DSProlog is an experimental knowledge discovery support environment (KDSE) for
knowledge discovery in PCM-data. Also, it may be seen as an integrated development
environment (IDE) for design, implementation and test of subsymbolic
and symbolic classifier systems for deterministic patterns in PCM-signals (e.g. in audio-signals).
Tools provided by DSProlog cover the following areas:
interactive classifier system design,
data mining large PCM-file collections,
similarity based retrieval of PCM-data,
spectrographic subsymbolic classifier systems,
hierarchical symbolic classifier systems,
template based initialization of classifier systems,
scientific visualization of PCM-data, classifiers and classifier decisions,
database access to classifier decisions (SQL supported),
automated annotation of deterministic events in PCM-signals,
integrated project management,
filtering, visualization, browsing and sonification of PCM-data,
interfacing with ISO-Prolog.
The name DSProlog is a combination of DSP - digital signal processing - and
Prolog, a logic programming language created around 1972 by Alain
Colmerauer and Philippe Roussel. The idea behind DSProlog is, that
classifiers for complex patterns in signals may be represented as
expressions in declarative logic and be computed in a similar way
than Prolog predicates are. DSProlog is a way to bridge the gap
between dense PCM-data and symbolic descriptions of deterministic
patterns in the signal.
Example: Creating a symbolic classifier for a sweep
Step 1: Create spectrogram of wavelet (dt = 20ms)
Step 2: Derive subsymbolic classifier from spectrogram
Step 3: Annotate sweep (dt = 600ms) with subsymbolic classifier
Step 4: Derive symbolic classifier from set of annotations
Conditions and License
Currently, the software is provided on the basis of a strict non-commercial
research-cooperation agreement. Users will have to take on a little
job in order to advance the project. Currently open jobs are:
proofreading the manual, creating standardized PCM-data sets (with simple waveform patterns like sweeps and clicks) and testing the software. Please contact Dr. Sebastian Huebner:
sebastian@sejona.de
Hübner, S.:
Wissensbasierte Modellierung von Audio-Signal-Klassifikatoren. Zur Bioakustik von Tursiops truncatus,
Doctoral Thesis, University of Potsdam, 2007.
Hübner, S.:
Bioacoustic Classifier System Design as a Knowledge Engineering Problem,
Conference contribution at the
International Expert Meeting on IT-Based Detection of Bioacoustical Patterns, Vilm, 2007
url:https://www.tierstimmen.org/meeting_vilm2007.
