David M. Almeida
Research Interests:
Academic Research
Projects:
- Development
of Active Map Algebra, a Map
Algebra-inspired computational framework for Wireless Sensor Networks,
with Prof. Kate Beard,
University of Maine Dept of Spatial Information Science and Engineering, Orono, Maine.
- Investigation of Open Source geospatial software models and standards at CARSI (Center for the Advanced Research of Spatial
Information) Lab,
with Prof.
Sean Ahearn, Hunter
College,
NY,
NY
- Investigation
of use of Virtual Globes and
high-resolution DEMs for visualization of
flooding and storm surge events, with Prof. Frank Buonaiuto, Hunter College, NY, NY
(as coursework).
- Developed MAPS (Multi-Agent Pedestrian
Simulation), in conjunction with a Planner from the NYC Dept of City
Planning (initially as coursework).
- Automated analysis of NASA MODIS time series
and Investigation of integration of NASA EOS satellite data with
Environmental Protection Agency BASINS, with Dr. Wenge
Ni-Meister, Hunter College, NY, NY (as
coursework).
Industrial Research
Projects:
- Developed
the FIGS (Field Information GIS System) prototype for the United
Nations (Nov 2006-Feb 2007), using FOSS Geospatial Software
components. This attempted to use FOSS (Free and Open Source
Software) geospatial components to put together a field station-based
prototype for a Field-based Humanitarian Information System (Geospatial
Database Technology Consultant).
Projects Below were supported by American Express/Lehman
Brothers Internal Technology Grants during 1989-1992.
- Speaker-Independent
Speech Recognition (SISR) Project: Worked along with several research
institutions (including Carnegie Mellon and SRI) and vendors to find
Speaker-Independent speech recognition systems that would perform
acceptably in the harsh and noisy environment of the trading floor.
Specific work included integration of candidate systems into Multi-modal
Input System software, domain-specific speech corpus collection,
background noise collection, labeling of collected speech, grammar design
for specific speech recognizers, and rationalization of grammars into a single
model to work with the API. (Approx. $500 thousand. Technical Lead)
- SISR
Project (above) also included the Multi-modal Computer Input Project:
Constructed system software to support Multi-modal application input by
imposing a Finite State Machine model on X-Windows applications. The
API allowed keyboard and mouse input to mix seamlessly with speech
recognizer and other forms of input. New input devices or systems
could then be added and tested on systems that incorporated the API.
- Speaker-Dependent
Speech Recognition Project: Attempted to use a rule-based system and
pattern matching to capture training data from Verbex
speech recognizers in order to automatically re-train them at a later
time. (Approx. $1.5 million. Technical Lead)
- Handwriting
Recognition Project: Investigated the use of handwriting recognition
systems for trading ticket entry. Worked with several vendors (Grid,
AT&T, Scenario, Nestor) to architect
prototype systems for trading floor use. (Approx. $ 350,000.
Technical Lead)