Project Groups

Adaptive Noise Cancellation in a Weather Forecast System

• Student : Vincent Gire
• PI : Vincent Gire
• Synopsis: The aim of the project is to reduce noise in FSK demodulation. It will involve 1) modelling the transmission channel to take into account potential multi-paths existing between transmitter and receiver, and 2) the reduction of periodic noise by adaptive noise cancellation.

Cardiology: Defibrillator

• Students: Nick Dargahi, Michelle Hewlett, Thomas John, Baharan Kamousi
• PIs : Bryant Lin, Amin Al-Ahmad
• Synopsis:

Channel Equalization in Underwater Acoustic Communications

• Student : Leili Baghaei Rad
• PI : Leili Baghaei Rad
• Synopsis:

Cognitive Memory: Control Systems

• Students: Sufen Fong, Martin Trautmann
• PI : Bernard Widrow
• Synopsis: Investigating how cognitive memory can be used for control problems. To keep things simple, the inverted pendulum is studied to enable comparison of control properties of cognitive memory to that of other control algorithms.

Cognitive Memory: Curiosity

• Students: Dave Kale
• PIs : Bernard Widrow, Sudeepto Chakraborty
• Synopsis: Emulating a baby (Curiosity) exhibiting curiosity, given certain "inborn" knowledge pre-loaded into her (cognitive) memory, as she moves randomly about a virtual landscape acquiring knowledge about her environment.

CT Reconstruction

• Students: Jongduk Baek, Hoon Cho, HyukJoon Kwon
• PI : Jongduk Baek
• Synopsis: In CT reconstruction it is important to have uniform x-ray source intensity. However, actual x-ray source has intensity variation, and this intensity is time varying. Intensity calibration is important because non-uniform intensity can give rise to image artifacts. The goal is to perform intensity calibration using LMS algorithm.

Decoding Neural Signals from Premotor Cortices

• Students: Zuley Rivera Alvidrez
• PI : Zuley Rivera Alvidrez
• Synopsis: Neural networks will be used to decode neural signals from monkeys' premotor cortices. Neural network techniques will be applied to the signal processing of neural signals recorded from the brains of monkeys while they perform simple reach tasks.

Dynamic Prediction of Tumor Movements

• Student :Paul J. Csonka
• PI :Lei Xing
• Synopsis:Using computer vision to continuously predict tumor location some number of frames into the future, in order to more accurately radiate them.

Eye-tracking 1

• Students: Jonathan Hsieh, MingGang Guo, Jason Min
• PI : Joyce Liao
• Synopsis: Building a model of eye-movement following a dot on the screen. Creating analytical tools to show real-time eye-movement and automatically detecting abnormalities.

Eye-tracking 2

• Students: Sufeng Li, Michael Shimasaki
• PI : Joyce Liao
• Synopsis: Explore the response of eye movement in relation to a moving object viewed on a screen. Design an adaptive filter that models normal eye movement in reponse to a continuously moving object modelled as a first order Markov process. This will enable detection of abnormalities by comparison with the output of the adaptive filter.

LMS Adaptive Digital Background Calibration of Charge Redistribution ADCs

• Student : Stephen O'Driscoll
• PI : Stephen O'Driscoll
• Synopsis: Investigate whether or not LMS adaptation is a useful calibration technique for charge redistribution analog to digital converters (ADCs). LMS adaptive calibration techniques have been used successfully in pipeline ADCs as discussed in a number of publications from UC Berkeley over the past 3-4 years.

Non-linear Power Amplifier Distortion Processing using LMS

• Student : Liquan Tan
• PI : Liquan Tan
• Synopsis: By using LMS to implement automatic pre and/or post distortion processing, non-linearity in the power amplifier can be addressed much more elegantly. The difference between this and conventional pre/post distortion treatment is that the LMS algorithm may be implemented in analog circuits, eliminating the need for an ADC or digital processing block.

Self-reference Wideband Noise Cancellation

• Student : Kapil Rai
• PI : Kapil Rai
• Synopsis: Wideband signal and additive interference will be decomposed into L-subspace dimensions and cancellation performed adaptively in every dimension with a synthetic reference signal. Time permitting, multiple antennae will be considered to obtain some advantage in the spatial domain.