Speech Recognition in a Web Browser

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 开发者_高级运维building any own plug-ins, is it possible to implement Speech Recognition using Flash ...

---

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).

---

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-speech attribute 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 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).