ref_cof.1

speech signal process tools

.\" 1.1 ESI
.TH REF_COF 1\-ESPS "5/21/87"
.ds ]W "\fI\s+4\ze\h'0.05'e\s-4\v'-0.4m'\fP\(*p\v'0.4m'\ Entropic Speech,Inc.
.SH NAME
ref_cof - computes reflection coefficients for frames of sampled data.
.sp
.SH SYNOPSIS
.B ref_cof
[
.BI \-P " param_fil"
]
[
.BI \-x " debug_level"
]
[
.BI \-r " range"
] [
.BI \-n " nframes"
] [
.BI \-f " filtername"
]
[
.BI \-s " syncflag"
] [
.BI \-a " analysis_meth"
]
.I in_file out_file
.sp
.SH DESCRIPTION
.PP
.I ref_cof
takes an ESPS sampled data file,
.I in_file,
and produces an ESPS FEA_ANA analysis file
.I out_file
containing the reflection coefficients corresponding to one or 
more fixed-length sampled-data frames.  
The reflection coefficients, along with the computed values for
.I raw_power
and
.I lpc_power,
are then stored in FEA_ANA records.  No pitch pulse information is
written to the file.
.PP
By default,
the program uses the "standard" method for computing the autocorrelation
function and the Levinson algorithm for computing the reflection coefficients
from the autocorrelation values, but other analysis methods
may be specified by using the 
.B \-a
option or through use of the 
.I method
variable in the parameter file.  Note that the standard
autocorrelation method may yield poor results when using a small
number (i.e.  < 100) of data samples.
.PP
If the \fB\-f\fP option is used, the function will optionally filter
the data with the named pre-emphasis filter prior to performing the analysis.
The filter coefficients are read from the parameter file and stored
in the header of the output file.
.PP
The program can also optionally multiply the autocorrelation function
by a sinc function (\fB\-s\f\P option) prior to computing the
reflection coefficients.  This has the effect of reducing the
spectral resolution if the spectrum of these coefficients is plotted.
.PP
If 
.I in_file
is "\-" then the input is read from the standard input and if
.I
out_file
is "\-" then the output is directed to the standard output.
.sp
.SH OPTIONS
The following options are supported:
.TP
.BI \-P " param_fil"
uses the parameter file 
.I param_fil
rather than the default, which is
.I params.
.br
.sp
.TP
.BI \-x " debug_level"
The program has two debugging levels: zero and non-zero.
.br
.TP
.BI \-r " first" : "last" "\fR [1:1000]"
.TP
.BI \-r " first" :+ "incr"
In the first form, a pair of unsigned integers specifies the range of
sampled data comprising the first frame.  Either
.I first
or
.I last
may be omitted; then the default value is used.  If 
.IR last " = " first " + " incr,
the second form (with the plus sign) specifies the same range as the
first.  If the specified range contains points not in the file, zeros
are appended. Both forms of the option override the values of 
.I start
and
.I fsize
in the parameter file.  The \fB\-p\fP option overrides possible 
values in the parameter file.  
.TP
.BI \-n " nframes [1]"
Specifies the number of frames to process, and overrides the 
.I nframes
value that may be in the parameter file.  The default applies only
if there is no value in the parameter file.  
.TP
.BI \-a " method"
Specifies the spectrum analysis method.  Two methods are supported:
"AUTOC" means the autocorrelation method [1] and STRCOV means
structured covariance [2].  If neither is specified, and no value is
in the parameter file, the autocorrelation method is used.  The
option value overrides any value in the parameter file.  
.TP
.BI \-f " filtername"
The program will perform pre-emphasis filtering with
.I filtername
the name of the pre-emphasis zfunc in the parameter file.
If this option is specified the program will read the numerator order and
denominator order from the parameters
.I nsiz
and
.I dsiz,
respectively.  The program then reads the numerator coefficients from
the array
.I filtername_num
and the denominator coefficients from the array
.I filtername_den.
.TP
.BI \-s " n"
The program will multiply the autocorrelation function by a sinc
function in order to reduce the spectral resolution of the
coefficients.  The autocorrelation function is multiplied by the
function
.I sinc (x/n) 
before computing the reflection	 coefficients.  In the frequency
domain this has the effect of convolving the spectrum with a boxcar
function of width
.I "f\ds\u/n,
where
.I f\ds\u
is the sampling frequency.  The value of
.I n
is recorded in a header comment.
.SH ESPS PARAMETERS
The following parameters are read from the parameter file:
.IP
.I "start - integer"
.IP
The first point in the input sampled data file that is processed.
A value of 1 denotes the first sample in the file.  
This is only read if the \fB\-r\fP option is not used.  
.sp
.I "fsize - integer"
.IP
The number of data points used to compute each set of reflection coefficients.
This is only read if the \fB\-r\fP option is not used.  
.sp
.I "nframes - integer"
.IP
The number of input data frames to process and of output sets of reflection co
efficients to produce.  If no value is given in the file, a value of 
one is assumed.  Both the parameter file value and the default value
may be overrided by means of the \fB\-n\fP option.  
.sp
.I "order - integer"
.IP
The number of reflection coefficients computed for each frame of input data.
If no value is given in the file, a default value of 15 is used.  
.sp
.I "nsiz - integer"
.IP
The number of filter coefficients in the numerator of the pre-emphasis filter
transfer function.
.sp
.I "method - string"
.IP
The spectrum analysis method to use. The autocorrelation (AUTOC)
and structured covariance (STRCOV) are currently supported.
.sp
.I "dsiz - integer"
.IP
The number of filter coefficients in the denominator of the pre-emphasis
filter transfer function.  A value of zero means that an FIR filter will be im
plemented
and the denominator coefficient array need not be entered.
.sp
.I "filtername_num - float array"
.IP
The numerator coefficients.  They are specified in order starting with a\d0\u.
The 
.I filtername
is specified by the \fB\-f\fP option.  
.sp
.I "filtername_den - float array"
.IP
The denominator coefficients.  They are specified in order starting
with b\d0\u.
The 
.I filtername
is specified by the \fB\-f\fP option.  
.SH ESPS HEADER
A new file header is created for the FEA_ANA file.
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.  
.PP
The program writes the usual values into the common part of the output header.
In addition,
.I ref_cof
writes the following values into the specified generic header items
in the output file:
.sp
\fIorder_vcd\fP = 0
.sp
\fIorder_unvcd\fP = order
.sp
\fImaxpulses\fP = 1
.sp
\fImaxraw\fP = 1
.sp
\fImaxlpc\fP = 1
.sp
\fIfilters\fP = NO
.sp
\fIspec_rep\fP = RC;
.sp
The following generic header items are also set appropriately:
.I "src_sf, start, nan, frmlen."
If a pre-emphasis filter was used the sizes, zeros (numerator), 
and poles (denominator) are recorded in generic header items
.I "filtername_siz, filtername_zeros, filtername_poles,"
where \fIfiltername\fP is taken from the \fB\-f\fP option.  
.PP
A generic header item \fIsinc_flg\fP is added to indicate whether or
not the \fB\-s\fP option was used.  
The corresponding value of
.I n 
is recorded in the header comments.
An item \fImethod\fP
is added with the appropriate spectrum analysis method,
and an indication of whether the DC component was removed
prior to spectral analysis
is added in the item \fIDC_removed\fR.
.sp
.SH SUBROUTINES CALLED
..br
getatal (3\-ESPS)
.sp
.SH  SEE ALSO
.PP
anaspec (1\-ESPS), plotspec (1\-ESPS), FEA_ANA (5\-ESPS), FEA (5\-ESPS),
SD (5\-ESPS).
.sp
.SH BUGS
.PP
None known.
.SH FUTURE CHANGES
.PP
The program will be expanded to include other methods for calculating the
autocorrelation function and the reflection coefficients than those
currently employed. 
.SH REFERENCES
[1] L. R. Rabiner and R. W> Schafer, 
.I Digital Processing of Speech Signals
Prentice Hall, NJ. 1978
.PP
[2] J. P. Burg, 
.I et al.,
"Estimation of Structured Covariance Matrices,"
.I Proceedings of the IEEE,
April, 1982
.sp
.SH  AUTHOR
.PP
Brian Sublett; modified for FEA_ANA files by John Shore; modified for 
structured covariance by David Burton