Web 浏览器中的语音识别
如何通过 Web 浏览器界面实现在线语音识别。
是否可以在不使用任何流服务器（完全基于浏览器的应用程序）的情况下实现这一点？在不构建任何自己的插件的情况下，是否可以使用 Flash 实现语音识别...
How to implement Online Speech Recognition through a Web Browser Interface.
Is it possible to implement this without using any Streaming Server (completely browser-based application)? Without building any own plug-ins, is it possible to implement Speech Recognition using Flash ...
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（3）
我不相信 HTML 支持音频捕获。有一个工作组制作了 http://www.w3.org/TR/xhtml+voice / 但我不相信除了 Opera 之外的任何浏览器都实现了这一点（参见 http:// /dev.opera.com/articles/voice/）。
您可以查看 WAMI 工具包。 WAMI 工具包是来自 MIT 的一个有趣项目 - http://wami.csail.mit.edu/。用他们自己的话来说，“WAMI：可通过 Web 访问的多模式应用程序。WAMI 是一种向任何网页添加语音识别功能的简单方法。” WAMI 为您提供了一个 Java 小程序，可以在您的网页中运行以执行音频捕获以进行语音识别。在他们的示例中，识别是在服务器上执行的。
我怀疑人们可以在浏览器中使用 Flash、ActiveX 或 Java，并在网页中运行本地语音识别。但是，您需要安装识别引擎（或验证该引擎是否可用）。
I don't believe there is HTML support for audio capture. There is a working group that produced http://www.w3.org/TR/xhtml+voice/ but I don't believe this is implemented in any browser except Opera (see http://dev.opera.com/articles/voice/).
You could look at the WAMI toolkit. WAMI toolkit is an interesting project from MIT - http://wami.csail.mit.edu/. In their own words "WAMI: Web-Accessible Multimodal Applications. WAMI is a simple way to add speech recognition capabilities to any web page." WAMI gives you a java applet that can run in your web page to perform audio capture for speech recognition. In their examples, the recognition is performed on the server.
I suspect one could use Flash, ActiveX or Java within the browser and run local speech recognition from within your web page. However, you would need to install a recognition engine (or verify that one is available).
我认为这是不可能的；使用 Flash，您当然可以从用户的麦克风捕获音频，但目前 WebKit 之外无法提供浏览器级识别功能。
旁注：
现在，作为 HTML5 规范的一部分，您可以将
x-webkit-speech属性附加到元素上 - 您可以已经在谷歌翻译中看到了这一点。这似乎是最好的非闪存选项。我发现 http://caniuse.com/#feat=audio-api 来跟踪可能发生的事情来自此评论对一个已解决的问题。
I do not think it is possible; using Flash you can certainly capture audio from the user's microphone but nothing offering browser-level recognition is available at the moment outside of WebKit.
Side note:
Right now, as part of the HTML5 specifications, there is a
x-webkit-speechattribute you can tack on to a<input>element - you might have seen this in action in Google Translate. That seems to be the best non-Flash option available.I found http://caniuse.com/#feat=audio-api to keep track of what is possible where from this comment on a closed question.
pocketsphinx 可以工作，但不足以处理超过几个单词，而且准确性很差。 CSharp 非常出色，但在网页内部署并不容易。好的服务需要远程服务器（Bing API、Google API 等）。
pocketsphinx works but is not good enough to handle more than a few words and the accuracy is poor. CSharp is excellent but it is not easy to deploy inside a webpage. The good services require remote servers (Bing API, Google API, etc).