Hi!
Next week we begin a development course for iphone but in two days, I have to submit a budget in hours about how long it would take to develop a small App
The App will serve for the Deaf, which work as postmen, which at the touch of a ring (interphone) in someone Portal,needs to identify someone has answer (there has been feedback). Do not need to acknowledge what he says, but someone has spoken. If so, the phone app will launch a mp3 saying "postal letter carrier, open me please"
this has to do without Internet access.In Android, it is necessary to stream the audio go to Google's servers for proccess.We need, in the iPhone, to do so without internet access.
well ... I have have no idea yet aboud iPhone development, How do you see this app? would yoy help me, give me some clue?
I have to write the budget and if the clients accepts, Iīll take the course.
The App will serve for the Deaf, which work as postmen, which at the touch of a ring (interphone) in someone Portal,needs to identify someone has answer (there has been feedback). Do not need to acknowledge what he says, but someone has spoken. If so, the phone app will launch a mp3 saying "postal letter carrier, open me please"
this has to do without Internet access.In Android, it is necessary to stream the audio go to Google's servers for proccess.We need, in the iPhone, to do so without internet access.
I imagine that you could take advantage of the fact that human voices tend to be all over the place (frequency space-wise) and that ringing sounds tend to be constrained to rather narrow bands.
Pull a Fourier transform on the sounds you hear (note: there's libraries that'll do this work for you - check Accelerate.framework). White noise (aka static) will be broad spectrum, which is to say at every frequency, but amplitudes should be low. Ring noises should be spikes on very discrete frequencies at regular intervals. Human voice will feature irregular spikes at irregular intervals.
(Note to the haters: Yes. I know. The X axis is frequency, the Y axis is amplitude, and there's no "time" axis. Hush. FT portions of your sampled audio - like 3 to 7 seconds at a time. Long enough to pick up a ring, not too long to make the human hang up.)
Now you'll be telling a lot of voice machines about postal letter carriers, so be sure to keep the mike going while playing your MP3. If it picks up a BEEP you can flash a message to your carrier like "Sounds like an answering machine".
Voice recognition (or sound)
Linear Mode
