🚀 go-pugleaf

Dario Antonio Palhares Carvalho de Freitas Diogo <falphaxt@cc.fc.ul.pt> wrote:
>
> 
> What is (phase?) vocoder? I gather it's an advanced audio DSP algorithm, 
> but what's it about axactly?
> 
> I've already serched the internet for information but couldn't find 
> anything specific.
> 
> If you can help me, please e-mail your answer to the address below.
> 
> 
> Thanx in advance...
> 
> > -----------------------------------------------------------------------------
>                   falphaxt@cc.fc.ul.pt ----- Dario Diogo
> -----------------------------------------------------------------------------
> 
> 
> 
VOCODERS have been used for secure transmission of voice for almost
50 years.  The original VOCODERS would analyse voice spectrum and 
the transmit the amplitude of each frequency band, and the amplitude
and frequency (pitch) of the voice or the amplitude of noise for
unvoiced sounds.

Modern VOCODERS use 10th order Linear Predictive Coding to analyse
voice.  Sampled voice (8 KHz sample rate) is analysed 50 times a
second to determine the shape of the vocal tract.  The ten "areas",
and the amplitude and frequency (pitch) of the voice or the amplitude
of noise for unvoiced sounds is sent every fiftieth of a second.

See Federal Standard 1015:  Analog to Digital converson of Voice by
2400 bit/second Linear Predictive Coding.  Published by General Services
Admisintration, Office of Information Resources Management.
(Date on this standard is 28 Nov 1984.)

  --  Russ

In article <4bg21b$1qp@news1.deltanet.com>,
J. Russell Lemon <lemonjr@cyber.net> wrote:
]Dario Antonio Palhares Carvalho de Freitas Diogo <falphaxt@cc.fc.ul.pt> wrote:
]>
]> 
]> What is (phase?) vocoder? I gather it's an advanced audio DSP algorithm, 
]> but what's it about axactly?
]> 
]> 
]VOCODERS have been used for secure transmission of voice for almost
]50 years.  The original VOCODERS would analyse voice spectrum and 
]the transmit the amplitude of each frequency band, and the amplitude
]and frequency (pitch) of the voice or the amplitude of noise for
]unvoiced sounds.
]
]Modern VOCODERS use 10th order Linear Predictive Coding to analyse
]voice.  Sampled voice (8 KHz sample rate) is analysed 50 times a
]second to determine the shape of the vocal tract.  The ten "areas",
]and the amplitude and frequency (pitch) of the voice or the amplitude
]of noise for unvoiced sounds is sent every fiftieth of a second.
]
]See Federal Standard 1015:  Analog to Digital converson of Voice by
]2400 bit/second Linear Predictive Coding.  Published by General Services
]Admisintration, Office of Information Resources Management.
](Date on this standard is 28 Nov 1984.)
]

Actually, this is not correct.  A *Phase* Vocoder does an inverse fft overlap
and add resynthesis of a sound from a set of fft data frames.  It is called a
phase vocoder because of the way it generates time-varying frequency bands via
an "unwinding" of the phase component for each frequency sub-band.  There are
several documents around that deal specifically with the working of the phase
vocoder, and there is one built into both the Cmusic (MIT) and Cmix
(Princeton) synthesis packages.


An LPC resynthesis process is usually called something other than VOCODER, in
modern computer music circles, but I cannot argue with the use of the general
term -- only the specific term "Phase Vocoder".

-- 
---
Douglas Scott                       | Silicon Graphics, Inc.
dscott@esd.sgi.com                  | Digital Media Systems 
(415) 933-6404  Fax: (415) 390-6159 | http://reality.sgi.com/employees/dscott