HQANAL
NAME
HQanal - High quality formant analysis
SYNOPSIS
HQanal
(-i item) (-o offset) (-t) (-c) (-r) (-6) file
DESCRIPTION
HQanal
performs excitation-synchronous formant analysis of a speech waveform using LPC and
root-solving following procedure specified by Melvyn Hunt.
Input should be a 10kHz speech waveform (preferably generated from a
20kHz acquired waveform using
down210)
and a set of excitation markers. The output is a set of raw formant estimates, synchronous
to the excitation markers in the voiced regions, and at 20ms intervals during unvoiced intervals.
The LPC covariance method is used, and pre-emphasis is applied to the voiced regions only.
Options
and their meanings are:
-I
Identify program and version number.
-i item
Select input item number.
-o offset
Set offset in micro-seconds between tx markers and start point of analysis window. Negative is
earlier, positive later, default is 0.
-t
Write tx item of analysis windows used in analysis.
-c
Write LPC coefficients used in analysis
instead of formant estimates.
-r
Write LPC analysis residual as lx waveform.
-6
Use 6-pole LPC analysis in voiceless regions.
INPUT ITEMS
1.xx
10kHz sampled speech waveform.
3.xx
Excitation markers.
OUTPUT ITEMS
12
High Quality formant estimates.
14
Optional LPC coefficients.
2
Optional residual waveform.
VERSION/AUTHOR
1.2 - Mark Huckvale (from
lpc
and
rootlpc
by Lynn Whitaker)
SEE ALSO
"Generation of controlled speech stimuli by pitch-synchronous LPC analysis of
natural utterances", M.J. Hunt & C.E. Harvenberg, 12th Intl. Congress on Acoustics,
Toronto Canada, 1986.
BUGS
Good formant analysis relies on good data preparation as much as on
a good program.
The formant item produced by
HQanal
is not compatible with
fmconv,
use
HQsynth
for re-synthesis.
Wed Jul 17 22:31:36 2013