Hello,

I'm trying to figure out a way to transform audio files from the  
standard amplitude over time format to some kind of frequency over  
time format, preferably with frequency values spaced along a  
logarithmic scale.  This is for a machine learning experiment.

I've been trying to figure out if I can use csound to perform the  
transformation.  It looks like the spectrum opcode may do what I'm  
looking for, but I'm having a hard time figuring out what its output  
signal looks like.  Is the wsig data type documented anywhere, or is  
there a better way to do what I'm trying to do?

If I haven't explained what I want well enough let me know and I'll  
try to clarify.

Thanks for any help.

-Greg

Hi,

Have you looked at the pvs and pv opcodes?
Peiman


Gregory Cooksey wrote:
> 
> Hello,
> 
> I'm trying to figure out a way to transform audio files from the  
> standard amplitude over time format to some kind of frequency over  
> time format, preferably with frequency values spaced along a  
> logarithmic scale.  This is for a machine learning experiment.
> 
> I've been trying to figure out if I can use csound to perform the  
> transformation.  It looks like the spectrum opcode may do what I'm  
> looking for, but I'm having a hard time figuring out what its output  
> signal looks like.  Is the wsig data type documented anywhere, or is  
> there a better way to do what I'm trying to do?
> 
> If I haven't explained what I want well enough let me know and I'll  
> try to clarify.
> 
> Thanks for any help.
> 
> -Greg
> 
> -- 
> Send bugs reports to this list.
> To unsubscribe, send email to csound-unsubscribe@lists.bath.ac.uk
> 
> 

-- 
View this message in context: http://www.nabble.com/exponentially-spaced-spectral-transformation--tf4594662.html#a13119044
Sent from the Csound - General mailing list archive at Nabble.com.

The pvs opcodes have a lot more options for getting data out of them than the 
spectrum.  To my knowledge there isn't yet anything that will write a wsig to a 
table, for instance.  However, they use linear frequency.  If you can use a linear 
frequency scale they would be ideal.

Have you considered using something like Matlab or Octave?  A wavelet transform 
within one of those programs seems to be exactly what you're looking for.  The 
spectrum opcode creates something very close to a wavelet transform, but I don't know 
of an easy way to get data out of a wsig short of writing a new opcode.


John W. Lato
School of Music
The University of Texas at Austin
1 University Station E3100
Austin, TX 78712-0435
(512) 232-2090

peiman wrote:
> Hi,
> 
> Have you looked at the pvs and pv opcodes?
> Peiman
> 
> 
> Gregory Cooksey wrote:
>> Hello,
>>
>> I'm trying to figure out a way to transform audio files from the  
>> standard amplitude over time format to some kind of frequency over  
>> time format, preferably with frequency values spaced along a  
>> logarithmic scale.  This is for a machine learning experiment.
>>
>> I've been trying to figure out if I can use csound to perform the  
>> transformation.  It looks like the spectrum opcode may do what I'm  
>> looking for, but I'm having a hard time figuring out what its output  
>> signal looks like.  Is the wsig data type documented anywhere, or is  
>> there a better way to do what I'm trying to do?
>>
>> If I haven't explained what I want well enough let me know and I'll  
>> try to clarify.
>>
>> Thanks for any help.
>>
>> -Greg
>>
>> -- 
>> Send bugs reports to this list.
>> To unsubscribe, send email to csound-unsubscribe@lists.bath.ac.uk
>>
>>
> 

How so Chuck?

I know we have things like SPEAR et al (our original poster may also find
SPEAR of use...)

but by a similar implication, converting PVOC analysis data to a directly
viewable & editable format is something i have thought might be handy /
useful at some point... (although thinking out loud, i'm not sure what it
would offer that SPEAR & /or SDIF style resynthesis does not exactly.. in
terms of 

*ease of use in performing meaningful adjustments on data before
resynthesis, or 
* arbitrarily creating "spectra" for synthesis based on non analysis / real
world models..... 

I have some related thoughts & issues about ATS i was going to ask , but
i'll put them in a separate post I think immediately to follow this one...

As far as the original post goes, I also thought at some point i heard
something about the possibility of a more exponentially spaced analysis that
was achievable by focussing different window sizes on different frequency
bands... idle speculation though on my part, but someone might actually
enlighten us to what that is or might be... (?)


Chuckk Hubbard wrote:
> 
> I don't know if this fits your needs, but it would be fairly simple to
> produce a logarithmic *graph* of the linear frequency analysis produced by
> the pvs opcodes.  But the data itself would still be spaced linearly.  If
> that doesn't help, I don't know of anything.
> 
> -Chuckk
> 
> 
> 
> 

-- 
View this message in context: http://www.nabble.com/exponentially-spaced-spectral-transformation--tf4594662.html#a13152002
Sent from the Csound - General mailing list archive at Nabble.com.