man.diff

来自「speech signal process tools」· DIFF 代码 · 共 548 行 · 第 1/2 页

message is printed.  The reflection coefficients are stored in the
field "refcof", unless \fIrc_fname\fP is assigned a different name.
The field has data type REAL and dimension \fIac_order\fP.
.TP 
.I "lpc_flag \- integer"
.TP
.I "lpc_fname \- string"
.IP
If \fIlpc_flag\fP is set to 1, linear prediction coefficients
are found from the sample autocorrelation of the windowed data frame
using \fIget_atal\fP(3-ESPS). 
The parameter \fIac_order\fP determines the number of prediction 
coefficients which are computed. If \fIac_order\fP=0, a warning
message is printed.  The coefficients are stored in the
field "lpc_coeffs", unless \fIlpc_fname\fP is assigned a different name.
The field has data type REAL and dimension \fIac_order\fP.
.TP 
.I "lpccep_flag \- integer"
.TP
.I "lpccep_fname \- string"
.TP
.I "lpccep_order \- integer"
.TP
.I "lpccep_deriv \- string"
.IP
If \fIlpccep_flag\fP is set to 1, cepstral coefficients
are computed from reflection coefficients found from the windowed
data frame. 
The program \fIget_atal\fP(3-ESPS) is used to compute \fIac_order\fP 
reflection coefficients; 
\fIlpccep_order\fP cepstral coefficients are obtained using 
\fIrc_reps\fP(3-ESPS).
If \fIac_order\fP is not defined or is 0, it is set to \fIlpc_order\fP;
if \fIlpccep_order\fP is zero, it is set to \fIac_order\fP.
If \fIac_order\fP < \fIlpccep_order\fP, 
the reflection coefficients are padded with
\fIlpccep_order\fP\-\fIac_order\fP zeros 
before the cepstral coefficients are computed.
If \fIlpccep_order\fP < \fIac_order\fP, \fIacf\fP sets 
\fIlpccep_order\fP to \fIac_order\fP.

The cepstral coefficients are stored in the
field "lpc_cepstrum", unless \fIlpccep_fname\fP is assigned a
different name. The field has data type REAL and, if 
\fIlpccep_deriv\fP 
is not specified, has dimension \fIlpccep_order\fP. If the string 
\fIlpccep_deriv\fP is defined, the routine
\fIgrange_switch\fP(3-ESPS) is used to to parse the string to determine
which elements of the \fIlpccep_order\fP cepstral coefficients are
to form the output field. For example if the entry 
.br
\fBstring lpccep_deriv = "0,11:20";\fP
.br
appears in the parameter file,
the output field has 21 elements taken from positions 0, and 11 through
20 in the cepstral sequence.  If \fIlpccep_deriv\fP specifies 
points outside the range [0:\fIlpccep_order\fP-1], \fIacf\fP 
prints a warning and exits.
.TP
.I "lar_flag \- integer"
.TP
.I "lar_fname \- string"
.IP
If \fIlar_flag\fP is set to 1, log area ratios (LARs)
are computed from the reflection coefficients found from the windowed
data frame. The reflection coefficients are found as above, and the 
LARs are obtained using \fIrc_reps\fP(3-ESPS).
The parameter \fIac_order\fP determines the number of 
LARs which are computed. If \fIac_order\fP=0, a
warning message is printed.  The LARs are stored in the
field "log_area_ratio", unless \fIlar_fname\fP is assigned a
different name. The field has data type REAL and dimension
\fIac_order\fP. 
.TP
.I "lsf_flag \- integer"
.TP
.I "lsf_freq_res \- float"
.TP
.I "lsf_fname \- string"
.IP
If \fIlsf_flag\fP is set to 1, line spectral frequencies
(LSFs) 
are computed from the reflection coefficients found from the windowed
data frame. The reflection coefficients are found as above, and the 
LSFs are obtained using \fIrc_reps\fP(3-ESPS).
The parameter \fIac_order\fP determines the number of 
LSFs which are computed. If \fIac_order\fP=0, a
warning message is printed.  The LSFs are stored in the
field "line_spec_freq", unless \fIlsf_fname\fP is assigned a
different name. The field has data type REAL and dimension
\fIac_order\fP. The parameter \fIlsf_freq_res\fP has default value
10.0 (see \fIrc_reps\fP(3-ESPS)).
.TP
.I "fftcep_flag \- integer"
.TP
.I "fftcep_fname \- string"
.TP
.I "fftcep_order \- integer"
.TP
.I "fftcep_deriv \- string"
.IP
If \fIfftcep_flag\fP is set to 1, the FFT cepstrum is 
computed from the windowed data frame using the routine
\fIfft_cepstrum_r\fP(3-ESPS). 
The cepstral data is stored in the
field "fft_cepstrum", unless \fIfftcep_fname\fP is assigned a
different name.
The parameter \fIfftcep_order\fP determines the order of the FFT, i.e.
the FFT produces 2^\fIfftcep_order\fP frequencies.
If \fIfftcep_order\fP=0, a
warning message is printed.  The field has data type REAL and, if 
\fIfftcep_deriv\fP 
is not specified, has dimension 2^\fIfftcep_order\fP. If the string 
\fIfftcep_deriv\fP is defined, the routine
\fIgrange_switch\fP(3-ESPS) is used to to parse the string to determine
which elements of the 2^\fIfftcep_order\fP cepstral coefficients are
to form the output field. For example if the entry 
.br
\fBstring fftcep_deriv = "0,11:20";\fP 
.br
appears in the parameter file,
the output field has 21 elements taken from positions 0, and 11 through
20 in the cepstral sequence.  If \fIfftcep_deriv\fP specifies 
points outside the range (0:2^(\fIfftcep_order\fP)-1), \fIacf\fP 
prints a warning and exits.
.TP
.I "fft_flag \- integer"
.TP
.I "fft_order \- integer"
.IP
If the flag \fIfft_flag\fP is set to 1, the FFT of the windowed data
frame is stored in \fIfea_file\fP.  The order of the FFT is
\fIfft_order\fP, i.e. 2^\fIfft_order\fP negative and positive
frequencies are computed.  If \fIfft_flag\fP is set, \fIfea_file\fP
has FEA subtype FEA_SPEC, and the data format is SPFMT_SYM_EDGE, 
and is stored in decibels (log power) (see FEA_SPEC(5-ESPS)).
If \fIfft_order\fP = 0, a warning message is printed; the default
values of \fIfft_order\fP is 10.
.TP
.I "warp_param \- float"
.IP
If \fIwarp_param\fP is not 0.0, the bilinear transform routine
\fIblt\fP(3-ESPS) is applied to the 
lpc cepstrum and FFT cepstrum
using the parameter
\fIwarp_param\fP. 
Warping is performed before elements are extracted from the cepstral
sequences (see parameters \fIlpccep_deriv\fP and \fIfftcep_deriv\fP).
If the parameter is outside the range (-1,1),
\fIacf\fP warns and exits.
.PP
The values of parameters obtained from the parameter file are printed
if the environment variable ESPS_VERBOSE is 3 or greater.  The default
value is 0.
.sp
.SH ESPS COMMON
.PP
ESPS Common is read provided that Common processing is enabled and 
that the 
.I filename
entry in Common matches 
.I sd_file,
in which case parameters present in Common override values from
the parameter file, which in turn may be overriden by command
line options (see the discussion in ESPS PARAMETERS and under
each option).   
Common is not read if 
.I input
is standard input.  
If 
.I output
is not standard output and 
.I input
is not standard input, the Common parameters 
\fIfilename\fP (= sd_file), \fIprog\fP (= acf), 
.I start,
and
.I nan
are written to ESPS Common.
.PP
ESPS Common processing may be disabled by setting the environment variable
USE_ESPS_COMMON to "off".  The default ESPS Common file is .espscom 
in the user's home directory.  This may be overidden by setting
the environment variable ESPSCOM to the desired path.  User feedback
of Common processing is determined by the environment variable
ESPS_VERBOSE, with 0 causing no feedback and increasing levels causing
increasingly detailed feedback.  If ESPS_VERBOSE is not defined, a
default value of 0 is assumed.
.sp
.SH ESPS HEADER
.PP
A new file header is created for the FEA file \fIfea_file\fP.  If the FFT is
computed, the header is of FEA subtype FEA_SPEC.  
The sampled data header
from the input header is added as a source in the output file header, and
the command line is added as a comment. The input file \fIinput\fP is
set as the reference header for tags.
.PP
The program writes the usual values into the common part of the
output header.  \fIAcf\fP creates and writes the following generic
items in the output file:
.nf
.sp
start = \fIstart\fP (LONG)
nan = \fInan\fP (LONG)
frmlen = \fIframe_len\fP (LONG)
src_sf = sample frequency of \fIinput\fP (FLOAT)
step = \fIstep\fP (LONG)
window_type = \fIwindow_type\fP (CODED) (not written for RECT window)
warp_param = \fIwarp_param\fP (FLOAT)
preemphasis = \fIpreemphasis\fP (FLOAT)
start_time (DOUBLE)
record_freq (DOUBLE)
.fi
.PP
The value written for \fIstart_time\fP is computed by taking the
\fIstart_time\fP value from the header of the input file (or zero, if
such a header item doesn't exist) and adding to it the relative time
from the first record in the file to the first record processed.  The
generic header item \fIrecord_freq\fP is the number of output records
per second of input.
.PP
The following generic header items are created based 
on entries in the parameter file:
.nf
.sp
source_field_name = \fIsd_field_name\fP (STRING)
sd_flag = \fIsd_flag\fP (SHORT)
sd_fname = \fIsd_fname\fP (STRING)
pwr_flag = \fIpwr_flag\fP (SHORT)
pwr_fname = \fIpwr_fname\fP (STRING)
zc_flag = \fIzc_flag\fP (SHORT)
zc_fname = \fIzc_fname\fP (STRING)
ac_flag = \fIac_flag\fP (SHORT)
ac_fname = \fIac_fname\fP (STRING)
ac_order = \fIac_order\fP (STRING)
rc_flag = \fIrc_flag\fP (SHORT)
rc_fname = \fIrc_fname\fP (STRING)
lpc_flag = \fIlpc_flag\fP (SHORT)
lpf_fname = \fIlp_fname\fP (STRING)
lpccep_flag = \fIlpccep_flag\fP (SHORT)
lpccep_fname = \fIlpccep_fname\fP (STRING)
lpccep_order = \fIlpccep_order\fP (STRING)
lpccep_deriv = \fIlpccep_deriv\fP (STRING)
lar_flag = \fIlar_flag\fP (SHORT)
lar_fname = \fIlar_fname\fP (STRING)
lsf_flag = \fIlsf_flag\fP (SHORT)
lsf_fname = \fIlsf_fname\fP (STRING)
lsf_freq_res = \fIlsf_freq_res\fP (FLOAT)
fftcep_flag = \fIfftcep_flag\fP (SHORT)
fftcep_fname = \fIfftcep_fname\fP (STRING)
fftcep_order = \fIfftcep_order\fP (SHORT)
fftcep_deriv = \fIfftcep_deriv\fP (STRING)
fft_flag = \fIfft_flag\fP (SHORT)
fft_order = \fIfft_order\fP (SHORT)
.fi
.sp
.SH SEE ALSO
.PP
.nf
\fIframe\fP(1-ESPS), \fIpwr\fP(1-ESPS), \fIrefcof\fP(1-ESPS), \fIsgram\fP(1-ESPS), 
\fIzcross\fP(1-ESPS), \fIget_rfft\fP(3-ESPS), \fIfft_cepstrum_r\fP(3-ESPS), 
\fIget_auto\fP(3-ESPS), \fIget_atal\fP(3-ESPS), \fIrc_reps\fP(3-ESPS),
\fIgrange_switch\fP(3-ESPS), FEA(5-ESPS), FEA_SPEC(5-ESPS)
.fi
.sp
.SH BUGS
.sp
.SH FUTURE CHANGES
.sp
.SH AUTHOR
.PP
Man page and program by Bill Byrne (based on frame.c by John Shore).