Superb-IT 2007 Participants
|
Maria J. Bustamante Project Title: Real-time Video and Image Processing of Rats Whiskers Movement Patterns in Order to Analyze Brain Response |
||
| As modern day society becomes more dependent on technology, researchers try to find a way of making technology useful not only for entertainment or business purposes, but also an essential resource to counteract health issues. The Brain Machine Interface (BMI) research deals with associating the human brain and machines together in order to make possible certain natural human actions such as moving or grabbing objects. At this stage, the BMI research is using rats as subjects for the numerous experiments. In the same way humans use their fingertips to feel surfaces; rats move their whiskers to recognize certain patterns. As the rodents feel different surfaces, they create distinct whisker movement patterns. In order to correlate the various brain signals as they feel the different patterns, we need a precise tracking device that can easily show the connection between the movement of whiskers and brain signals in real time. The frequency of the whisks is around 8 to 12 Hz; therefore, the real time video must be processed at a frequency of 100 frames per second, since we want it to be well above the Nyquist rate (~20Hz) for best quality results. Matlab is being used to develop a program that processes real time videos. To better track the whisks, a small red dot is placed on the rat's whisker. The program tracks the dot and removes the whisker from the background; therefore forming a binary image with the region of interest being the dot, which is represented as ones, and the background and whisker being represented as zeros. The program also detects the centroid of the region of interest, which is the key to provide a graph of the whisker movement vs. time. Once the movement of the whiskers is properly tracked, several tasks such as micro stimulation of the brain will be performed in a more accurate manner. | |||
|
|
|||
|
Maribel Hudson Project Title: A Low Cost Rapid Prototyping of Millirobots System | ||
| The purpose of this project is to implement an automated assembly routine that can be used to pick and place small, surface-mount resistors using strictly visual feedback. The system that will be used is the low-cost version of the orthotweezers. It consists of a three-axis stage (a flexure based two-axis stage, and a servo driven, Newport-based Z stage), two independently actuated tweezer arms, and a microscope with a digital camera. The motors controlling the system are driven using the Servio board and Python for the serial interface. The vision functions are provided by OpenCV in combination with Python wrappers that allow them to be called from withing Python scripts and applications. | |||
|
|
|||
|
Lisa Miller Project Title: Approximate Nonstationary Convolution Using a Block Separable Matrix |
||
|
When an image is generated wholly by computer, every object within it is rendered in precise focus. This causes the image to look unnatural because anything viewed through a lens (including our own eye lenses)will always have a blurring of objects that are not the same distance from the lens as the object specifically focused on. There is a range in clear focus and the rest of the image is blurred with a degree depending on its distance away from the object(s) in focus, this is normally referred to in photography as depth of field. In order to create realistic looking computer generated images, artificial blurring must be added after the fact. This post-process method has been done very successfully for still images using a convolution matrix, but the method runs into problems when trying to blur video images on the fly. The required matrix multiplication is very resource heavy; direct matrix multiplication is O(N2) with N being the size of the image in pixels. Therefore, this method is generally too slow to be processed in real time. My project involves using a quad-tree, implemented as a C++ data structure, to compress the convolution matrix into blocks of varying sizes. Each block of a block separable matrix can be processed in O(M) time, where M is the width of the block. If a constant upper bound for the number of blocks is fixed, the algorithm should be able to run in O(N) time. I will be attempting to get a faster blurring time using this method by processing the blocks with a tree traversal, while maintaining acceptable image quality. Additionally, I will be attempting to allow the degree of blur and the shape of the blurred field to be readily changeable, which is not possible with other methods currently in use such as a Fast Fourier Transform, which is O(NlogN). This project has immediate application to the entertainment industry, especially video games, and to more serious endeavors such as virtual reality for the military and health- care fields. |
|||
|
|
|||
 |
Javier Rosa Project Title: Mapping Genomic Markers Associated with the Presence of Schizophrenia in a Population of Highly Related Individuals |
||
| Schizophrenia is a psychiatric illness which affects 2.5 million people worldwide. We intend on using genetic data gathered from a relatively isolated population in Israel that can be traced back to 5 males that were alive 250 years ago in order to find genetic variance that is highly associated with the presence of the disease. In order to do this we will create new statistical methods and algorithms designed specifically for such isolated populations. The findings of our study could be used to find new treatments for Schizophrenia or to screen for individuals that are at high risk for the disease. | |||
|
|
|||
 |
Kevin Simmons Project Title: Visualizing the 4D Blur Matrix |
||
| Computer graphics can be classified into subfields of geometry, animation, rendering, and imaging. Rendering has uses in architecture, video games, TV special effects, and design visualization, each employing a different balance of features and techniques. The three steps involved in rendering are the following: create a set of depth images, blur each depth image, and composite the blurred depth images to form a single vision-realistic rendered image. Many rendering algorithms have been researched, and software used for rendering may employ a number of different techniques to obtain a final image. The purpose of my project is to create an image that can be blurred. The user will see shades of grey which means the image will transition from white to gray and then finally to black. My program will have a graphical user interface that will include buttons, sliders, and checkboxes that will play a key role in displaying the image. I will be using various mathematical equations and algorithms that will be implemented into the C++ programming language. | |||
|
|
|||
 |
Juan Yaquian Project Title: Serial-Parallel Control Interface for Tunable Analog Circuits Performances |
||
| Today, technology is becoming extremely dependant on wireless capabilities and future technology will be rendered almost useless unless it has strong wireless capabilities and provides high performance at high frequencies. However, current technology encounters complications when operating at the necessary high frequencies. To address the issues, novel approaches must be sought in the areas of circuit design and device modeling in an effort to provide more reliable technology at the mm-wave frequencies. One complication that is currently being addressed by researchers at the Berkeley Wireless Research Center (BWRC) deals with the level of distortion in a signal at the front-end of a receiver operating at high frequencies. The retrieved signal at the output of the front-end must be rid of its high level of distortions if the digital applications at the back-end of the receiver are to be able to properly use the received signal. Researchers in my group are actively pursuing the front-end linearization schemes in the circuit design level to address this problem. A typical front-end makes use of a low noise amplifier (LNA) and a mixer. Research has shown a LNA topology capable of broadband high linearity if all its voltage bias knobs, when applied at the right value, serve as a sort of lock combination in that when the correct combination is applied, a linear amplification is achieved. One key issue to enable this highly linear LNA is to allow flexibility of multiple bias voltage tune ability. This relies on multiple bias control wires connecting to each knobs of the LNA. A large number of pin counts and pad area are required if parallel connection is to be used in the silicon implementation. A more desirable approach is to serialize the parallel control into a sequential sequence and use a synchronized clock to latch each of the bias control and distribute it to corresponding bias knob. Moreover, the design of a Serial-Parallel Control Interface for the LNA can be generalized and reused in any other analog circuits whose performances are bias voltage dependent and tunable. | |||
|
|
|||