lp_syn.1

speech signal process tools

.\" Copyright (c) 1996 Entropic Research Laboratory, Inc.; All rights reserved
.\" @(#)lp_syn.1	1.5 1/26/96 ERL
.ds ]W (c) 1996 Entropic Research Laboratory, Inc.
.TH  LP_SYN 1\-ESPS 1/26/96
.SH NAME

.nf
lp_syn \-  speech LPC synthesis from an excitation signal or a parametric source
.fi
.SH SYNOPSIS
.B
lp_syn
[
.BI \-P " param_file"
] [
.BI \-{pr} " range"
] [
.BI \-s " range"
] [
.BI \-i " int_const"
] [
.BI \-t " synth_rate"
] [
.BI \-f " f0_ratio"
] [
.BI \-u " up_ratio"
] [
.BI \-d " damp"
] [
.BI \-z
] [
.BI \-x " debug_level"
]
.BI lpfile
.BI exfile
.BI outfile
.SH DESCRIPTION
\fILp_syn\fR uses the LP (linear prediction) or reflection 
coefficients stored in the file
\fIlpfile\fR to synthesize a speech signal from an excitation source or a 
parametric source in the file \fIexfile\fR.  \fILpfile\fR is a FEA
file with the field \fIspec_param\fR to store the spectral coefficients
and with the generic header item \fIspec_rep\fR specifying the type of spectral
representation -- \fI"RC"\fR for reflection coefficients or \fI"AFC"\fR for
LP coefficients.  \fIExfile\fR may be a FEA_SD file of SHORT data type as
an excitation source, or a FEA file as a parametric source.  If \fIexfile\fR
is a FEA file, it must contain three fields: \fIrms\fR for the frame RMS value,
\fIF0\fR for the frame fundamental frequency, and \fIprob_voice\fR for the
frame voicing state.
.PP
The synthesizer uses a Rosenberg-polynomial glottal-flow pulse, open-phase
damping, and per-sample gain correction.
.PP
.SH OPTIONS
.PP
The following options are supported:
.TP
.BI \-P " param_file \fR[params]\fP"
uses the parameter file 
.I param_file
rather than the default, which is \fIparams\fP. 
.TP
.BI \-r " first:last"
.TP
.BI \-r " first:+incr"
Determines the range of frames from input file, \fIlpfile\fR.  In
the first form, a pair of unsigned integers gives the first and last frame
of the range.  If \fIfirst\fR is omitted, 1 is used.  If \fIlast\fR is
omitted, the last frame in the file is used.  The second form is equivalent
to the first with \fIlast = first + incr\fR.
.TP
.BI \-p " "
Same as the \fB-r\fR option.
.TP
.BI \-s " first:last"
.TP
.BI \-s " first:+incr"
Same function as the \fB-r\fR option, but specifies the range of input frames
in seconds.
.TP
.BI \-i " int_const \fR[0.95]\fP"
specifies the coefficient of a first-order integrator to be applied to
the output if parametric source is used.  This constant is applied in
the fricative regions only.  When synthesizing from residual
excitation \fIint_const\fR would usually be set to 0.0.
.TP
.BI \-t " synth_rate \fR[1.0]\fP"
specifies the time-scale modification rate.  This speeds up or slows
down the rate of speech without distorting the pitch.  Use values less
than 1.0 to slow the speech, use values more than 1.0 to speech it up.
This can only be used during parametric source synthesis.
.TP
.BI \-f " f0_ratio\fR[1.0]\fP"
specifies the output fundamental frequency scale factor if and only if
parametric source is used.  The output fundamental frequency is
multiplied by this factor.
.TP
.BI \-u " up_ratio \fR[1]\fP"
specifies the upsampling ratio to use during synthesis from a parametric
source.  This produces an output file with sampling rate of
\fIup_ratio\fR times the value of \fIsrc_sf\fR generic header item in
the input \fIlpfile\fR.  If \fIsrc_sf\fR does not exist in
\fIlpfile\fR, a value of 2 times the value of \fIbandwidth\fR generic
header in \fIlpfile\fR is assumed.  Valid values are integers greater or
equal to one.
.TP
.BI \-d " damp \fR[0.1]\fP"
specifies a parameter that controls the amount of open-glottal-phase
damping that is applied to the lattice filter state vector when
parametric source synthesis is performed.  When LP coefficients are
computed using standard frame-synchronous analysis (e.g. hanning
window; 25ms duration), it is usually best to set \fIdamp\fR to 0.0.
When parameters are computed using \fIps_ana\fR, values greater than
zero often improve the naturalness of the synthetic speech.
.TP
.BI \-z
inverts the sign of input spectral parameters before using them for
synthesis.
.TP
.BI \-x " debug_level \fR[0]\fP"
If 
.I debug_level
is positive,
.I lp_syn
prints debugging messages and other information on the standard error
output.  The messages proliferate as the  
.I debug_level
increases.  If \fIdebug_level\fP is 0 (the default), no messages are
printed.  
.SH ESPS PARAMETERS
The following ESPS parameters have the same meanings as the command-line
options supported
.TP
.I "start - integer"
.IP
The first frame in the input file \fIlpfile\fR that is processed.  A
value of 1 denotes the first frame in the file.  This is only read
if the \fB\-p\fP option is not used.  If it is not in the parameter
file, the default value of 1 is used.  
.TP
.I "nan - integer"
.IP
The total number of frames in the file \fIlpfile\fR to process.  If 
.I nan
is 0, the whole file is processed.  
.I Nan
is read only if the \fB\-p\fP option is not used.  (See the 
discussion under \fB\-p\fP).
.TP
.I int_const - float
.PP
.I synth_rate - float
.PP
.I f0_ratio - float
.PP
.I up_ratio - int
.PP
.I damp - float
.PP
.SH ESPS COMMON
No ESPS common parameter processing is supported
.PP
.SH ESPS HEADERS
The usual \fIrecord_freq\fR, \fIstart_time\fR header items, and all
supported parameters are stored as generic header items.
.PP
.SH FUTURE CHANGES
.PP
.SH EXAMPLES
The example shell script included below shows how one might use some
of the analysis and synthesis programs supplied with ESPS.  The script
takes one argument, the basename of a speech file with the extension
".sd".  Read the comments in the script for details.
.PP
.nf
.na
.ne 10
#!/bin/sh
#
# Determine the sample rate of the original speech file.
sf=`hditem -i record_freq $1.sd`
#
# Establish the window size and frame step for periodic analyses.
size=.02
step=.005
#
# Get analysis step size and window length in samples.
ssize=`echo $sf $size \\* p q | dc`
sstep=`echo $sf $step \\* p q | dc`
#
# Standard rule-of-thumb computation for LPC order.
order=`echo $sf 1000 / 2 + p q | dc`
#
# Compute reflection coefficients using a standard set of parameters
refcof -z -r1:1000000 -e.97 -x0 -wHANNING -l$ssize -S$sstep \\
       -o$order $1.sd $1.rc
#
# Get a high-resolution estimate of F0 and a reasonably accurate
#  voicing-state estimate.
get_f0 -i $step $1.sd $1.f0
#
# Compute the LPC residual (approximates the glottal flow derivative).
get_resid -a 1 -i 0.0 $1.sd $1.rc $1.res
#
# Blank out the residual signal in the unvoiced regions.
mask $1.f0 $1.res $1.resm
#
# Find the points of glottal closure in the voiced regions.
epochs  -o $1.lab $1.resm $1.pe
#
# Perform epoch-synchronous LPC analysis.
ps_ana -f $1.f02 -i $step -e .97 $1.sd $1.pe $1.rc
#
# Perform synthesis using these parameters.
lp_syn $1.rc $1.f02 $1.syn
#
# Alternatively, the better F0 and voicing-state estimates from get_f0
# can be used by merging them with the preemphasized RMS feature from
# ps_ana:
rm -f $1.f03
mergefea -fF0 -fprob_voice  $1.f0 $1.f03
mergefea -a -f rms $1.f02 $1.f03
#
# Perform synthesis using these new parameters.
lp_syn $1.rc $1.f03 $1.syn2
.fi
.ad
.PP
.SH ERRORS AND DIAGNOSTICS
.PP
.SH BUGS
Synthesis using LPC (AFC) parameters does not work in the parametric
excitation mode.  This will be fixed in the next release.  Synthesis
using reflection coefficients (RC), does work in both modes.
.SH REFERENCES
Talkin, D. and Rowley, J., "Pitch-Synchronous analysis and synthesis
for TTS systems," \fIProceedings of the ESCA Workshop on Speech
Synthesis\fP, C. Benoit, Ed., Imprimerie des Ecureuils, Gieres, France,
1990.
.PP
.SH "SEE ALSO"
refcof (1-ESPS), get_resid (1-ESPS), mask (1-ESPS), get_f0 (1-ESPS),
ps_ana (1-ESPS), transpec (1-ESPS)
.PP
.SH AUTHORS
David Talkin, Derek Lin
.PP