[Resolved] about get the recognize result

Home Forums OpenEars [Resolved] about get the recognize result

Viewing 12 posts - 1 through 12 (of 12 total)

  • Author
    Posts
  • #1024104
    sunnysuhappy
    Participant

    dear Halle
    i find that if i want to get the recognize result, i has two ways.one is that using real time ‘startListeningWithLanguageModelAtPath’, and when it get a recognized word, it callsbacks.another is using ‘pathToTestFile’ it will give a callback.
    My questiin is that, if i have a way to stop current realtime input and get the result that has been input.

    #1024112
    Halle Winkler
    Politepix

    Welcome,

    It isn’t designed to stop at a specific time and receive a hypothesis of everything up to that time.

    #1024128
    sunnysuhappy
    Participant

    ok
    thanks for your reply,and i want to know if I can input audio buffer to recognize instead of a wav path.

    #1024129
    Halle Winkler
    Politepix

    That is what is done by default – not sure I understand the question yet, can you clarify?

    #1024152
    sunnysuhappy
    Participant

    soory, I mean if I have a way that i can input custom audio buffer in it.instead of the api startListen.
    Now i think i find the way to achieve my target.use rapidEars to have realtime recognize.but i find that rapidEars has crash when use it for server seconds, and it seems having memory leak, it always lead to a large memory useage increased every time to 100M and more

    #1024153
    sunnysuhappy
    Participant

    This situation happens when i use ReturnNBest mode.

    #1024154
    Halle Winkler
    Politepix

    OK, please read the note about using N-Best with RapidEars in its documentation – it’s explained that it’s an experimental feature and the developer has the job of keeping the overhead low enough for it to function. It is more likely that it has too much to do than that there is a leak.

    #1024155
    sunnysuhappy
    Participant

    Ok thanks.
    If i want to recognize lots of sentences. so i think I should clean lauageModel when recognize new words to save memory. but it seems no apis.

    #1024156
    Halle Winkler
    Politepix

    It is very easy to use multiple language models and switch between them with OpenEars. Please take a look at the OpenEars sample app, which has an example of dynamically generating a second language model and switching between models. It would probably also be useful for you to use the tutorial (https://www.politepix.com/openears/tutorial) and read the documentation for OpenEars to get an easy overview of the API.

    #1024162
    Halle Winkler
    Politepix

    There’s a new version of RapidEars out today, so please check it out if you are having issues with other RapidEars methods.

    #1024341
    sunnysuhappy
    Participant

    ok thanks very much

    #1024479
    Halle Winkler
    Politepix

    Hello,

    Just a quick note that I sent you an important email to the address you registered this forum account with – if you haven’t seen my email, could you check your spam folder or commercial folder to see if it ended up in there? Thanks!

Viewing 12 posts - 1 through 12 (of 12 total)
  • You must be logged in to reply to this topic.