abu_developer wrote:
Speech recognition in Scratch is difficult because:
1) basic sampling rates need to be about 2X the highest frequency ... human speech then would need to be sampled at a minimum of about 5000 times per second.
2) on my macbook (2GHz) machine, I can get a counter to count to 42 in 1 second.
3) if you were able to store the "amplitude" or volume of each of those samples in a list you would get roughly 5000 items in a list every second.
Not trying to be a nay-sayer, but I don't think Scratch is quite the tool you want to do this kind of programming!
CHEERS
I agree, Scratch is not fast enough
Offline
I beleive google did that, but I'm not sure it's possible with Squeak. It would take hours of work just to make it recognize one type of voice, so I think this would be pretty much impossible unless you use another programming language, which is out of the answer, sorry!
Offline
WindowsExplorer wrote:
I beleive google did that, but I'm not sure it's possible with Squeak. It would take hours of work just to make it recognize one type of voice, so I think this would be pretty much impossible unless you use another programming language, which is out of the answer, sorry!
Google did that but it still recognizes words for other words
Offline
LS97 wrote:
WindowsExplorer wrote:
I beleive google did that, but I'm not sure it's possible with Squeak. It would take hours of work just to make it recognize one type of voice, so I think this would be pretty much impossible unless you use another programming language, which is out of the answer, sorry!
Google did that but it still recognizes words for other words
Yea, and google only put that on an Adriod search app. Do you think the Nyan Cat Theme song is addictive?
Offline
oh, come on, let's close this: squeak is so slow, it would be only possible with plug-ins, making it hard to implement online. also, what about other languages? there are maybe two-three voice recognitions in the world that partly work, and all being developed by big companies like google (and there are probably some we don't know about, made for CIA etc...)
Offline
Writing a plug-in seems the only option to support speech recognition, but as pointed out, running code online from browsers gets more complicated. I am curious if plug-ins have been used to provide support to some feature other than speech recognition. In summary, I understand it is harder, but do you think it is possible to write / deploy plug-ins in Scratch?
PS: The accuracy of voice / speech recognition software has been improved considerably. Of course the state-of-art systems use sophisticated statistical algorithms. Constraining the grammar to few isolated words reduces the computational cost significantly.
Offline
One suggestion I have, is that we use the default windows 7 speech recognition feature.
I am using this right now to detect what I'm saying and writing it in this message. It is actually pretty good at understanding what I'm saying. I'm not sure how we will implement this, but I'm sure it could be done.
Offline
Thanks for the reply. Yes, invoking Microsoft's speech recognition engine is one option. A limitation is that users of a "speech-enabled" Scratch application would have to have the engine locally installed. Users could not simply execute it from Internet using a browser.
We have been working with speech-enabled applications at the university I am affiliated to. I will discuss with our group the suggestion of using Microsoft's engine and give some feedback via this forum.
Offline
I did a little experiment with this using <(loud)> and a list storing the values that would be compared with certain tolerances to a new sample. It actually worked reasonably well, so long as you said the word or words with the same tone every time. Test it out here: http://scratch.mit.edu/projects/Scratcher456/1993408 Oh, and yes, I am aware that this thread is kind of dead.
Offline
No.
Voice recognition is very hard to program. I don't even know if it's possible in Squeak. Plus, how would they implement it in the current block building interface? Plus it might be hard to use and have lots of user-defined technical inputs, and that can be confusing to younger Scratchers.
Offline