation as in hub4 acoustic model for US english or voxforge models for indian english. I had earlier posted a similar question
Speech recognition response is poor in sphinx4
You can also refer to a this site http://nsh.nexiwave.com/ which provides all the necessary details for configuring Sphinx4 . based on these configurations i suppose you should be able to get a dignified output