Project WAM: A Wavelet Auditory Model for characterizing complex acoustic
signals
PI/Contractor John Benedetto, University of Maryland
John J. Benedetto
Department of Mathematics
University of Maryland
College Park, MD 20742
jjb@math.umd.edu
301-405-5161
http://www.norbertwiener.umd.edu
Background
The Wavelet Auditory Model (WAM) constructed by the contractor implements an
alternate method for characterizing acoustic signals (sounds) using a mathematical
model based on the mammalian auditory system. This model is based on a
theoretical framework that uses an iterative algorithm for reconstruction from
irregular samples. This is an effective method for dealing with speech compression
problems. The published technical abstract for the work on which this is based is as
follows:
“A time-scale representation of (acoustic) signals, motivated by the structure
of the mammalian auditory system, is presented. Drawing from the theory of
irregular sampling and frames, a theoretical framework is developed in
which an iterative algorithm for reconstruction is constructed. Numerical
examples are included which illustrate the validity of such a representation
as a new and effective method to deal with speech compression problems.” A
Wavelet Auditory Model and Data Compression; Authors: Benedetto J.J.;
Teolis A.; Source: Applied and Computational Harmonic Analysis, Volume 1,
Number 1, December 1993 , pp. 3-28(26); Publisher: Academic Press
Due to the novel framework used to attack speech compression, the possibility
exists to extend the implementation for arbitrary acoustic signals. By utilizing the
mathematical approach of frames, the power of wavelet transforms to provide
temporally specific characterization of complex acoustic signals could be
substantially increased. The Government would like to see the current theoretical
software re-written into a user-friendly environment so that analysts can explore
this approach.
Scope
The scope of this project is two-fold.
 Implementing WAM in its current form in a user-friendly, documented
format
 Extending WAM from a compression application into an application for
detecting and characterizing features and anomalies in complex acoustic
signals
WAM is uniquely suited to provide temporally specific characterizations of portions
of the acoustic signal with varying levels of coherence and vastly different
characteristics. Moreover, it should be able to overcome some of the limitations
commonly associated with incomplete or interrupted sampling strategies. In
particular, we expect that it will be a powerful tool to address background “noise”
found in multimedia samples.
The University of Maryland’s Norbert Wiener center for Harmonic Analysis and
Applications will provide the needed computing and academic resources for
successful completion of the project.
Deliverables
Deliverables will include reports, software, and limited support to the Government
in implementing the software.
 Reports: three reports a year beginning with Fall 2009, to correspond to
progress, documentation of effort, and projected future steps from the
previous academic period (fall, spring, summer)
 Software: user-friendly implementation of current WAM theoretical
framework (FY10)
 Software: user-friendly implementation of extended WAM for detection,
characterization and evaluation of features and anomalies in acoustic signals
 Support: getting all software to work on the project Laptop and Hard-drive,
and either the Gallery for Advancing Signal Processing (GASP) system, or
another system designated by the government.
Budget (to be fleshed out with overhead costs, etc)
FY09—Fourth Quarter
Salary/overhead for Research Programmer
$ 15 K
Laptop & 1T Hard-drive for test-data/software/deliverables $10K or as supplied
by Government
FY10
$ 125 K
Salary/overhead for Research Programmer
Limited salary/overhead for PI
Travel
Misc. Supplies
Option Year 1 (FY11):
$150 K
Salary/overhead for Research Programmer
Limited salary/overhead for PI
Travel
Misc. Supplies
Option Year 2 (FY12):
$150 K
Salary/overhead for Research Programmer
Limited salary/overhead for PI
Travel
Misc. Supplies
Timeline
FY09:
Ramp up
Project summary
FY10:
Coding, Documentation, Verification and Validation for WAM in its current form
Option Year One FY 11:
Extensions for WAM, to include detection and characterization of features and
anomalies in acoustic signals, with emphasis on some or all of the following:
Noise reduction
Synthetic noise
Option Year One FY 12:
Additional Extensions for WAM, to include detection and characterization of
features and anomalies in acoustic signals, to include some or all of the following:
Noise reduction
Synthetic noise