Summer 1999

Project Archive

David Auerbach (Physics) - Swarthmore


Advisors: Ahmed M. Abdelatty Ali, Dr. Jan Van der Spiegel, Dr. Paul Mueller

ABSTRACT: This paper describes the use of recurrent neural networks for phoneme recognition. Spectral, Bark scaled, and cepstral representations for input to the networks are discussed, and an additional input based on algorithmically defined features is described that can also be used as input for phoneme recognition. Neural networks with recurrent hidden layers of various sizes are trained to determine, using the various input representations, whether a stop consonant is voiced or unvoiced, and whether the stop consonant is labial, alveolar, or palatal. For voicing detection the peak accuracy was 75% of the phonemes not used to train the network identified correctly, and for placement of articulation, the peak accuracy was 78.5% of the testing set identified correctly. Using the algorithmically defined features and a three-layer feedforward network, an average accuracy of 80% for voicing and 78% for placement of articulation. Implications of these results and further research needed are discussed. Go to Full Paper.

Darnel Degand (MEAM) - University of Pennsylvania


Advisors: Dr. Jim Ostrowski, Robert Breslawski
Collaborators: Kapil Kedia, Vincent Marshall

This project concerns the development of modular components that can be used to quickly put together robotic locomotion devices. This system will allow a single robot to complete multiple tasks or navigate through various terrain simply by changing its configuration. The modules will include modular connectors that can be easily detached and reattached from the mainframe, allowing one module to be replaced with another. Several applications include walking, climbing, and possibly swimming. This will lead to decreased costs because a single modular robot will be used in place of many specific robots. Work has been done on developing a leg module that includes three degrees of freedom; hip rotation, raising the leg, and bending the knee driven by three motors. CAD/CAM tools are presently being utilized to generate the basic modules in a manner that allows for easy production and assembly of multiple units. The modular legs can be plugged onto a main body to rapidly configure a robot with four, six, or more. Software modules that can be tailored to the given modular configuration will be developed. Go to Full Paper.

Hector E. Dimas (Electrical Engineering) - University of Pennsylvania


Advisors: Dr. Dwight L. Jaggard, Aaron D. Jaggard, *Omar Manuar

This research involves the overlaps of antenna theory, fractal geometry, and numerical calculations. Its goal is to investigate how random or periodic antenna array geometry can be improved through the use of fractals. Using a program we developed in Matlab, we plotted the radiation of linear/planar antenna arrays. We were able to correctly analyze radiation of three two-dimensional arrays. Another program whose purpose was to generate fractals was used to allow us to characterize the radiation properties of periodic and random arrays. Go to Full Paper.

Ian J. Gelfand (MSE) - University of Pennsylvania

Flat Contacts for Carbon Nanotube Circuits

Advisors – Prof. Alan T. “Charlie” Johnson, Dr. J. Hone

ABSTRACT: We attempted to construct flat contacts for carbon nanotube circuits using a mechanical polishing process. These contacts were built on an oxidized silicon wafer by etching trenches in the oxide surface. These trenches were then filled with Au and polished down to nanometer flatness. It is hoped that these contacts will allow us to assemble molecular circuits in a controlled manner. Go to Full Paper.

Jason Gillman (EE) - University of Pennsylvania

Photolithographic Processes for Creating a Proximity Sensor in Low Temperature Co-Fired Ceramic Tapes

Advisors: Dr. J.J. Santiago-Aviles, P. Espinoza

Abstract: This paper describes two photolithographic processes for patterning Low Temperature Co-Fired Ceramic (LTCC) tapes. The processes involve the use of two types of LTCC tape. In one case photoformable tape is used, and in the other, DuPont 951Ò tape coupled with DuPont RistonÒ 9015, a dry photoresist, is used. Although a full 8-layer sensor was not completed using either method, single spirals were fabricated and compared to past sensors fabricated using other methods. The parasitic resistance of a single spiral fabricated by the photolithographic methods was 20% of that found in spirals fabricated using screen-printing. The process steps used in creating the spirals as well as important parameters are discussed, and suggestions for improvements in the process are made. Go to Full Paper.

Jolymar Gonzalez-Esteves (Mechanical Engineering) - University of Puerto Rico, Mayaguez


Advisors: Dr. Jorge Santiago-Avilés and *Patricio Espinoza

Low Temperature Co-fired Ceramic (LTCC) and thick film technology, with their mechanical, electrical, and thermal properties make them an appropriate choice for the development of a pressure transducer. This research designs a pressure transducer using LTCC and thick film technology. We show the relation between the size of the diaphragm and resistance to pressure. We determine the best position of the piezo-resistors, which are accommodated forming a Wheatstone bridge. Two of the piezo-resistors measure the deflection on the tangential axis, and the other two measure the deflection on the radial axis. By using the Wheatstone bridge we can obtain more accuracy in the output of the transducer. Go to Full Paper.

Kapil Kedia (MEAM) and Vincent Marshall (EE) - University of Pennsylvania


Advisor: Dr. James Ostrowski (Mech. Eng.)
Collaborators: Darnel Degand and Robert Breslawski

This paper describes the design of an architecture to construct and control a modular robot. This architecture is to serve as a logical and electrical framework for a small mobile robot that has logical and functional separation between its modules. This architecture has a master/slave scheme, where a master “brain” module controls multiple function modules. The communication between the master and the slaves takes place over a serial bus on which the master initiates all communication in order to avoid bus access problems. The architecture was tested using BASIC Stamp II and PIC microprocessors. Testing with both actuator and sensor modules was successful when using this hardware. Go to Full Paper.

Patrick Lu (Electrical Engineering) - Princeton University

IdentifyingKey Phoneme Features in Spectrograms

Advisors: Ahmed M. Abdelatty Ali, Dr. Jan Van der Spiegel, Dr. Paul Mueller

ABSTRACT: Spectrograms carry all necessary information for reliable human and computer perception of speech. This paper discusses the importance of spectrogram features used by a recognition algorithm developed by Ali et al. as they relate to human perception. Features, including MNSS, burst frequency, formant transitions, voicing onset time, and voicing/unvoicing information are defined and their importance to computer stop consonant recognition described. Confirming many previous findings, burst frequency and formant transitions were found to be most important in the perception of speech synthesized from spectrograms while other features played a secondary role. Software tools developed that should facilitate other similar investigations are described. Go to Full Paper.

Catherine Reynoso (Computer Science) - Hampton University


Advisors: Dr. Jay Zemel and Carlos Lopez

Our project is part of an effort to network an array of sensors to monitor the flow distribution in a duct. It involves the fusion of several powerful software-based information extraction instruments (nodes) with a set of pyroelectric sensors. The pyroelectric anemometer sensors function by responding to local changes in the flow of gas. The program, which will be later installed into the microcontroller to carry out the computations for the system is referred to as the lock-in amplifier. Our project specifically focuses on the development of the lock-in amplifier for one of these independent nodes.

Philip Schwartz (Physics and Chemical Engineering) - University of Pennsylvania


Advisors: Dr. P. Bloomfield, Drexel University, Dr. S. Frankel, University of Pennsylvania

A system allowing underwater robots to communicate was developed using ultrasonic transducers sending out a signature pulse of 5 bits to enable recognition of each individual robot. Steel tubes were fabricated to house ultrasonic transmitting and receiving transducers. The design affords electrical shielding of the electrical signals and the amplification electronics. The electronics were designed to be able to send in an energizing signal and send out the amplified received signal using only one line. Power for the amplifiers and the electrical pulses that activate the transducers to emit acoustic pulses were sent in over a BNC cable, and acoustic pulses received by the transducers and transformed into electrical signals were amplified and transmitted to recording electronics. The transducers were made of piezoelectric polyvinylidene fluoride (PVDF) and were backed with an inert PVDF cylinder. The outside electrode of the transducer was grounded to the steel tube and the inside electrode was connected to the wire using conductive epoxy. Signals were transmitted to the transducers using a C program and an arbitrary function generator. Signals were received from the transducers using a 20 MHz National Instruments NI5102 digital oscilloscope card for PCI bus and Virtual Bench 2.1.1 software. The hardware and software were tested to ensure capability of interpreting the received signals. The signals output by the function generator were created using Arbitrary Waveform Creation software. The electronics were designed and diagrammed in an iterative procedure using P-Spice.