man.diff

来自「speech signal process tools」· DIFF 代码 · 共 458 行 · 第 1/2 页

.\" Copyright (c) 1992 Entropic Research Laboratory, Inc.; All rights reserved
.\" @(#)sfconvert.1	1.1 11/10/92 ERL
.ds ]W (c) 1992 Entropic Research Laboratory, Inc.
.TH  SFCONVERT 1\-ESPS 11/10/92
.SH NAME
sfconvert \- converts sampling frequency in a sampled data file by using a
lowpass interpolation filter designed by the windowing method

.SH SYNOPSIS
.B
sfconvert [
.BI \-P "param_file"
] [
.BI \-s " new_sample_freq"
] [
.BI \-r " range"
] [
.BI \-v " deviation"
] [
.BI \-c " corner_freq"
] [
.BI \-R " rej_db"
] [
.BI \-t " trans_band"
] [
.BI \-w " sfwin_type"
] [
.BI" \-l " sfwin_len"
] [
.BI \-e " channels"
] [
.BI \-o " output_type"
] [
.BI \-d
] [
.BI \-f
] [
.BI \-x " debug_level"
] [
.I input.sd
]
.I output.sd
.SH DESCRIPTION
.PP
.I Sfconvert
takes an input ESPS Sampled Data (FEA_SD) file,
.I input.sd,
and resamples it with a new sampling frequency specified by
.B \-s
and
.B \-v
options to produce an ESPS FEA_SD file,
.I output.sd.
By default, the new sampling frequency is the input sampling frequency.  
.PP
If \fIinput.sd\fR 
is replaced by "\-", the standard input is read.  If \fIoutput.sd\fR
is replaced by "\-", the standard output is written.  If \fIinput.sd\fR
is missing from the command line, \fIsfconvert\fR gets its name from
ESPS Common.
.PP
The actual new sampling frequency is of the form 
(\fIup\fR/\fIdown\fR)*\fIsf_in\fR.  The variable \fIsf_in\fR is the sampling 
rate of
the input file and the variables \fIup\fR and \fIdown\fR are 
interpolation and decimation factors.  They are the smallest 
integers such that the actual new sampling frequency is deviated from 
\fInew_sample_freq\fR with the maximum tolerable deviation in percent, 
\fIdeviation,\fR from \fInew_sample_freq\fR.  The default of \fIdeviation\fR is zero.
.PP
\fISfconvert\fR supports multichannel input data.  Single or multiple 
channels selected are 
extracted from \fIinput.sd\fR, resampled, and written in a single or
multichannel \fIoutput.sd\fR.  Channels are selected with the 
\fB\-e\fR options.  Default is all channels.
.PP
\fISfconvert\fR supports data types of DOUBLE, FLOAT, LONG, SHORT, BYTE,
DOUBLE_CPLX, FLOAT_CPLX, LONG_CPLX, SHORT_CPLX, and BYTE_CPLX.  The data type
of \fIouput.sd\fR by default is the same as that of \fIinput.sd\fR.  But it
can be set to any data type by \fB\-o\fR option.  If input data type is complex
and output data type is set to real, imaginary part of input data is 
discarded.  If
input data type is real and output data type is set to complex, zeros are 
filled in the imaginary part of output data.  Converting one data type to
another with less dynamic range will result in clipping.
.PP
Actual resampling is
computed via floating point or double precision operation.  By default, if 
\fIinput.sd\fR
is of any data types other than DOUBLE or DOUBLE_CPLX, resampling is carried 
out in
floating point computation.  Otherwise, double precision is used.
\fB\-d\fR option forces all computation to be carried out in double precision,
regardless of input data types.
.PP
\fISfconvert\fR first designs a non-causal type I lowpass FIR filter (a
filter with odd length and with filter shape symmetrical about its center
point) by the windowing method.  By default, the Kaiser windowing method is 
used to 
design the lowpass filter with the transition bandwidth, \fItrans_band\fR by 
the \fB-t\fR option, between
passband and stopband, and with the rejection ratio, \fIrej_db\fR by the 
\fB-R\fR option, in dB values
from passband to stopband.  Such filter is then convoluted with
input sampled data.  
.PP
The parameters,
\fItrans_band\fR and \fIrej_db\fR determine the filter length designed by  the 
Kaiser windowing method and they have direct impact
on computational time.
The filter length is directly proportional to \fIrej_db\fR and 
inversely proportional to \fItrans_band\fR.
.PP
The lowpass filter can be designed by the windowing method using other window 
types,
such as rectangular, Hanning, Hamming, triangular, and cosine^4 windows.
The \fB-w\fR option specifies which window type to use and the \fB-l\fR option 
specifies the time duration of the filter length in seconds.  
 Kaiser windowing method allows the lowpass 
filter to be designed from the criteria of stopband rejection ratio and 
transition bandwidth, but these other windowing methods do not use such
criteria.
.PP
The paramters \fIrej_db\fR and \fItrans_band\fR for the Kaiser windowing
design method may coexist with \fIsfwin_type\fR and \fIsfwin_len\fR for
other windowing design method in command line or in paramter file.
A non-Kaiser windowing design is used only when the \fB-w\fR option
exists on command line, or when the \fIsfwin_type\fR paramter in a paramter
file is set to a non-Kaiser window and \fIKaiserflag\fR is set to zero in 
a paramter file.  Otherwise, Kaiser windowing design method is used.
.PP
In the case when the input data is too short to convolute with the designed
filter, zeros are padded to input data.
.PP
By default the corner frequency or cutoff frequency for the
lowpass filter is the Nyquist frequency,
(the output sampling frequency divided by two).  Since the transition
bandwidth must be nonzero, in many cases
it is desirable to have a broad transition region that occurs
entirely below Nyquist frequency.  The \fB-c\fR option sets the corner
frequency, \fIcorner_freq\fR, in Hertz.
.PP
.SH OPTIONS
.PP
The following options are supported:
.TP
.BI \-P " param_file"
The file \fIparam_file\fR is used for the parameter file instead of the 
default, which is \fIparams\fR.
.TP
.BI \-s " new_sample_freq \fR[input sampling frequency]\fP"
Sampling frequency of \fIoutput.sd\fR.  Default is the input sampling
frequency.
.TP
.BI \-v " deviation \fR[0]\fP"
Maximum tolerable frequency deviation from \fInew_sample_freq\fR in percent.
For example, \fIdeviation=2\fR means the actual output sampling frequency is
the requested sampling frequency plus or minus 2%.  This options allows 
\fIsfcovnert\fR to find the smallest integer conversion factors, 
\fIup\fR and \fIdown\fR, within the allowable range, and thereby
reduce the problem of generating long filter lengths in the Kaiser windowing
method due to a large interpolation factor.  Note that this only affects
memory requirement.
.TP
.BI \-r " first:last"
.TP
.BI \-r " first:+incr"
Determines the range of points from input file for frequency conversion.  In
the first form, a pair of unsigned integers gives the first and last points
of the range.  If \fIfirst\fR is omitted, 1 is used.  If \fIlast\fR is
omitted, the last point in the file is used.  The second form is equivalent
to the first with \fIlast = first + incr\fR.
.TP
.BI \-c " corner_freq \fR[output sampling frequency/2]\fP"
Corner frequency of the lowpass interpolation filter designed by the windowing 
method.  If this paramter is not specified or a value of zero is given, a
default value of the Nyquist rate of output data file is used.
.TP
.BI \-R " rej_db \fR[60]\fP"
Rejection ratio in dB values from passband to stopband in the lowpass filter
designed by the Kaiser windowing method.  \fISfconvert\fR ignores this option
if a non-Kaiser windowing design method is used.
.TP
.BI \-t " trans_band \fR[200]\fP"
Transition bandwidth in Hertz from passband to stopband in the lowpass filter
designed by the Kaiser windowing method.  \fISfconvert\fR ignores this option
if a non-Kaiser windowing filter design method is used.
.TP
.BI \-w " sfwin_type"
If Kaiser windowing method is not desired in designing the lowpass filter, 
another
window type can be used.  \fIsfwin_type\fR can be set to any of these
window types: WT_RECT (rectangular), WT_HAMMING (Hamming), WT_TRIANG
(triangular), WT_HANNING (Hanning), and
WT_COS4 (cosine^4).  If this option is selected, the \fB-R\fR and \fB-t\fR
are ignored.
.TP
.BI \-l " sfwin_len"
Time duration in seconds of the window selected by \fB-w\fR option. 
\fISfconvert\fR
ensures the window length in samples is an odd number by 
adding 1 to the multiplication product of \fIsfwin_len\fR and the 
maximum of \fIup\fR or \fIdown\fR, if the product
is even.  This is to ensure the lowpass interpolation filter is
a type I filter.  \fISfwin_len\fR must be greater than zero.  \fISfconvert\fR
ignores this option if the Kaiser windowing filter design method is used.
.TP
.BI \-e " channels \fR[all channels]\fP"
Determines which channels are extracted out of \fIinput.sd\fR for
sampling frequency conversion.  The format is that of a comma separated
list of integers and pairs \fIa:b\fR, where \fIa\fR and \fIb\fR are integers.
This specifies the channel numbers that are selected for output.  
For example, \fI2,5:8,12\fR specifies
channel 2, channels 5 through 8, and channel 12.  Additionally, an expression
\fIa:+c\fR may be written instead of \fIa:b\fR where \fIc\fR is an integer
such that \fIa+c=b\fR.  Thus \fI5:8\fR could be replaced with \fI5:+3\fR
in the example.

The numbering of channels begins with 0.  The channel numbers must be
specified in increasing order without repetitions.  If this option is not
specified, the default is to select all channels in the input file.
.TP
.BI \-o " output_type \fR[input data type]\fP"
This option specifies the data type of \fIoutput.sd\fR.  Available data types
are DOUBLE, FLOAT, LONG, SHORT, BYTE, DOUBLE_CPLX, FLOAT_CPLX, LONG_CPLX, 
SHORT_CPLX, and BYTE_CPLX.  By default, output data type is the same as
input data type.
.TP
.B \-d