I began working with the Silva Research Group in the fall of 2006 to build an optimized multi-channel imaging system for capturing signal information across the neural retina. I defended my thesis in September, 2007.

Title

Design of a Coordinated System for Real-time 3-D Image Construction via Confocal Microscopy

Abstract

In order to facilitate the imaging of signaling networks and to fully utilize the capabilities of confocal microscopy, it is desirable to build a system to coordinate several hardware components of an imaging platform. This document describes the integration of hardware with a confocal microscope and the design of a custom control application to optimize image sampling and 3-D image reconstruction of signaling networks. By implementing advanced timing controls and minimizing switching delays, speed and accuracy are optimized, and the sampling space and time resolution are maximized. The control application reliably demonstrated successful 3-D imaging at the maximum frame rate supported by the camera.

Links

Master's Thesis (PDF)

Defense Presentation (PDF)

Movie: FRET Imaging (AVI)

Movie: Fluor-4 Imaging (AVI)

Neurophysiology Review

This review of the neurophysiology of the interception process was a project assigned during the spring semester, 2007.

Abstract

From hunting prey to catching a baseball, the process by which primates intercept moving targets is a fundamental, if complex, component of neural processing. Over the past 30 years, understanding of the interception process, in primates in general and humans in particular, has advanced significantly. Based on observations of interception behavior and recordings of the corresponding neural activity, several theories describing the involved processes have been generated and tested. This document presents a high-level summary of the physiology of the most relevant neural components of the interception process, an outline of the prevailing models used to describe this process, and a short discussion of recent experimental results used to confirm or reject those models.

Introduction

In order to coordinate the bodily movement involved in everyday tasks such as eating, avoiding or catching a moving object, or even moving through a static environment, prediction of motion and structure based on extrapolation from visual clues is necessary. To understand the processes involved, several models of primate interception have been fit to observed behavior, then experimental data collected mostly from single-neuron recording devices during the interception process were used to confirm or reject model hypothesis.

Links

Review Paper (PDF)

Neurophysiology Lecture

In the spring quarter of 2007, I gave a one-hour presentation comparing two modeling methods of neural signaling. Specifically, this lecture presents the Integrate-and-Fire Model in relation to the traditional Hodkin-Huxley Model.

Links

Integrate and Fire Model (PDF)

Biofuels from Cellulosic Feedstocks

The culmination of a feasibility study of the engineering process behind cellulosic feedstock-based ethanol is a 30-minute presentation describing the benefits and drawbacks of thus energy source. Ultimately, this method is shown to be ill-advised, as the process is currently energy-negative. However, an economic survey is presented which shows the energy threshold at which this technology will be valuable.

Links

Biofuels from Cellulosic Feedstocks (PDF)

Hair Cell Mechanotransduction

A thirty-minute presentation about the mechanical transducive properties of the human inner ear cells.

Links

Hair Cell Mechanotransduction (PDF)

Bioelectronics CGE for Biowarfare Detection

A thirty-minute presentation reviewing existing nanotechnology and proposing new designs for the detection of pathogens and toxins.

Links

Bioelectronics CGE for Biowarfare Detection (PDF)

Nanowire versus Photonoic Biosensing

A group of researchers reports on new nanowire technology as it relates to nano-scale biosensing.

Links

Nanowires vs Photonic Biosensors (PDF)

Webcam-Based User Interface

As the personal computing industry pursues more user-friendly, inexpensive user interfaces, the concept of a touch-less interface is worthy of inspection. This report documents the implementation of a webcam-based user interface using an average personal computer and an inexpensive, low-quality web camera. Using the video feed from the webcam, this product would track the user�s eyes and find their focus on the screen, controlling a cursor. This product is useful for sterile environments such as hospitals as well as public terminals where sanitation is important. The disabled can make use of such a device, as well as anyone who would like to operate their television from across the room without having to find the remote.

Project Goals

Although this project pursues the vision of an entire software system of image processing, cursor control, graphical user interface (GUI), and a calibration routine, this is an unreasonable expectation during one semester of part-time work. Therefore, it will be the purpose of this project to demonstrate a proof-of-concept, documenting the core image processing and neural network methods that will interpret useful user input through an inexpensive web-camera in real time.

Links

Webcam-Based User Interface Project Proposal (PDF)

Webcam-Based User Interface Report (PDF)

Webcam-Based User Interface Presentation (PDF)

Movie: Demonstration of User Interface (AVI)

Image Deblur Algorithm

This report discusses a research project that investigated the use of Fourier optics to correct blurring of images caused by motion. More specifically, linear motion blur of images such as in photographing a speeding car or a shooting star has been considered. We have considered three different cases of linear motion blur and investigated ways of applying inverse filtering techniques to correct the blur.

Links

Image Deblur Project Proposal (PDF)

Deblurring Images Using Cepstrum Analysis (PDF)

Neural Network Jazz Classifier

In many instrumental musical genres, especially those featuring improvisation, a natural way in which listeners divide up a song into logical subsections is to segment the song according to which instrument is soloing during a given section. Improvisational Jazz is a prime example of such a genre, where performances usually consist of a sequence of individual solos. A system which could analyze an audio recording, locate solo sections and identify the soloing instrument could allow users of digital audio playback devices to easily navigate their way to sections of interest in a song. Such a system could also be used to do content based searches of online music, based on the presence or absence of solos by particular instruments. In this paper we present a method that performs the fundamental classification task of such a system. What follows is a description a method for analyzing a short clip of music, and identifying the dominant, i.e. soloing instrument using a combination of wavelet transform analysis for preprocessing and a neural network for classification.

Links

Musical Instrument Identification Using Wavelets and Neural Networks (PDF)

Jazz Classifier Presentation (PDF)

Electrooculogram (EOG)

The electrooculogram measures eye position using a steady electric dipole that is created by the corneal-retinal potential. The EOG is capable of measuring both horizontal and vertical eye movements. The horizontal measurement can be made by placing surface electrodes on the temple to the right and left of the eye and on the forehead. The relationship between the horizontal angle of gaze and the EOG voltage should be approximately linear for +/- 30 degrees of arc. The EOG output typically ranges from 50-3500 uV and the signal frequency from dc to 50 Hz. The electrooculogram is frequently the method of choice for recording eye movements in sleep and dream research, in recording eye movements from infants and children, and in evaluating reading ability and visual fatigue. It is also useful for the detection of RPE diseases.

Links

Electrooculogram Lab Report (PDF)

Galvonic Skin Response (GSR)

The galvanic skin response measures changes in the electrical resistance of the skin caused by an individual�s psychological state. These changes are specifically caused by the sweat glands and ducts, which create a potential difference between the lumen of the sweat duct and the dermis and subcutaneous layers of the skin. As a person sweats, the resistance of the skin decreases, causing the potential difference to decrease. The GSR measurement varies from 1-500k Ohms and over the frequencies .01-1 Hz. The main use of the galvanic skin response has been as a lie detector test.

Links

Galvonic Skin Response Lab Report (PDF)

Electroencephalogram (EEG)

The electroencephalogram is used to record the electrical activity within the brain. The intensity and patterns of this activity are called brain waves. These waves vary greatly between when a person is awake and asleep. General patterns of these waves that are observed in normal people can be classified into four wave groups, alpha, beta, theta, and delta. Other abnormal patterns can help detect seizure disorders such as epilepsy. EEG measurements are also used for the legal definition of brain death.

Links

Electroencephalogram Lab Report (PDF)

Electrocardiogram (EKG)

The 12 Lead EKG measures biopotentials generated by the heart. It can diagnose quite a few problems with the heart including but not limited to right bundle branch block, left bundle branch block, hyperkalemia, myocardial ischemia, and myocardial infarction. Heart problems are the cause of a large portion of deaths in the United States and the EKG is a vital tool for diagnosing heart problems.

Links

Electrocardiogram Lab Report (PDF)

Electromyogram (EMG)

The electromyogram measures the electrical potential generated by muscle cells when the cells contract. The potential of this measurement ranges from .1-5 mV and between the frequencies dc-10,000 Hz. The mean frequency of muscle fatigue is approximately 200 Hz. There is a nonlinear relationship between the potential and the force of the contraction. The RMS amplitude of the signal is approximately equal to logarithm of the force of the contraction. The EMG can be used to detect problems and diseases such as Myasthenia Gravis, Muscular Dystrophy, and denervation of the muscle.

Links

Electromyogram Lab Report (PDF)

Variable Hardware Inductance Simulator

This report documents the design and implementation of a variable hardware inductance simulator during the spring 2006 semester. This project was completed at the Institute for Advanced Technology for the purpose of research in the field of pulsed-power applications. Included are the design schematics and implementation details used to create a ‘black box’ capable of simulating a user-defined inductance between 100mH and 10H to an accuracy of 5%. The deliverable simulates an inductor for inputs between -120V and 120V with up to 20A of current at frequencies below 1 kHz.

The implementation process involved two competing designs based around the same basic circuit. The first employed a switching MOSFET to control the current, while the more favorable design used a bipolar junction transistor operating in the linear region. The latter design was chosen to be implemented on the full power scale.

The testing process was completed using a desktop power supply controlled by an isolated waveform generator. The simulator performed within 5% of the ideal input current for a variety of pulse shapes and lengths longer than 1msec. In order for the first 10msec of the input current to match the integration of the forst 10msec of the input voltage, an all-time bias current of 10-20mA is necessary. Therefore, except for the initial DC current, the product met specifications within its operating bandwidth.

The project cost $5028, and was $190 under budget. The project met all time and monetary restrictions; although, it involved more professional expertise than was initially estimated.

All relevant safety and ethical considerations have been addressed, and the deliverable is ready for integration into its application area.