The MERIT Summer Research Program
 
 
 
 
 
 
 
 
 
  The A. James Clark School of Engineering
  University of Maryland Home

 

BIEN Projects: Summer 2011

The BIEN (Biosystems Internships for Engineers) projects that will be offered during the Summer 2011 session are listed below. Project names are linked to their respective descriptions. Faculty members and project directors are linked to their home pages where available.


1. Micro-Robot Control and Coordination
Prof. Pamela Abshire

Most previous work in the field of robotic controls has assumed that the robots possess advanced sensing and movement capabilities. However, this will not be true in the case of extremely resource-constrained robots, for example sub-cm3 robots or "antbots". The focus of this project is on small robots with limited sensing and movement capabilities. These robots can be used for a number of applications including surveillance and infrastructure inspection and monitoring. Resources available to the micro-robots will be extremely limited, including power, sensors, communication bandwidth, and computation. Thus, in order to realize controllable micro-robots it will be necessary to implement custom designs for many of the system components including communication, computation, actuator drivers, and power transfer. Several promising methods to address these problems include asynchronous systems, "smart" sensors that detect when there is relevant data to communicate, and de-centralized controllers that coordinate swarm activities based on locally available, noisy information. BIEN interns will engage in hands-on construction and testing of prototypes and algorithms as well as design of system components such as distance sensors and simplified system architectures. This inter-disciplinary effort will be carried out in coordination with collaborators in Mechanical Engineering.

2. Nose on a Chip
Prof. Pamela Abshire

Modern engineering has produced exquisite sensors for many kinds of measurements, but olfactory sensors remain somewhat rudimentary. Interfacing electronics to biological systems leads to the possibility of creating devices that directly monitor the responses of living biological cells. Potential applications include cell-based sensing, medical diagnosis, drug screening, pathogen detection, and scientific research into cellular mechanisms. These advances provide a technological basis for the Nose-on-a-Chip, a hybrid bio-electronic olfactory sensor that monitors responses from biological cells known as Olfactory Sensory Neurons (OSNs). Major technical challenges in order to realize a functional Nose-on-a-Chip include: developing a microfluidic system for culturing the OSNs, monitoring the weak extracellular electrical responses of OSNs in response to odorants, and interpreting the responses in order to detect and classify the odors.

BIEN participants will develop passive and active microelectrode arrays and array readout techniques for detecting weak electrical signals from electrically active cells , conduct odorant screening experiments on OSNs, design algorithms that map the electrical signals from cells into an inferred odor, characterize new component technologies such as packaging methods and data acquisition software, and develop new functional capabilities such as cell steering or screening. This inter-disciplinary effort will be carried out in coordination with collaborators in Biology and Mechanical Engineering.

3. Computer Vision Methods for Understanding the Dances of Honeybees
Prof. Rama Chellappa

Prof. Chellappa’s cutting-edge computer vision research provides broadly applicable theories and algorithmic infrastructure to understand the environment from visual cues. One unique biological application developed by Prof. Chellappa’s group is to model and recognize the dances performed by honeybees, as these dances encode information regarding the quality and quantity of nectar, distance and bearing from the hive. From videos of bees in a hive executing the waggle dance and the round dance, the simultaneous automatic tracking and behavior analysis approach estimates the position, orientation and the current behavior of the honeybee, using a parametric shape model and a three-tier hierarchical motion model combining dynamics and Markovian properties with a tracker based on a particle filter. This method was extended for tracking multiple bees using multiple cameras in natural flight [34], which requires addressing the challenges in establishing correspondence between objects. BIEN participants will develop real-time implementations of these two tracking algorithms. Using the 3D flight paths of bees and the factorization algorithm, they will investigate the behavioral patterns observed in actual flight paths. They will also evaluate the effectiveness of the tracking methods for assessing the health of bee hives, so that alerts about potential disappearance of colonies can be generated. Finally, they will extend the behavior model by incorporating interactions between the dancing bee and the "dance attendees" who touch the dancing bee with their antennae. Prof. M. Srinivasan of the University of Queensland will serve co-advisor for this

4. Studies of correlation functions of laser beams that have passed through turbulence
Prof. Chris Davis

In this project the signals from several photodetectors that are illuminated by an expanded laser beam that has passed along a long or short path through clear air turbulence will be analyzed to determine multipoint correlation functions, scintillation effects, and polarization crosstalk. These experiments are relevant to free space laser communication systems, and the determination of local turbulence conditions in the atmosphere. Students working on the project will learn about real-time data acquisition and analysis, how to operate sensitive optical experiments, and will gain an understanding of wave propagation through random media.

5. Extraction of articulatory information from natural speech for speech and speaker recognition
Prof. Carol Espy-Wilson

This research proposes to design and implement an articulatory trajectory (in the form of vocal tract constriction variables) estimator from natural speech. Different model driven strategies (such as neural networks, particle filtering, etc) will be explored to obtain a suitable model that performs the necessary nonlinear transformation of the acoustic observations to get to the articulatory trajectory space. Exploration would also include appropriate acoustic parameterization of speech and their orthogonalization and dimensionality reduction.

To carry out this exploration we need a large vocabulary database that contains acoustic samples and their corresponding articulatory data. Unfortunately no such large vocabulary database exists at present. Previously we have proposed [1] an unsupervised technique to obtain articulatory annotation for medium sized databases. This project would employ that technique; optimize its performance and running time by incorporating efficient and robust algorithms and by using better programming paradigms.

[1] H. Nam, V. Mitra, M. Tiede, E. Saltzman, L. Goldstein, C. Espy-Wilson, M. Hasegawa-Johnson, A procedure for estimating gestural scores from natural speech, Proc. of Interspeech, pp. 30-33, Japan, 2010.

6. Integrated Bio-Micro-Systems for Detecting and Studying Bacterial Films
Prof. Reza Ghodssi

Bacterial infections are the leading cause of disease worldwide, and the development of antimicrobial drugs has been one of the highest priorities of biomedical research. It has been shown that certain types of bacteria communicate with each other through small signaling molecules. This capability, called quorum sensing, allows bacteria to perform population-coordinated actions and to overcome the host's immune system. Bacteria aggregate and form a pathogenic matrix known as a biofilm, which is impenetrable to conventional antibiotics. A promising new approach for combating bacteria is to develop drugs that disable their communication, making them less pathogenic and more susceptible to antibiotics. Researchers at the MEMS Sensors and Actuators Lab (MSAL) in the Electrical and Computer Engineering (ECE) department are developing microfluidic devices and implantable MEMS for studying in vitro and in vivo bacterial quorum sensing. Growth of clinically relevant bacterial biofilms is optically monitored in microfluidic devices, and biofilm growth is correlated with environmental factors. The use of microfluidic technology will ultimately allow large numbers of potential drugs to be tested rapidly with small sample volumes. In parallel, we are developing implantable MEMS sensors to detect in vivo bacterial infections by using surface acoustic waves (SAW). The implantable device will be able to monitor bacterial quorum sensing continuously at sites susceptible to bacterial infection, such as artificial organ or joint implants. The sensor will thus provide an early alert system and prevent severe infection and invasive surgery.

This project is a collaborative effort involving groups in the Bioengineering, Electrical and Computer Engineering, and Materials Science and Engineering departments as well as the Institute for Systems Research at the University of Maryland. It offers interdisciplinary learning opportunities to MERIT students. The student will use advanced bioengineering equipment, a microfluidic test station and an acoustic wave sensor measurement setup. The goal of the project will be to perform measurements of biofilm growth using standard laboratory tools. The results of this study will be used to validate the experiments and to further refine the micro/nano devices.

7. Neuromorphic VLSI and Algorithms Inspired by Bat Echolocation
Prof. Timothy Horiuchi

Prof. Horiuchi’s research bridges microelectronics and neurophysiology via development of VLSI models of neural systems at the spiking neuron level and their demonstration on robotic platforms. His group has been developing bat robots (biologically-realistic sonar sensors), analog VLSI implementations of neural circuits, and software-based approaches to test models of behavior in closed-loop robot environments. Undergraduates have long played an important role in linking the lab’s chip design and behavioral simulations research to demonstrations on robots and the exploration of system-level behavior. Projects in the Horiuchi laboratory focus on two main areas: “neuromorphic” VLSI design of circuits that implement neural algorithms in custom CMOS, and the development of algorithms and sonar hardware that implement models of bat-inspired echolocation behavior. Student teams have worked on projects that have spanned across multiple labs. Future projects include: (1) Modeling neurons of the superior colliculus (with Prof. Moss as co-advisor); (2) Developing new algorithms of bat flight behavior and for adaptive echo vocalization behavior (with Prof. Moss); (3) Adaptive synaptic models for implementing working-memory models (with Dr. Katrina MacLeod of Biology at UMCP); (4) Implementing spike-based winner-take-all models of the attentional spotlight; (5) Designing new circuits for synaptic integration and conductance-based models of the neuron. Most projects involve combinations of theory with algorithm development and testing on live sensory signals.

8. Synthetic Flocking and Bioinspiration
Prof. P. S. Krishnaprasad

There is a tremendous level of interest and excitement in the investigation of collective behavior in nature, found across many length scales, as in large graceful, dynamic flocks of European starlings evading peregrine falcons, small highly synchronized groups of marine mammals such as spinner dolphins foraging for food, mobile swarms of midges, and further down in length scale, in rafts of bacterial assemblies. Gathering reliable 3D data of such collective behavior to obtain insights into the underlying structure and quantitative characteristics is a challenge that is slowly being addressed in a variety of university centers. Building and analyzing mathematical models based on such understanding would be a foundation for designing control laws to synthesize flocks, and implementing them in software applications to achieve versatile collective behavior in robotics. In the summer of 2010, the Intelligent Servosystems Laboratory will organize a small team to explore such modeling and analysis studies of synthetic flocking, inspired by biological data, and using novel mathematical formulations. Applications to robotics will also be possible using a small number of physical platforms interacting with a collective of virtual platforms in a distributed processing environment. Students with enthusiasm for mathematics, biology, physics and software are likely to find the project rewarding. The project will be supervised by Professor P. S. Krishnaprasad and his Ph.D. students.

9. Signal Processing in the Human Brain
Prof. Jonathan Z. Simon

Brain activity is observable via a variety of tools. Most fall into the broad category of brain-imaging (e.g. fMRI) and are too slow to measure real-time neural computations. An alternative is magnetoencephalo-graphy (MEG), which is sensitive to neural processes changing as fast as every millisecond. MEG is related to the more commonly used clinical tool electroencephalography (EEG), but it has key advantages due to its use of neural magnetic fields: the brain is magnetically, but not electrically, transparent. Since the entire brain is active simultaneously, however, the neurally generated magnetic fields become a mix of signals generated in many cortical areas, and they require a variety of signal processing techniques to determine the underlying neural processes performed in individual areas. Summer students will be given the opportunity to apply both traditional and cutting-edge signal processing techniques to neural data acquired via MEG, as well as being able to take part in conducting the experiments in which the MEG data is itself acquired. The goal in conducting these auditory experiments, and their data analysis, is to characterize, understand, and quantify the neural computations performed by the brain.

10. Security and Privacy Protection of Biometric Data (tentative title)
Prof. Min Wu

Imagine that you no longer need to bring any identification cards to prove who you are, and no longer need to remember different passwords to perform transactions or log-in to your personal accounts online. Biometric technology has become increasingly popular.

By associating people with their unique and hard-to-reproduce physiological or behavioral characteristics (such as fingerprint, iris, face, and voice), biometrics enables person identity recognition and authentication based on who you really are rather than what you have (e.g. ID cards) and what you know (e.g. passwords). In order for biometrics to be broadly adopted, one of the critical issues that must be addressed is security and privacy protection of biometric data.

Through this REU project, students will be involved in the research on security and privacy protection of biometric data. Some of the work includes robust, distinctive, and efficient feature extraction from biometric data; secure storage of biometric data over distributed network clouds and potentially adversarial environments; and secure and privacypreserving matching of biometric data. During this interdisciplinary study, students will be exposed to and integrate state-of-the-art techniques from relevant research areas of image processing, multimedia retrieval, information coding, and applied cryptography. Students will also learn to identify new technical/research issues related to secure biometrics and develop algorithms and software prototypes to showcase their work.

11. Crittercam Research and Development
Prof. Nuno Martins

Crittercam is a remote imaging research tool, developed by Greg Marshall and his team at the National Geographic Society, that enables scientists to study the behavior and ecology of wild, free-ranging animals in their natural habitats. This tool is currently being used for scientific research in a wide variety of locations and habitats around the world, both marine and terrestrial, and has led to a number of extraordinary new scientific insights and discoveries. As with any scientific instrument, however, improvements in the underlying technology can lead to better capabilities for observation as well as deployment in a greater number of habitats and conditions. The most significant technological constraints in the design of Crittercams are size and power, since they respectively determine which creatures the Crittercams can be deployed on and how long the recordings will last. This project represents a collaboration between the Remote Imaging Group at the National Geographic Society in Washington, D.C., and the Department of Electrical and Computer Engineering at the University of Maryland, College Park. BIEN interns will work to design, develop, and build an improved Crittercam prototype, focusing on one of several technical issues including increased sampling rate, incorporation of different kinds of sensors such as velocity sensors or audio sensors, or the mechanism that releases the Crittercam from its temporary host. This inter-disciplinary project will be carried out on campus at the University of Maryland as well as at the National Geographic Society, a short subway ride away in Washington, D.C.