FPGA Central

Hi,

Given a collection of audio files (raw streams) I would like to
determine which of the songs has an acoustic guitar in it.
Coming from speech and language processing I am aware of some
strategies for feature extraction, but not enough to build something
from scratch.

Can anyone point me to simple resources on how to accomplish something
like that? Optimally something example-driven, that is.

-Samuel

>Given a collection of audio files (raw streams) I would like to
>determine which of the songs has an acoustic guitar in it.

This is an extremely difficult if not impossible task.
Perhaps build a neural network and "teach" it what acoustic guitar is?!

DSP team wrote:
>>Given a collection of audio files (raw streams) I would like to
>>determine which of the songs has an acoustic guitar in it.
>
>
> This is an extremely difficult if not impossible task.
> Perhaps build a neural network and "teach" it what acoustic guitar is?!
>
>

How big a universe do you really have?
Are you trying to distinguish between a vocalist
with/without accompaniment?
with guitar versus vocalist with wind instrument?
with guitar versus bowed string instrument?
with guitar versus percussive string instrument?

Never try to solve the general problem.
Solve a *SPECIFIC* problem first.