EE Home Page > EE Undergraduate Labs > EE 442 Home Page > 2002-2003 Abstracts

Advisors:   Kostas Daniilidis

The final product of this project consists of a robot that will be controlled via the input of a PC.
The user end of the system will have PC available with a close-to-real-time image.  The user can
point to the desired location on the image with a mouse.  The two PC’s communicate with each
other through the Linksys Wireless CompactFlash Cards WCF12.  The information from the user-
end of the system is transferred to the robot-end through a serial cable and translated into commands
that can be sent to the Motorola 68HC11 that control the 9V motors of the car.  The final result is
the movement of the car to the desired position.   

Infrared Sensor Based Position Locator for Robot Gaming System

Authors:    Jonathan Ha
                Sangyop Lee

Advisors:  Insup Lee
                David Arney

The remote-controlled robot gaming system requires real-time location data for each robot car. 
An infrared emitter station will be placed in each corner of the game room and a circular infrared
detector array will be mounted on the roof of the robot car.  Only one emitter station will send
its infrared signal pulse to the car at any given time, implementing a round-robin time-sharing
system.  Once received by the robot, the intensity of the signal will be measured by the change
in output current from the infrared detector.  It was determined that there is a direct correlation
between this change in current and the relative distance and alignment between the infrared emitter
and detector.  By correlating three of the strongest measurements from the circular array to a
parabolic function, the relative angle from the car to the corner can be determined.  Using at least
two of these angle measurements, the position of the robot can be triangulated, producing (x, y)
coordinates that are sent to the main computer.

Matching Order Graduation System  

Authors:    Kian Lahji
                 Thomas Leri
                 Tyler Mullins

Advisor: Jorge Santiago

Most graduation ceremonies lack a personal feel for the graduating students in that the procedure typically involves waiting in line to receive a diploma, with each specific graduate eventually walking across the stage to receive the diploma once that student’s name is announced. Unfortunately, at large venues, it becomes increasingly difficult to distinguish once graduate from the next; this becomes a burden on family and friends who attempt to witness their loved one in such a joyous occasion. The MarchingOrder project is designed to allow a more personal feel to graduation ceremonies. The web-based system enables students to create their own “Graduation Page” which will be displayed on a public monitor while they walk across the stage to receive the diploma. Thus, an identifiable picture of each student will be available for the audience to view so that each graduating student will be discernible to the crowd from afar. In addition, administrators are able to monitor the status of their students by searching their records online; this eliminates the need for cumbersome and tedious paperwork which is currently the standard.

Non-contact NIR Single Pulse Lase Functional Imaging System

Authors:    Darwin Choi
                 Adrian Lau
                 Johnny Wong

Advisors:  Britton Chance
                Nader Engheta

An NIR Single Pulse Laser Functional Imaging System is reproduced with an additional non-contact
functionality, enabling the system to remotely scan the target medium from a distance. The hardware
system is put together using optoelectronic devices. The signal analysis methodologies are based on
Monte Carlo simulation results and experimental calibrations. From the analysis process, functional
information representating the blood oxygenation level could be obtained. By continuously obtaining
data, a real-time metabolism rate monitor is designed using the system.

Authors:    Moacir Daniel Araujo
                 Jorge Trujillo

Advisor:    Kostas Daniliidis

This project is an implementation of a facial tracking system that uses
color-based information to track a subject’s face in real time. The
system consists of a USB web camera mounted on a custom made RS-232
controlled motor that pans so as to always keep the desired face in the
center of the screen. To operate, the system is first initialized and
color information from the user is extracted and stored. This
information is then used in a statistics based algorithm that operates
on individual pixel color values to identify the location of the
subject’s face. To minimize the computational overhead involved with
scanning every pixel of each frame when searching for the face, a smart
search algorithm that searches on average one twenty-fourth of the total
number of pixels was devised. The full implementation of the system
allows for real-time tracking of a subject’s face at a rate of over
fifteen frames per second and can follow this subject over a 180º field
on the horizontal axis and a 90º on the vertical axis.

Implementation of Software Architecture on Sony AIBO

Authors:    Yen Chu Cheng
                 Kian Neng Lim

Advisor:    Dan Lee

The Robot World Cup (RoboCup) competition constitutes part of the artificial intelligence (AI) community’s initiative to promote intelligent robotics and AI research. The ultimate goal of the competition is the development of robotic soccer players that can beat a human World Cup champion team by 2050. The University of Pennsylvania has been taking part in the Sony AIBOcategory of the competition since 1999. As a result of Sony’s continued effort in developing the AIBO, these robotic dogs are now able to communicate with one another via a wireless card. In the most recent competition, the Penn team did not have a software architecture to make full use of the robot’s wireless capabilities to execute coordinated strategies. This lack of coordinated strategies gives rise to this project. There are two specific project goals:

• To design a software architecture/network that will ensure efficient and reliable communication between the robodogs

• To implement a decision-making network that will optimize team strategies

Autonomous Robot Vacuum Cleaner

Authors:   Jon Angelo
                Shih-hwa Kim
                
Advisor: 

The Autonomous Robot Vacuum Cleaner (ARVaC) vacuums the surface area of a flat, closed floor space without human control. ARVaC uses infrared (IR) proximity sensors to detect sharp edges, which serve as pivots for movement. The information related to navigation is then processed by a 68HC11 microcontroller, which control the locomotion algorithm and the vacuum cleaner.

New DAC to Improve Audio Experience

Author:    Nihaar Gupta
               Rupesh Kapadia

Advisor:   Dwight Jaggard

The component described which performs the digital to analog conversion of the PCM Redbook
CD audio format with the precision of 16 bits and zero oversampling conversion technology. This
component will be compared to a DAC that performs the conversion at an 8x overasmpling rate
and uses a one-pole low pass filter at 100 kHz. The purpose is to demonstrate that SNR and Total
Harmonic Distortion are not as important measurements of audio quality as jitter, time smear and
phase linearity.  The net result is improved tonal, spatial, temporal qaulities.  

Optical Interconnect System

Authors:    Jonathan Krafcik
                 Kiran Pallegada

Advisor:    Jorge Santiago

This project involved the design, fabrication, and evaluation of an optical interconnect system developed with the resources available at Penn. The silicon-based Microfabrication Lab was used to fabricate an SU-8 optical waveguide that couples with a visible red laser diode on its input side and a photodiode on the output. HC6811 microcontrollers were programmed that modulate data onto the output of the laser diode and then receive the electrically converted optical signal from the photodiode. Figures of merit include the expected operating speed in comparison to current copper wire technology (considering the resource constraints on the project) and also the bite error rate (BER) as an indicator of the optical waveguide's functionality.        

Authors:    Litton Chen
                 Dan Keeley

Advisor:    Insup Lee
                 David Arney
 
The fighting robots are controlled by Java-embedded JStamp micro-controller boards. 
The robots are capable of running independently of each other.  These boards control
all operations onboard the robots including the weapon system, sensor system, and motor
control.  The weapon system consists of two servos in a pan-and-tilt setup with an infrared
transmitter mounted on top for a range of firing directions.  The sensor system consists of
four sensors covering the wheels of the robot to detect hits on the robot and disable the
appropriate wheel at that time.  Finally, the JStamp board will control the motors that will
drive the robot.  The robot will accept commands as to what action to perform via a serial
link to a nearby computer.

Adaptive Controls for Multiple UAVS

Authors:    Florence Chan
                Chintan Jhunjhunwala
                Aalok Mehta

Advisor:   George Pappas

Ever imagine what would happen if the planes of the Blue Angels flew without their pilots?

The hierarchical controller developed is a hybrid system that provides an unmanned aerial vehicle (UAV)
with the autonomy necessary to alter its flight dynamically.  Each level in the hierarchy gradually transforms
mission commands into signals that adjust the control surfaces of a virtual UAV.  A simulation environment
based on commercial off-the-shelf flight simulation software displays the flight paths of several autonomous
UAVs working together to concert flight formations.  In the future, the simulator could be expanded to
simulate tasks that require other forms of coordination among several intelligent agents.

Pookie: A Digital Circuit Simulator Including  the Motorola 68HC11 

Authors:    David Cohen
                 Nicholas Ng
                 John Wolff

Advisor:    Jonathan Smith

This software project is a digital circuit simulator, similar to Xilinx, that includes many of the
components used in the EE lab in digital logic circuits.  Unlike Xilinx, it includes, as a component,
a simulation of the Adapt11 board that is used in the EE lab in EE 300 and EE 400 to teach
the Motorola hc11 microcontroller.  Its pins can be connected to any other pins in the simulation,
but the behavior of the component is described by an emulator of the hc11.  The core emulator
engine was taken from the open-source Wookie project and integrated into the Pookie system. 
The user can load Motorola byte code files (.s19) into the “memory” of any hc11 in the simulation,
and can simulate the behavior of the circuit while the hc11 is running the program.  This will allow
students to simulate the actual systems that they are building in EE 400. 

Device Design and Implementation for High-Power
Single Wall Carbon Nanotube Operation

Authors:    Steve Lee
                 Jawaad Mahmood

Advisor:    Scott Paulson

While low-bias transport properties are well understood in carbon
nanotubes, high power characteristics of the carbon nanotubes will play a
crucial role as carbon nanotubes become ready for commercial use. The work
presented in this paper will describe the design and implementation for high
power carbon nanotube operation as resistors. Chemical Vapor Deposition will
be used to grow the nanotubes and electron beam lithography will be used to
wire them for electrical measurements. It is expected that 1 mW of power will
be dissipated through a nanotube of 1 mm length and 0.7 nm radius under ultra
high vacuum, and at least 0.5 mW of power will be dissipated in the
presence of an extra heat sink. Through preliminary models, it is known that
there is strong heat coupling (phonon-phonon interactions) between the
nanotube and its environment. Thus, by changing its environment, the heat
dissipation properties can be optimized. The end product will be able to
handle 25 mA at 40 V.

Self Demodulating Nonlinear Ultrasonic Communicator

Authors:    Ryan Hinkle
                 Michael Nigro
                 Asif Rahman

Advisor:     Daniel Lee

The Self-Demodulating Nonlinear Ultrasonic Communicator makes possible the communication of highly
directed audio signals through exploitation of the "beamlike" properties of ultrasonic waves.  The system
preprocesses a standard voice signal before using it to modulate a 40 kHz ultrasonic carrier signal.  This
signal is then transmitted through air at a range of five meters to a speaker array of 37 ultrasonic transducers
which outputs 3.5 Watts, or 120 dB, of sound energy to achieve a 65 dB self-demodulated audible signal
output one meter from the array.  The medium in which self-demodulation occurs can either be the air or
any sound reflective surface, making reflection of the sound "beam" possible.  A beam angle (defined as
the angle from center at which 3 dB attenuation occurs) of 15 degrees has been achieved, which represents
a highly significant improvement over a standard piezoelectric speaker, which has virtually no beam angle.
Signal distortion relative to the original signal, measured by Total Harmonic Distortion, has been reduced
through use of a frequency equalizer along with other, more experimental methods.  A goal of 15 percent
distortion has been set for the product.

Skating Robot Dog

Authors:    Henry Chakardjian
                 Edward Graf
                 Tae Lee

Advisor:    Dr. Afifi

This design project intends to add another dimension to robotic movement by
giving a four-legged robot the ability to inline skate. Because of budget
constraints the robot must be built from the ground up. The anticipated result
is a four-legged robot with inline skating capabilities. The robot will create
it’s own momentum from rest. The focus of this project is the design and
construction of the four legs because the legs accomplish the skating motion.
The implementation of the leg motions will be done in stages. The first stage
is to coordinate motion between the two front legs while there is outside
support for balance. The second stage is to coordinate motion among all four
legs with the support. The final stage is to coordinate motion among all four
legs and to maintain balance without outside support.

Theremin-Based Interface for Computer Input

Authors:    Zachary Davis
                 Harlan Flagg

Advisor:    Dwight Jaggard

At present, there are several devices that can be used for point and click control of a computer; however,
for those with severe disabilities or other physical limitations, none of these devices is a feasible option.
Inspired by the theremin, a musical instrument, a new computer interface device is to be constructed which
uses the variable capacitance induced by the user's body to control the action of the pointer on the computer
screen.  The user will be able to move the pointer and click by hand movements.  This interface does not
require the user to hold any objects or perform any fine movements.

IR Information Center for Handheld Devices

Author:    Steve Battle
               Jacob Rank

Advisor:   Siddharth Deliwala

Current information distribution methods are not keeping up with the increased use of mobile technology. In many cases, bulky and unwieldy paper documents are used where digital files on portable devices would be much better.

The goal of this project is to create an embedded information server that can interface wirelessly with handheld computing devices. Administrators are able to configure the server remotely over the Internet to control the data accessible to end users. These users will then be able to browse the items available for download, select the desired file, and have it transferred automatically to their waiting handheld device.

The administrative functions are achieved through uploads to a FTP server process running on the embedded system. Remote data files will be accessed via a FTP client process and sent to users’ handheld devices over the IrDA wireless transmission protocol. The entire system is implemented on a Rabbit 3000 microcontroller unit with integrated IR transceiver and Ethernet control IC.

Bandwidth Reduction Techniques In a Distributed
Interactive Environment

Author:     Viviana Acosta

Advisor:    Dr. Greenwald

The Internet gaming community continually seeks ways to minimize bandwidth consumption on networked
3D games, or distributed interactive virtual environments.  Fly8, an example of such a networked game, is
written in C using UDP transmission atop Sun/Solaris Unix computers.  Remote users fly planes across a
single host server while aiming to become the sole survivor at the end of the game.  After initial packet
transmission measurements, the control scripts produced packet transmission logs, and as expected, traffic
accumulated as remote users were aggregated. The solution to reducing network traffic, thereby reducing
bandwidth consumption, was to use a prototype script that incorporated position-based dead reckoning.
The thesis of Jiang Ping Shi, a former doctoral student of the project advisor Dr. Greenwald, articulates
position-based dead reckoning, in which graphical trajectory is computed by position estimation.  Finally,
in comparing the control scripts with the prototype scripts, the resulting graphical models, and UPD packet
transmission logs, the prototype scripts indicate significant improvement in minimizing bandwidth consumption,
certifying position-based dead reckoning, an example of multiple types of dead reckoning, as a prime choice
for reducing bandwidth in distributed interactive virtual environments.

FPGA Implementation of Artificial Human Cochlea
Front End FPGA Implementation of an Artificial Human Cochlea

Authors:    Aamer Ghouse
                 Aditya Talwar
                 Henry Tam

Advisor:   Jan Van der Spiegel

Traditional methods of speech recognition have very limited complexity and impose considerable grammar
constraints. Systems in existence today have critical problems understanding different voices and do not have
robust vocabularies. A more robust approach suggested by Ahmed Ali, based on Seneff ’s Generalized
Synchrony Detector (GSD), relies on the acoustic – phonetic characteristics of speech. This paper describes
an attempt, based on the research by Ahmed Ali, to model an artificial cochlea using sixteen digital filters. The
specific part of the cochlea of interest here, the basilar membrane, acts as a collection of bandpass filters that will
be mimicked to develop an artificial digital cochlea. The final product is expected to be cost – efficient and will
consist of sixteen bandpass digital filters implemented on four Xilinx FPGA’s. The Field Programmable Gate
Array’s (FPGA) will be mounted on a Printed Circuit Board (PCB) which will act as the stand-alone module.
The digital cochlea will be used in conjunction with a neural network that will extract features and phenomes (
the simplest unit of speech) from speech signals.

Remote Controlled RoboDog

Authors:    Catherine Gooi
                 Candace Lin

Advisor:    Daniel Lee

Annually, students worldwide participate in a robotic dog (Sony AIBO) soccer competition called Robocup. 
To win this competition it is essential to master the motions of the dog, perfecting its runs, kicks and passes. 
To facilitate the coordination of movements, a graphical user interface is created in Matlab for easy control
of the Sony AIBO robotic dog.  This interface features a click-and-drag figure of the dog and options for
speed control and the recording of a series of user-defined movements.
Communication between the AIBO and the user interface is established through an 802.11b wireless network. 
The AIBO modifies its position according to a matrix of joint angles.  Accordingly, each modification of position
made to the “click and drag” figure corresponds to a new joint angle matrix being sent to the AIBO.  The interface
allows a sequence of positions to be recorded and replayed.  Recording simply involves storing the set of angle
matrices for each new position.  The speed at which the positions are replayed is adjustable using a scroll bar. 
Speed control is implemented by finding intermediate points between consecutive positions, and varying the
number of those intermediate points.  By increasing the number of intermediate points the AIBO slows down
and by decreasing it speeds up.  
Each of these features is implemented by writing Matlab functions.  These functions are called when corresponding
buttons on the graphical user interface are pressed by the user.

Modular Robotics Using the Sony AIBO

Authors:    Gabriel Jinich
                 Eduardo Vieira de Fonseca

Advisor:     Daniel Lee

The Sony AIBO Robot Pet is a modular system.  Legs and head clip on to the body and connect to a central
processor through the proprietary Open-R Bus system.  This approach to design will allow Sony to license
Open-R technology to third parties so that they may build and market new modular elements.  This
project seeks to analyze both the Open-R Bus system and the functionality of a leg element so that a wheel
module may be implemented and controlled by pre-existing software via the Open-R Bus. 

CMOS Camera-On-Chip with Polarization Difference Imaging

Authors:    Kunal Ghosh
                 King Yeung

Advisors:   Jan Van der Spiegel
                 Nader Engheta

This aim of this project is the successful design and implementation of a CMOS Active Pixel Image
Sensor (APS) for Polarization Difference Imaging (PDI). With advances in CMOS technology,
previous work has focused on the design of imaging system applications that exploit the low-cost
and easy System-on-Chip (SoC) integration advantages of CMOS over CCDs (Charge-Coupled
Devices), traditionally the technology-of-choice for digital cameras and imaging systems. The work
presented in this project is novel in that it seeks to enhance the existing CMOS APS image processing
technology with the design and implementation of PDI processing on-chip. PDI is a biologically-inspired
imaging technique that improves the visibility of objects in scattering media by amplifying the signal from
targets whose polarization-difference magnitude is distinct from the background. Experimental results
obtained from a target in a scattering medium demonstrate that a manmade polarization-difference system
can render readily visible surface features invisible to conventional imaging. This work has therefore been
inspired by the quest to integrate a PDI system with a CMOS imager. The CMOS imager for Polarization
Difference Imaging has been designed and implemented in a 5 V 0.5 um double poly, triple metal conventional
CMOS process. A 72 x 72 pixel array is used resulting in a core chip area of 2.4 mm x 2.4 mm. On-chip
synchronization pulses are generated for the horizontal and vertical digital readout circuitry. PDI processing
on-chip is implemented using switched-capacitor circuits that sum, difference and scale the orthogonally
polarized components of light. The CMOS imager for PDI is highly suitable for future intelligent active vision
systems, especially under light scattering conditions.

Active Noise Cancellation From Fixed Point Sources

Authors:    Wuen-E Chang
                 Hoon Kim
                 Benjamin Resnik

Advisor:    Dwight Jaggard

Much sound pollution annoys or disturbs people in everyday life such as overhead airplane rumbles,
traffic, and construction.  While many passive systems reduce these noises, this project aims to reduce
the stress that people endure by destructively eliminating sound via active noise cancellation.  The project
creates an active noise cancellation system that drives an amplified speaker to cancel one- or two-
frequency noise signals in one-dimension.  Active noise cancellation attenuates noise by creating a
duplicate of the original sound wave with a 180° phase shift in each frequency.  The frequencies are
limited from 70Hz to 1,500Hz.  

A feedback system using two microphones will calibrate the system and measure the noise and noise’s
attenuation.  The digital signal processor will use this data to adjust the system's response.  The software
algorithm conducts a fast Fourier transform on the signal to determine frequencies in the noise signal. 
The DSP generates a cancellation frequency at based on the FFT.  Noise at a higher frequency than
the cancellation frequency requires a consistent positive phase shift and the inverse applies.  If the
frequencies are the same, then a phase shift of 180° is applied.  The anticipated result is a 20 dB
attenuation regarding the unwanted noise within a few seconds.

Component Image Transform Analyzer (CITA)

Authors:    Will Luis
                 Victor Wan

Advisors:   Nader Engheta

Human Visual systems lack the ability to detect the polarization of light.  The aim of this project is to develop
a real time imaging system that maps the polarization of light to a form that is suitable for the polarization blind
viewer. 
There are two systems developed in this project to obtain the polarization information.. The first consists of a
Ferroelectric Liquid Crystal (FLC) synchronized with the frame rate of a CCD video camera.  The FLC rotates
incoming by 90 degrees in response to a voltage signal.  The FLC used in tandem with a polarizer enables the
capture of the strength of the polarization field along two orthogonal axis – one polarization state per frame of
the video stream. The second imaging system is consists of four cameras, with each camera detecting a different
axis of polarization. 
The vision processing system employed consists of a frame grabber and PC as a host system.  Upon obtaining
the images, the vision processing system subtracts orthogonal axes of polarization and outputs the result onto a
computer monitor.

Application of Contour Mode Disk Resonators in Oscillator Circuits

Authors:    Patrick Baude
                 Matthew Cohen

Advisor:    Stephane Evoy

One of the main obstacles to further the integration of wireless communication systems is the reliance
on off-chip filtering elements.  Bandpass filters that are currently fabricated on integrated circuits have
very low Q ratings. One possible solution is to use microelectromechanical systems (MEMS) as
replacements for the existing integrated bandpass components.  The use of MEM structures has been
researched and a major deterioration in resonator Q arises from energy losses at the clamping points. 
A contour-mode disk resonator minimizes clamping losses by designing the structure to have its clamping
point at the center, a vibrational node.  Modeling is done to predict the motion of such disks.  Disks are
fabricated with diameters ranging from 2 to 50 microns, corresponding to resonant frequencies in the
hundreds of MHz range.  An oscillator is designed  and simulated using these MEMS as the frequency
selective element.

Interfacing Circuitry for Nanostructure-based Sensors

Authors:    Vishnupriya Dasgupta
                 Saramma Varghese

Advisor:    Stephane Evoy

As new technologies become more microscopic, nanotechnology seems a possibility in mainstream
applications now more than ever.  There is a severe lack of circuitry to interact with many nanostructures,
however.  Poly-silicon is a material that is currently being used in analog electronics technology, and the
material has many properties which make it ideal for experimentation.  For these reasons, a poly-silicon
device was fabricated to simulate a nanoscale semiconductor sensor.  A power efficient circuit was then
developed to maintain constant power across the nanostructure.  To this end, the circuit maintained constant
power within a one percent margin of error in the presence of the device.  The device under test has the
property of changing resistance as temperature increases or decreases.  As the temperature is dropped,
the resistance shows a corresponding drop but the power across the device remains constant. 

Cawthorne-Cooksey Vestibular Rehabilitation Controller

Authors:    Nirav Batavia
                 Stephen Kossuth
                 David Lyman

Advisor:    David Solomon
                 Siddharth Deliwala

The Cawthorne-Cooksey Vestibular Rehabilitation Controller (CCVRC) creates an electronic take-
home rehabilitation device for patients with vestibular illnesses in order to improve on current primitive
rehabilitation methods. The system being developed is based on the popular Cooksey-Hawthorne
rehabilitation method for vestibular therapy. The patient wears a laser apparatus on his head and uses
a wall mounted panel attached to a microcontroller which is in turn connected to a Palm Pilot to create
a completely transportable system.  The patient will be instructed to move his head in order to align
the laser beam in the direction of the illuminated light on the panel. The solar cells on the panel sense
the light and the microcontroller records a variety of data which is then fed back to the Palm Pilot. 
This feedback can be collected over time and can therefore provide useful feedback to medical
personnel for the purpose of making more informed decisions regarding the rehabilitation program. 
Not only does the CCVRC help therapists, it is anticipated that it can also help improve recovery
time and reduce doctor office visits.  

GSM Signal Generator

Authors:    Jeremy Goldman
                 Andrew Gotwols
                 Sam Miller

Advisor:    Bill Townsend

The Global System for Mobile communications (GSM) Signal Generator will support Bit Error Rate testing of the TruePosition® Location Measurement Unit. The signal generator generates baseband digital data on a digital signal processor (DSP) card, connected to a PC via a parallel port, in the form of GSM bursts that will make up four of the GSM channels used globally. The digital data is sent to a digital-to-analog converter, a daughter card for the DSP card, and converted to the analog baseband frequency. The baseband analog signal can then be routed to a radio frequency (RF) card where it will be upconverted to the proper frequencies defined by the GSM standards. Both the DSP card and the RF card are controlled from the PC via a graphical user interface

Computer Monitoring System (Electrical Powered Saturn)

Authors:    William Anderson
                 Daniel Lin

Advisors:   Simon Hauger

For several years now, the students of the automotive program of West Philadelphia High School have
built cars to be entered in various competitions across the country.  In these competitions, the car that
make the best use of its power source wins.  This is judged by the vehicle fuel efficiency, the ratio of
the speed of the vehicle to the energy used in the battery.  Unfortunately, an electric powered Saturn
built at the school does not have a display system yet.  The overall goal of this project is to create a
data collection and display system that allows the driver of the electric powered Saturn to operate it
at its most efficient point.  This overall goal will be accomplished by doing the following by scaling down
and feeding the DAQ card the following voltage signals:
a. Motor RPM
b. Battery current
c. Speed
d. Incline
From these four voltage signals, all of the other necessary data points will be calculated.

Neuromorphic Processing for Invariant Feature Extraction

Authors:    Toufique Harun
                 Yu Ming Hu
                 Jason Naftal

Advisor:    Nabil Farhat
 
 It has been hypothesized that spiking neural networks can be used to extract invariant features
from images.  This project's goals are two fold.  First we will simulate a system to extract invariant
features from simple shapes in order to recognize them by operating on the characteristic of the
inter-spike interval histogram of the network.  In order to apply this in real world applications, a
faster hardware version needs to be built.  A fully featured system is beyond the scope of this
project, but we will build a neuron with threshold modulation and an inter-spike interval analyzer.

Neural Networks for Radar

Authors:    Vito Sabella

Advisor:

Radar Target Identification has been a difficult task that has plagued the Military for years.  Modern,
high-tech ISAR radar can give you long-range, distance independent snapshots of aircraft with incredible
precision, but attempting to identify targets based on that data is still a difficult art.
The Radar Target Identification system builds upon the concept of Neural Networks to build a system
that can accurately recognize a target from a handful of its Radar signatures.  Utilizing a supervized version
of the Adaptive Resonance Theory (ART) algorithim, the system can categorize and identify aircraft targets.

Electronic Umpiring System

Authors:    Matt Dilmaghani
                 Kevin Lanik
                 Ari Peskoe

Advisor:

The Electronic Umpiring System will automatically determine whether a pitched baseball is deemed
a ball or strike according to the official rules of baseball.  This objective will be accomplished using
infrared LED’s that shine upwards from home plate in a plane perpendicular to the path of the ball. 
Infrared sensors (phototransistors) will be placed at the bottom of angled channels in the front plane
of home plate.  Sensors will detect infrared light reflected from the ball as it passes over the plate.
Software embedded in a microcontroller will determine the X-Y position of the ball and whether it
is a ball or a strike.

Efficient Sampling of RGB Images

Authors:    Shelley Zhang
Advisor: Dr. G. Buchsbaum

Multispectral images, of which regular RGB color images are an example, are signals available in different spectral bands. Multispectral images have gained considerable usage in remote sensing from space as well as numerous other applications in geosciences and medical imaging. While gross spatial features like edges in the multispectral image bands are similar, details in spatial low frequency features vary among the spectral bands.

A MATLAB model is developed and applied to a number of images. The model is based on Nyquist sampling requirements and Fourier similarities between the spectral bands to minimize the number of required samples. Two methods are used to reduce samplings. The first uses the spectrum of (R+G+B)/3 in high frequecies to replace that of the individual RGB spectra. The second uses transformation into (R+B+G)/3, R-aG, R+bG-cB and bandlimits the the latter two spectra. A reduction of 33% of total samples could be obtained with a signal to noise power ratio of approximately 40dB. The tradeoff curve of error as function of number of saved samples is provided

Rear Infra-Red Sensors for the Smartchair

Authors:    Sham Shah
                 Chung-Tah Tsao

Advisor: Vijay Kumar

A distance sensing mechanism was built on the autonomous wheelchair, Smartchair, to allow the Smartchair to maneuver around obstacles. The system consists of eight infrared sensors that operate autonomously and are strategically placed on the rear end of the Smartchair so as to collect maximum information about the proximity areas of the Smartchair. A microcontroller is used to input timing waveforms into the sensors and control each of the eight sensors. It receives information from the sensors in form of high or low bits, which it then decodes into a decimal number. The decimal number is then relayed to the microprocessor that uses a comprehensive mathematical formula to convert the decimal number into the appropriate distance. The distance is then reported to the user in real-time using a serial port, and using the information from the sensors, the Smartchair will find an alternate path to reach its destination. The sensor system was integrated onto the current Smartchair user interface and can be used in conjunction with the autonomous navigation function or the manual navigation function.

Interactive E Commerce and Communication Software