Note recognition in python? (real time)

Started by h12 February 6, 2009
I'm trying to get a signal being recorded from a microphone to be converted
to a list of musical notes played. 

I am new at this, and have gotten results from getting and audio signal
with pymedia or the sound card directly, then using Numpy's FFT I have
gotten what I think is a list of signal strengths at various frequencies.
It, however, does not seem to be verry accurate. 

I was wondering if this is the process to use, or I am on the wrong track
all together. 



h12 wrote:

> I'm trying to get a signal being recorded from a microphone to be converted > to a list of musical notes played. > > I am new at this, and have gotten results from getting and audio signal > with pymedia or the sound card directly, then using Numpy's FFT I have > gotten what I think is a list of signal strengths at various frequencies. > It, however, does not seem to be verry accurate. > > I was wondering if this is the process to use, or I am on the wrong track > all together.
Start with the basics. Hopefully, in 3..4 years you will be able to approach the problem. VLV
"h12" <turborage@gmail.com> writes:

> I'm trying to get a signal being recorded from a microphone to be converted > to a list of musical notes played. > > I am new at this, and have gotten results from getting and audio signal > with pymedia or the sound card directly, then using Numpy's FFT I have > gotten what I think is a list of signal strengths at various frequencies. > It, however, does not seem to be verry accurate. > > I was wondering if this is the process to use, or I am on the wrong track > all together.
There is already an application out there that does this: transcribe http://www.seventhstring.com/xscribe/overview.html -- % Randy Yates % "She has an IQ of 1001, she has a jumpsuit %% Fuquay-Varina, NC % on, and she's also a telephone." %%% 919-577-9882 % %%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO http://www.digitalsignallabs.com
>"h12" <turborage@gmail.com> writes: > >> I'm trying to get a signal being recorded from a microphone to be
converted
>> to a list of musical notes played. >> >> I am new at this, and have gotten results from getting and audio
signal
>> with pymedia or the sound card directly, then using Numpy's FFT I have >> gotten what I think is a list of signal strengths at various
frequencies.
>> It, however, does not seem to be verry accurate. >> >> I was wondering if this is the process to use, or I am on the wrong
track
>> all together. > >There is already an application out there that does this: transcribe > > http://www.seventhstring.com/xscribe/overview.html > >-- >% Randy Yates % "She has an IQ of 1001, she has a
jumpsuit
>%% Fuquay-Varina, NC % on, and she's also a
telephone."
>%%% 919-577-9882 % >%%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO
>http://www.digitalsignallabs.com >
Thanks, problem solved.
On Fri, 06 Feb 2009 13:07:29 -0500, Randy Yates <yates@ieee.org>
wrote:

>"h12" <turborage@gmail.com> writes: > >> I'm trying to get a signal being recorded from a microphone to be converted >> to a list of musical notes played. >> >> I am new at this, and have gotten results from getting and audio signal >> with pymedia or the sound card directly, then using Numpy's FFT I have >> gotten what I think is a list of signal strengths at various frequencies. >> It, however, does not seem to be verry accurate. >> >> I was wondering if this is the process to use, or I am on the wrong track >> all together. > >There is already an application out there that does this: transcribe > > http://www.seventhstring.com/xscribe/overview.html
MS SongSmith doesn't provide transcription, either, but the underlying processing could probably do it. At least it sure seems like it. If you haven't heard of SongSmith, it's an MS application whose function is to automatically put a band or orchestra behind a vocal track. It's pretty awful at it, but there are a few websites that have interesting collections of classic songs where the vocal/melody was isolated and run through SongSmith, like this one: http://i.gizmodo.com/5145015/the-strangest-microsoft-songsmith-remixes Roxanne is classic. I like "Creep" on here, too: http://www.npr.org/blogs/allsongs/2009/01/songsmith_brilliant_and_awful_1.html It seems like a transcription function would have to be under there somewhere. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php