Add support for Speech Recognition

[kubawolanin]

Hello!
I've been thinking about implementing speech recognition capability in HABPanel.
I'm utilizing webkitSpeechRecognition for that (but may end up with SpeechRecognition)
So far I've managed to add an icon to the dashboard header and display the speech output on the top, like this:

The "bullhorn" icon is temporary - apparently bootstrap's Glyphicon collection doesn't have any microphone icon ;-)

The idea is to display the output text at the time of speaking - after the result is final, it should hide the <span> element containing the text and then pass the output to the VoiceCommand item in openHAB.

Now two questions:

1. Is it possible for me to get Voice Command Item (from the Rule Voice Interpreter setting)? Ideally, I'd like to retrieve it from REST, but couldn't find it. If it's not possible, then we'd need our users to select it in Settings.
2. @ghys would you mind me extending current SpeechService or you'd rather see it somewhere else? Currently I'm coding simple POC inside dashboard.view.controller but I'm thinking on better place for the logic.

Thank you,
Kuba

[kaikreuzer]

1.) You should note that VoiceCommand is outdated by now, see here. So I would suggest to directly use the new REST endpoint (and you then do not have to rely on the existence of a specific icon).

[kubawolanin]

Hi @kaikreuzer thanks for the response.
I wasn't aware of that, so I'll take a look. At a glance it looks exactly what I'm looking for :)
As for the icon - I'm not relying on glyphicons ;) svg to the rescue!

[ghys]

Cool, I actually considered doing it along with the TTS support but due to a number of potential issues I figured it could be delayed.
So I'll share my thoughts from that time:

• Safari doesn't support it, therefore no iOS, no IE/Edge/Firefox either, only Chrome: see http://caniuse.com/#search=speech - in any case the icon should be hidden if the feature is not available for any reason;
• be aware there might not be room to display the text (think phones and tablets in portrait mode), maybe an overlay is more appropriate.

Extending the SpeechService with STT looks good to me.

[xylo]

This sounds like a nice feature. Do you know if the speech recognition works offline (not server-based)?

[kubawolanin]

Hey @xylo
No, unfortunately. I'm targeting webkit-based browsers here, meaning that it will work only on Chrome (ver >=33). And AFAIK Chrome is utilizing Cloud Speech API for that.
I didn't test it on Firefox, because I'm using webkitSpeechRecognition either way, but according to Mozilla's docs on that matter it could work if enabled in browser's settings.

[xylo]

@kubawolanin
Thanks for the info.

[kubawolanin]

Implemented in #95 - closing this issue.