rel40.me

speech signal process tools

.np
The programs \fIframe\fP, \fIwindow\fP, \fIpwr\fP, and \fIzcross\fP
have all been changed to no longer copy all of the input fields to the
output.  Now, the output file contains a single field in each case
(frame of data, windowed frame, power, zero crossing rate) \- the
information computed by the particular program.  If combined
information is needed, use \fImergerfea\fP.  These programs have also
been fixed to handle correctly various cases of tagged and segment
labelled files.  
.np
A new option (\fB-a\fP) was added to \fIcopysd\fP to permit adding a
constant to the data before writing it.  Together with \fB-s\fP
(scale), this means that \fIcopysd\fP can both scale and shift when
copying.  Note that \fIfea_func\fP can also be used (now that sampled
data files are FEA files.)  Also, the \fB-d\fP option on \fIcopysd\fP
has been extended to support conversion to any data type.  Generic
header items in the output file record the scale and addition
constants.  
.np
\fIsetrange\fP has been changed to work on arbitrary FEA files.
\fIsetrange\fP accepts a range specified in seconds, transforms this
to a range of records, and writes the results to ESPS Common (in terms
of the parameters \fIstart\fP and \fInan\fP) for use by other
programs.  The revised version of \fIsetrange\fP uses the generic 
header items start_time and record_freq in performing the
transformation.  \fIsetrange\fP will still work on old-style SD files.
.np
\fIbtosps\fP has been generalized so that it will convert headerless
sampled data of any ESPS data type to ESPS FEA_SD format; this
includes complex data.  Also, \fIbtosps\fP will now convert
multi-channel data.  New options include \fB-t\fP (input data type),
\fB-n\fP (number of channels), \fB-T\fP (start time), \fB-c\fP (ASCII
comment), \fB-C\fP (ASCII comment file), and \fB-x\fP (debug level).
For real (non-complex) data that is not on a pipe, \fIbtosps\fP will
also determine the maximum absolute value and write the result in the
generic header item \fImax_value\fP.  The conversions provided by
\fIbtosps\fP can also be achieved using \fIaddfeahd\fP
(1\-\s-1ESPS\s+1), \fIaddgen\fP (1\-\s-1ESPS\s+1), and \fIsetmax\fP
(1\-\s-1ESPS\s+1), but \fIbtosps\fP is more convenient in the case of
sampled data.  Note that the \fB-s\fP option (number of leading bytes
to skip) has been changed to \fB-S\fP, and that a comment must now be
supplied for record keeping purposes.  
.np
\fIaddfea\fP has a new option (\fB-T\fP) to permit the explicit
setting of a FEA subtype code in the output FEA file.  This permits
conversion, for example, of ASCII data to multi-channel FEA_SD.  In
principle, you could go from ASCII to FEA_ANA, etc. 
.np
\fImake_sd\fP has a new option (\fB-S\fP) to set the sampling rate of
the output FEA_SD file.  Also, \fImake_sd\fP was modified to preserve
the data type of the field specified in the input file.  
.np
\fIfea_deriv\fP has a new option (\fB-T\fP) to set a FEA subtype code
in the output file.  Another new option (\fB-t\fP) causes tags to be
copied from the input to the output.  
.np
\fIxmcd\fP and \fImcd\fP were fixed to remove the temporary file when
killed.  
.np 
The previous version of \fIfft_filter\fP ignored any range
specification; this has been fixed.
.np
The parameter prompts for \fIauto\fP (when via \fIeparam\fP) have
been improved.  
.np
\fIespsenv\fP has been updated to add ESPS_EDR to the environment 
variables reported.  
.np
\fIAddgen\fP now adds the user's name to the comment put in the output
header.  
.np
An option (\fB-S\fP) was added to \fItofspec\fP to disable scaling.  
Input data can now be copied to the output FEA_SPEC file (e.g., for
display using \fIwaves\fP+) without scaling to an output range.  
.np
The order of the file parameters in \fIzero_pole\fP and \fIatofilt\fP
have been changed so that the order is input file followed by output
file (the general convention for ESPS).  This will break scripts that
use the old versions.
.np
The possible functions to be applied by \fIfeafunc\fP was expanded to
include squaring (sqr).  Note that, because \fIfeafunc\fP allows
specification of the output file type, it can be used to transform the
data type of any FEA field (the only restriction is that it will not
cross the boundary between complex and non-complex; e.g., you can go
from DOUBLE_CPLX to BYTE_CPLX but not from DOUBLE_CPLX to BYTE).  
.np
The X version of \fIimage\fP now parses the standard option for X
geometry.  
.np
User-level programs implemented as shell scripts have been modified
to use full paths when calling ESPS programs (examples include
\fIsf_convert, plotsgram, plotsd, \fPetc.).  This avoids problems that
occur when there are name conflicts with local programs.
.np
Man pages are now included for \fIemakecats\fP and \fImakewhatis\fP.  
.np
Default parameter files for Paddfea \fIdspsgram, fft, refcof, testsd,
transpec, sgram,\fP and \fIbtosps\fP have been revised to include
parameter choices and limits.
.np
A new option on \fItofspec\fP reverses the order of elements in the
field.  This is useful when converting files for display in
\fIwaves\fP+, since the effect is turn turn the display upside down.
.np
The \fB-r\fP for \fIlpcsynt\fP has been changed to the ESPS convention
of specifying the input record range.
.sh 1 "ESPS LIBRARY ADDITIONS AND CHANGES"
.sh 2 "Changes and Additions to Support Machine-Independent I/O (MIIO)"
.lp
Two items were added to the universal portion of the ESPS header
structure: \fIedr\fP and \fImachine_code\fP.  The former indicates
whether the file is stored in EDR or NATIVE format.  The latter
indicates the machine type that produced the file.  See ESPS
(5\-\s-1ESPS\s+1).
.lp
The information contained in the \fIedr\fP and \fImachine_code\fP
header items actually is stored in a short preamble that comprises the
first few bytes of an ESPS file.  The \fIread_header\fP routine first
reads the preamble (which is always stored in EDR format), determines
whether the file is an ESPS file, determines the file format (EDR or
NATIVE), and then either exits (if not an ESPS file) or reads the rest
of the information for the ESPS header structure.  Users should keep
in mind that the C structure shown in the man page ESPS
(5\-\s-1ESPS\s+1) (and defined in header.h) does \fInot\fP correspond
to the actual format stored in the file, which is why headers I/O
must be done with \fIread_header\fP and \fIwrite_header\fP.  
.lp
When a new header structure is created (by \fInew_header\fP or
\fIcopy_header\fP), the \fIedr\fP flag in the header is set to 1 (EDR
format) if the user has a UNIX environment variable ESPS_EDR set to
"on"; otherwise, \fIedr\fP is set to 0 (NATIVE mode).  This initial
setting can be changed by programs (e.g., in some cases one might want
to have a command-line option to force a particular format regardless
of the environment variable) before calling the I/O routines.  The
format used in writing the header is determined by the value of the
\fIedr\fP flag when the header is passed to \fIwrite_header\fP.
Similarly, the format of file records is determined by the value 
of the \fIedr\fP flag when the header is passed to (for example)
\fIput_fea_rec\fP.  (Yes, you can get into serious trouble if you fool
around with inconsistent header structures.)  
.lp
All of the library routines that support header and record I/O were
modified to support MIIO, including \fIread_header, write_header,
get_fea_rec,\fP and \fIput_fea_rec.\fP.  The \fIget\fP and \fIput\fP
routines for the various non-FEA file types were also modified (i.e.,
for SD, FILT, SCBK, and SPEC files).  The I/O routines for the various
FEA subtypes (e.g., FEA_ANA, FEA_SPEC, etc.) did not have to be
modified since they work by calling \fIget_fea_rec\fP and
\fIput_fea_rec\fP.
.sh 2 "Changes and Additions to Support Complex Data" 
.np
All of the FEA support functions have been generalized to support the
new complex data types BYTE_CPLX, SHORT_CPLX, LONG_CPLX, FLOAT_CPLX,
and DOUBLE_CPLX.  In particular, the following have been generalized: 
.sp .5
.nf
.ta 1.5i
  \fIadd_fea_fld\fP \-	add definition of new field to ESPS FEA file
  \fIallo_fea_rec\fP\-	allocate memory for an ESPS FEA file record
  \fIcopy_fea_rec\fP \-	copy data from one ESPS FEA record to another
  \fIget_fea_rec\fP \-	get the next data record from an ESPS FEA data file
  \fIprint_fea_rec\fP \-	print an ESPS FEA data record
  \fIprint_fea_recf\fP \-	print fields in an ESPS FEA data record
  \fIput_fea_rec\fP \-	put an ESPS FEA data record onto the file
  \fIget_fea_siz\fP \-	get the size information for a ESPS FEA file field

.fi
Each complex type has a real and an imaginary part of the
corresponding integer or floating type.  For details see the FEA man
page FEA (5\-\s-1ESPS\s+1).
.np
The following new functions support the processing of complex FEA files:
.sp .5
.nf
.ta 1.5i
  \fIis_file_complex\fP \-	does file contain any complex data types
  \fIis_field_complex\fP \-	is data type of field complex
  \fIis_type_complex\fP \-	does type code represent complex data
.fi
.np
A real-part addition (\fIrealadd\fP) function was added to the
functions that support complex arithmetic (see \fIcomplex\fP
(3\-\s-1ESPS\s+1).  
.np
Various support functions that operate on arrays of different data
types have also been generalized to support the complex types, including:
.sp .5
.nf
.ta 1.5i
  \fIarr_alloc\fP \-	allocate storage indexed as multidimensional array
  \fImarg_index\fP \-	allow indexing of allocated storage as multidimensional array
  \fIarr_free\fP \-	free storage allocated by arr_alloc or marg_index
.fi
.np
The following signal FFT functions have been added to support complex
processing: 
.sp .5
.nf
.ta 1.5i
  \fIget_cfft\fP \-	Compute fast Fourier transform of a FLOAT_CPLX data sequence.
  \fIget_cfftd\fP \-	Compute fast Fourier transform of a DOUBLE_CPLX data sequence.
  \fIget_cfft_inv\fP \-	Compute inverse fast Fourier transform of FLOAT_CPLX data 
  \fIget_cfftd_inv\fP \-	Compute inverse fast Fourier transform of DOUBLE_CPLX data
.fi
.sh 2 "Changes and Additions to Support FEA_SD" 
.np
The usual access functions have been provided for FEA_SD: 
.sp .5
.nf
.ta 1.5i
  \fIinit_feasd_hd\fP \-	initialize a FEA file header for subtype FEA_SD
  \fIallo_feasd_recs\fP \-	allocate memory for FEA_SD file records
  \fIget_feasd_recs\fP \-	get data records from an ESPS FEA_SD data file
  \fIget_feasd_orecs\fP \-	get overlapping data from an ESPS FEA_SD data file
  \fIput_feasd_recs\fP \-	write data records to an ESPS FEA_SD file
.fi
.sp .5
Note that FEA_SD supports independent specification of the data type
stored in the file and the data type stored in memory.  (This is a
feature of the old-style SD file type that we have preserved.)  Thus,
one specifies the desired in-memory type for an array of FEA_SD
records when allocating via \fIallo_feasd_recs\fP, and the
\fIget_feasd_recs\fP or \fIget_feasd_orecs\fP take care of any
necessary conversions on input.  
.np
\fIeopen\fP (ESPS file, read header, and check type) was modified so
that, if a FEA_SD header is requested when an old-style SD file is
read, the SD header is converted to a FEA_SD header and returned.
This change makes it trivial to convert many programs from SD to
FEA_SD (provided that they were written using \fIeopen\fP).  By
changing a call like
.nf
.ft CW

   in_file = eopen("filter", argv[optind++], "r", FT_SD, NONE, 
                   &ih, &fpin);
\fRto\fP
.ft CW

   in_file = eopen("filter", argv[optind++], "r", FT_FEA, FEA_SD,
                   &ih, &fpin);

.fi
.ft LR
the revised program will not only read the new FEA_SD files, but it
will continue to read the old-style SD files.  For more information
about conversions, see the new document "FEA_SD Files \- Feature Files for Sampled Data in ESPS".
.np
To further support the conversion of old programs, the following I/O
routines for SD files were modified so that they also work on FEA_SD
files.
.sp .5
.nf
.ta 1.25i
  \fIget_sd_recd\fP \-	get data from an ESPS sampled-data file in DOUBLE format
  \fIget_sd_recf\fP \-	get data from an ESPS sampled-data file in FLOAT format
  \fIget_sd_recs\fP \-	get data from an ESPS sampled-data file in SHORT format
  \fIget_sd_orecd\fP \-	get overlapping data from an ESPS SD file in DOUBLE format
  \fIget_sd_orecf\fP \-	get overlapping data from an ESPS SD file in FLOAT format
  \fIget_sd_orecs\fP \-	get overlapping data from an ESPS SD file in SHORT format
  \fIput_sd_recd\fP \-	put data onto an ESPS sampled-data file from DOUBLE data
  \fIput_sd_recf\fP \-	put data onto an ESPS sampled-data file from FLOAT data
  \fIput_sd_recs\fP \-	put data onto an ESPS sampled-data file from SHORT data
  \fIget_sd_type\fP \-	get the type of data in an ESPS sampled-data (FEA_SD or SD) file
.fi
.sp .5
That is, if these routines are called with a FEA_SD header, they will
do the right thing (assuming that the FEA_SD header properly describes
the data in the file).  In particular, suppose an existing SD program 
opens files via \fIeopen\fP and does I/O via the foregoing functions.
Then, if the calls to \fIeopen\fP are modified as described above,
the I/O calls need not be modified \- they will work regardless of
whether the input file is SD or FEA_SD.  
.sh 2 "Changes and Additions to Support Enhanced Parameter Processing" 
.lp
The syntax of ESPS parameter files has been extended to include
optional limitations on a parameter's value.  Two cases are supported:
(1) a list of discrete choices for the parameter (all data types); (2)
a minimum and maximum value for the parameter (numeric data types).
The new information is added to the symbol table constructed by
\fIread_params\fP (3\-\s-1ESPS\s+1) and is made accessible to user
programs by means of access functions.  The limitations are not
enforced by the lowest level ESPS parameter routines (e.g., the
existing \fIgetsym\fP() routines have not been modified).  Higher
level functions and programs can choose whether or not to exploit the
information and how to do so.  For example, the new \fIxprompt\fP
(1\-\s-1ESPS\s+1) program enforces the parameter limitations if they
are given.  
.lp
The following functions have been added to support parameter file
processing:
.sp .5