Can OpenEars Outputs Anything I Say?

[pmpascua]

Hi, Im interested in using the OpenEars library for my project and I want to ask you some questions

Does the software convert any speech I can make into voice? If I say “Pao”, would it output ‘Pao’? What about “BLAHBLAHBLAH”, would it output that? Or is it just limited to dictionary words?

If I say a number, lets say five, does it output ‘5’ or ‘five?

How long do I make a pause till I get the output?

[Halle Winkler]

Welcome,

OpenEars can recognize words from a vocabulary that you set. If you make a vocabulary with “Pao” it can recognize “Pao”, but you might have to make an entry in its lookup dictionary for how that should be pronounced. “5” versus “five” is the same — you add the one to your dictionary that you want returned. You can adjust the pause but it defaults to .7 seconds which I think is the best timeframe for most applications.

[pmpascua]

Is there a way to recognize anything? I had an app that makes people fill out forms that contains first name, last name, address, etc and a speech recognition is nice to have compared to typing a keyboard

[Halle Winkler]

Nope, it (and all speech recognition) always works from a predefined vocabulary list. If you can get text of their name, you can add the name to the list, but that might not help you since presumably you want them to make their very first name entry via speech, correct? So without knowing their name in advance, you can’t recognize their name.

[pmpascua]

:( so how do I add vocabularies? on the devices’ lookup dictionary?

[Halle Winkler]

Just run through the quickstart tutorial, it will make it all clear.

[pmpascua]

Hi, I followed the quick start tutorial just today and so far its working. One thing I don’t like is to predefining the vocabularies such that it accepts all possible words spoken. Is there a way to at least get the vocabulary lists on the device instead of typing all possible words?

For the numbers, I’ve added ‘1’ along with ‘WORD’ & ‘STATEMENT’ and when I said ‘1’, it returns ‘WORD’. why is that? Also, when ‘1’ and ‘ONE’ is added, ‘ONE’ always returns; Does it mean that its more likely to return words instead of numbers?

[Halle Winkler]

Is there a way to at least get the vocabulary lists on the device instead of typing all possible words?

I don’t quite understand this question — can you clarify what you are wanting to do here a bit more?

For the numbers, I’ve added ’1′ along with ‘WORD’ & ‘STATEMENT’ and when I said ’1′, it returns ‘WORD’. why is that? Also, when ’1′ and ‘ONE’ is added, ‘ONE’ always returns; Does it mean that its more likely to return words instead of numbers?

This is probably because “ONE” is finding a more accurate pronunciation match in the dictionary for your accent. If “1” is giving undesired results and “ONE” isn’t, go with “ONE”. It’s always necessary to do a bit of refinement on a vocabulary for best results.

[pmpascua]

I don’t quite understand this question — can you clarify what you are wanting to do here a bit more?

Nevermind, I’ll figure it out on my own.

I just found a critical problem. Let say I made a view controller to be presented and it contains a button to toggle whether to turn on or off OpenEars. When turned on, it calls the [pocketspinxController startListening] and when turned off, it calls the [pocketspinxController stopListening]. However when I turned it on again,either on same view or when the view is dismissed and presented again, it either crash or just froze on [pocket startListening]. Any Ideas why?

[Halle Winkler]

You can only have one listening loop running at once. If you have one running, you have to stop it before starting a new one.

[pmpascua]

So how do I stop it? [stopListening] method doesn’t work

[Halle Winkler]

That’s surprising. I just ran many tests this week which involved stopListening being used repeatedly in a session and it worked fine, and I haven’t gotten any reports of issues with stopListening. Is stopListening not working under one specific circumstance or is it never working for you? Are there more details you can give me about anything interesting about the case you’re describing?

[pmpascua]

Hmm how would I explain this…

So I had a button that toggles the voice recognition on the view controller. Whenever a view controller is loaded it allocates and initiate an instance of the pocketspinxcontroller. when I turn it on, I had message saying that pocketspinx is calibrated & listening and at that point I was able to start making speech to text stuff. Turning it on again would either freeze or got a message “Setting up the continuous recognition loop has failed for some reason, please turn on OpenEarsLogging to learn more.”

[Halle Winkler]

Are you using the lazily allocated memory management from the tutorial? It sounds a bit like your approach to instantiating the PocketsphinxController object is leading to multiple PocketsphinxControllers in the same session, both trying to use the mic stream. The sample app also has an example of how to do this safely.

[pmpascua]

i’ll just create a shared instance of the pocket sphinx controller in case that I created another view controller that might use voice recognition.

By the way, do you have something like an feature like one Siri had, where it detects the intensity of sounds listened and it looks like a wavy thing while you talk in iOS 7?

[Halle Winkler]

OK, I would strongly recommend having it as a property of one view controller and managing it as shown in the sample app and tutorial. It isn’t advised to put it in a singleton and can create difficult troubleshooting. With MVC it isn’t necessary to use a singleton for this kind of controller — just make the view controller responsible for PocketsphinxController that is the parent of the other view controllers in which you want to receive OpenEarsEventsObserver callbacks (which can be instantiated anywhere). It will stay instantiated in one place but provide results to as many places as you need, and you can control it either with delegate calls or notifications to the parent view controller in which the one PocketsphinxController is instantiated.

By the way, do you have something like an feature like one Siri had, where it detects the intensity of sounds listened and it looks like a wavy thing while you talk in iOS 7?

OpenEars doesn’t provide any visual interface elements, but it gives you hooks for reading the decibel levels of both incoming and outgoing speech so you can create UI elements like you described. Take a look at the docs to learn more.

[Author]

Posts