[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Dissertation: Data Reprocessing in Signal Understanding Systems

All the i's are crossed, and the t's are all dotted, and another PhD
has been minted!  My dissertation covers the evaluation of a signal
interpretation framework in auditory scene analysis, so I'm posting
this announcement on the chance that the work might be of interest to
people here.

You can find a link to the gzipped postscript document at


I'm currently working as a post-doc at the University of Massachusetts
at Amherst, and can be reached at klassner@cs.umass.edu.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

         Data Reprocessing in Signal Understanding Systems


                         Frank Klassner

    Submitted to the Department of Computer Science in partial
       fulfillment of the requirements for the degree of
            Doctor of Philosophy in Computer Science
           at the University of Massachusetts Amherst

                        September 1996


Signal understanding systems have the difficult task of interpreting
environmental signals: decomposing them and explaining their
components in terms of an arbitrary number of instances of perceptual
object categories whose properties can interact with one another.
This dissertation addresses the problem of designing blackboard-based
perceptual systems for interpreting signals from complex environments.
A ``complex environment'' is one that can (1) produce signal-to-noise
ratios that vary unpredictably over time, and (2) can contain
perceptual objects that mutually interfere with each others' signal
signature, or have arbitrary time-dependent behaviors.  The
traditional design paradigm for perceptual systems assumes that some
particular set of fixed front-end signal processing algorithms (SPAs)
can provide adequate evidence for reliable interpretations regardless
of the range of possible scenarios in the environment.  In complex
environments, with their dynamic character, however, a commitment to
parameter values inappropriate to the current scenario can render a
perceptual system unable to interpret entire classes of environmental
events correctly.

To address these problems, this research advocates a new view of
signal interpretation as the product of two interacting search
processes.  The first search process involves the dynamic,
context-dependent selection of signal features and interpretation
hypotheses, and the second involves the dynamic, context-dependent
selection of appropriate SPAs for extracting evidence to support the
features. For structuring bidirectional interaction between the search
processes, this dissertation presents the Integrated Processing and
Understanding of Signals (IPUS) architecture as a formal and
domain-independent blackboard-based approach. The architecture is
instantiated by a domain's formal signal processing theory, and has
four components for organizing and applying signal processing theory:
discrepancy detection, discrepancy diagnosis, differential diagnosis,
and signal reprocessing.  IPUS uses an iterative process of
``discrepancy detection, diagnosis, reprocessing'' for converging to
the appropriate SPAs and interpretations.  Convergence is driven by
the goal of eliminating or reducing various categories of
interpretation uncertainty.

This dissertation discusses the IPUS architecture's features, the
basic problem of auditory scene analysis (the application domain used
in testing IPUS), and evaluates performance results in experiment
suites that test the utility of the reprocessing loop and the ability
of the architecture to apply special-purpose SPAs effectively.
Although the specific research reported herein focuses on acoustic
signal understanding, the general IPUS framework appears applicable to
the design of perceptual systems for a wide variety of sensory