Instead of working with hundred thousend of soundfiles use
most fundamental feature of S_TOOLs-STx has to be made aware to the
user. STx processes the audio signals in soundfile documents as a
whole, such as originally recorded. The soundfiles are frozen as
originals, in the sense of a secure database, whereas the associated
metadata are treated dynamically and therefore free to be updated if
order to access and process individual sound events conveniently, such
as phonemes, words, traffic noise evets, animal sounds etc. on a
soundfile basis, an open concept of Soundfile segment definition and
annotation on the metadata level has been implemented. It imposes no
restrictions in access to the sound whatever, but opens the option of
multiple annotations to the same sounds if necessary, preserves the
original context of the recordings, enables relative addressing and
avoids the cumbersome management of hundreds of thousands of individual
sound files in dedicated data bases.
Concept of Soundfile-Segments
your soundfiles contain recorded music or interviews with phonetically
balanced text, or different environmental sounds of noise, machinery
noise or animal sounds recorded unattended under free field conditions
by automated monitoring systems. Recordings of this type usually have a
duration of several hours and frequently much more (up to weeks),
depending on the recording sessions arranged. In general, the principal
goal consists in collecting as much as relevant audio information on
the subjects under consideration as possible. By accessing the recorded
sound material the immediate problem of segmentation arises.
Phonetically balanced text has to be transcribed and annotated to the
word and phoneme level, environmental sounds have to be identified as
events and described accordingly as well as longterm monitoring
recordings of animal sounds. Usually for each segment a new soundfile
would be created in order to access the sounds. S_TOOLS-STx adopts
instead the strategy to create segment identification and soundfile
addresses (soundfile name, start sample, length in samples and
additional metadata), providing a unique reference for each sound
segment. The advantages of the concept described are:
- the original soundfile (essence) remains frozen in as an
- segmentation on a metadata level can be organised according
to tailored cirteria, depending on the questions to be posted to the
- dynamic and multiple segmentation and annotations can be
realized without duplicating the essence
- the original context (the signal before and/or after the
segmented region) remains accessible by relative addressing
- segment lists can be compiled using arbitrary metadata
elements for comparative listening, parameter statistics and evaluation
address specifications for segment definition in S_TOOLS-STx
can add attributes to any DataSet element by editing the properties of
the selected element. Attributes not automatically set by S_TOOLS-STx
are called 'user-defined' attributes. Set an attribute in the bottom
edit box with the format
songLyric=God save our gracious Queen; and delete it by
assigning no value (e.g. songLyric=;).
order to see user-defined attributes in the Detail, press the
button  to
for element attributes.
you need to edit the user-defined attributes often, you might like to
try developing a
to facilitate data entry.
can copy a DataSet element's attributes by right-clicking on the
attribute and selecting
A selection of the following will be available, depending the element
- Reference -
the path from the root of the DataSet to this element. This can be used
in macros to reference this element.
- ID - the ID of
this element. The ID is unique on this level of the DataSet.
- Segment - the
name of the selected segment
- AudioRef - the
Reference and the segment name separated by a semicolon (e.g.
- Filename - the
absolute path of the file associated with this element.
- More - click
Attributes -> Clipboard dialog box, where you can
select multiple attributes as well as formatting options.
A segment in a soundfile
is defined by an unique identifier ("name") in
the current soundfile, by a start address (in samples, s, ms, or
numerical expressions of them), by an end address
and optional taggings or annotations (such as textual
transcriptions) manual classifications or other metadata
associatet to the sound segment. The
metadata are stored in soundfile.name.xml text files, which can be
accessed inside or outside of S_TOOLS-STx.
Note: The user is responsible for
the administration of wave-files and their associated metadata files, especially in
case of migration procedures.
must be unique / each soundfile
in samples (Fs=44100Hz)
expressions are allowed in segment addressing
stands for begin of a segmernt
"e" for the end of a segment
"l" for the length (duration).
address expressions are
always evaluated in samples, independent wether specified in s, samples
or other units of time.
the "_" is used to separate from_to in order to remain reserved +/-*
for address expressions.