last day (14 days later) » 

11:26 PM
hello... what's the challenge?
 
@TAbraham the challenge is coming up with a challenge
 
@MartinBüttner "challenge is coming up with a challenge" what does that mean?
 
hey, I don't know if this is too easy of the challenge but what about identifying a person saying a phrase in an audio file..
 
identifying the person, or the phrase?
 
11:36 PM
the phrase..
 
@TAbraham "too easy"? :D
 
google can't do this
 
or the person, whichever one is harder..
 
definitely the phrase
I think that's beyond "appropriate PPCG difficulty"
 
yes, millions of dollars are spent on these programs and they suck anyway
 
11:39 PM
how about a particular word...
 
18
Q: Voice recognition: "Yes" or "No"?

Vi.Task Implement a program in minimum bytes of source or binary code that does voice recognition of a voice sample (me saying "yes", "yeah" or "no" in voice or in whisper, plainly or quirkily) based on training samples with maximum accuracy. The program should read train/yes0.wav, train/no0.wav, ...

 
or what about a really simple siri...
 
I think something that also produces audio would be nice.
E.g. steganography
 
@MartinBüttner I like that idea..
 
11:41 PM
 
yes.. great idea..
 
or self-made audio compression
(although people would probably just reimplement the easy parts of MP3)
anyway, I think in general we can find a lot of inspiration in the tag (without copying something exactly over to audio)
 
hey.. what about a person says one of the names of a hat and the program show a picture of it..
 
I don't know... I'm not too fond of voice recognition in general... it's been done for the simple case, and I don't think the challenge really gets more interesting by making it recognise a wider variety of words (you just pick a better classifier or train it for longer). I don't think it really adds anything.
I'd prefer if we used the weekly challenges to explore some genuinely new territory
 
ok..
 
11:49 PM
(that being said, I'm not the sole authority on this... so if everyone else thinks voice recognition is the way to go, then we'll definitely do that!)
 
how about input is some musical notes being played, and it outputs a score
1 note at a time so it is not too hard
 
that seems a bit more interesting. it's regression and not classification, and it also involves measuring the timing to figure out the note lengths.
well I'm off for tonight... keep brainstorming ;)
(we could also come up with an audio-pendant to Piet and ask for an interpreter)
 
gota go too... great thoughts.. I hope to see what great ideas we come up with for this challenge..
 

  last day (14 days later) »