Presentation Schedule
Fall 2016
Abstract
Slide #Last Name First Name Title & Abstract of Presentation Ph.D Advisor Day & Time Venue
4 Akbar Md Navid Widely Linear Estimation with Complex Data Mohammad SaquibNovember 28
11:00-12:00 amECSN
4.728
5 Arjona Angarita Ricardo Javier Performance Analysis of the IEEE 802.11 Distributed
Coordination FunctionAndrea Fumagalli
October 281:00-3:00pm
ECSN 3.524
6 Arunachalam Harish BabuA Review of Methods for Whole Slide Histopathological
Image Analysis for Tumor Detection Ovidiu Daescu
October 19 12:00-1:00 pm
ECSS 4.910
7 Cao Beiming Silent Speech Recognition Using Inversely Mapped
Articulatory DataJun Wang
October 271:00-2:00 pm
BSB 11.102E
8 Challapalli Niharika Tight Bounds for Compressed Sensing Algorithms Mathukumalli
VidyasagarOctober 12
10:00-11:00 amECSS 3.910
9 Chen Yingping Integrated Isolated Power Converter Using Active
Rectification and Closed-Loop CRM Control for Secondary Side Regulation in E-Meters
Dongsheng MaNovember 282:00-3:00 pm
ECSN 3.804
10Ershad
LangroodiMarzieh
Identification of Meaningful Skill Assessment Metrics Using the Wisdome of Crowd
Ann MajewiczOctober 13
9:00-10:00 amECSN 2.704
11 Erturk Feyzullah A Method for Online Incipient Fault Detection in SiC
MOSFETs Bilal Akin
October 129:30-10:30 am
ECSN 4.702
12 Gour Riti Robust Routing in Networks with Probabilistic Failures Jason JueNovember 16 1:00-2:00 pm
ECSS 4.910
13 Hao YiyaNondeterministic Sound Source Localization with
Smartphones in CrowdsensingIssa Panahi
October 2811:00-12:00 pm
ECSS 3.910
Table of Contents
Abstract
Slide #Last Name First Name Title & Abstract of Presentation Ph.D Advisor Day & Time Venue
14 Jha Sumit Analysis of Driver Visual Attention Carlos BussoNovember 102:00-3:00 pm
ECSN 4.702
15 Kumar SauravExtremum Seeking Control for Multi-Objective Optimization
Problems Robert D Gregg
October 272:00-3:00 pm
ECSN 4.728
16 Li SenAn Energy-Stored Quasi-Z-Source Inverter for Application to
Photovoltaic Power System Poras Balsara
November 17 10:30-11:30 am
ECSN 4.728
17 Li Xi Optical Antenna Enhanced Light Emitting Devices Qing GuNovember 183:30-5:00 pm
ECSN 4.728
18 Liu Jiawei Noise Subspace-Based Iterative Technique for Direction
Finding Mohammad Saquib
October 26 10:00-11:00 am
ECSN 4.728
19 Lotfi Mahsa Prediction of Time to Tumor Recurrence in Ovarian Cancer
using Sparse Regression Methods Mathukumalli
VidyasagarOctober 12
12:00-1:00 pmECSS 3.910
20 NimmalapudiSai Govinda
Rao Self-Correcting Op-Amp Input Offset Using Analog Floating
Gate Memory Andrew Marshall
November 810:00-11:00 am
ECSS 3.503
21 YelleswarapuVenkata Pavan
Kumar Concepts and Methods in Optimization of Phase Noise in LC
VCOs Kenneth O
November 29 10:00-11:00 am
ECSN 3.802
22 Zhao Zhongfan Literature Review of Fast Extremum Seeking Control for
Wiener-Hammerstein Plants Yaoyu Li
November 28 9:30-11:30 am
ECSN 4.702
Widely Linear Estimation with
Complex Data
Mean square estimation of complex and normal datais not linear as in the real case but widely linear. Thepurpose of this correspondence is to calculate theoptimum widely linear mean square estimate and topresent its main properties. The advantage withrespect to linear procedure is especially analyzed.
Abstract:
Md Navid AkbarNovember 28, 2016 │ 11:00 am-12:00 pm │ ECSN 4.728
PhD Advisor: Dr. Mohammad Saquib
Performance Analysis of the
IEEE 802.11 Distributed
Coordination Function
To present an analytical model to compute the802.11DCF throughput, in the assumption of finitenumber of terminals and ideal channel conditions. Themodel is applied to the basic access and RTS/CTSaccess mechanisms with an extensive throughputperformance evaluation.
Abstract:
Ricardo Javier Arjona AngaritaOctober 28, 2016 │ 1:00 pm-3:00 pm │ ECSN 3.524
PhD Advisor: Dr. Andrea Fumagalli
A Review of Methods for Whole
Slide Histopathological Image
Analysis for Tumor Detection
Histopathological image analysis, an important sub-field in Pathology
informatics, deals with identifying patterns and extracting features
from whole slide images that help in aiding tumor prognosis or
diagnosis. In most of the research studies, histopathological images
used for tumor analysis, contain data that is rich in information about
the cellular level details. This presentation covers various image
processing algorithms that are used in tumor studies, their limitations
and explains why it is difficult to develop a standard pipeline of
algorithms that are applicable for all types of tumor.
Abstract:
Harish Babu ArunachalamOctober 19, 2016 │ 12:00–1:00 pm │ ECSS 4.910
PhD Advisor: Dr. Ovidiu Daescu
Silent Speech Recognition
Using Inversely Mapped Articulatory Data
Silent speech recognition (SSR) is the process of converting non-audio (sensors) signals to
speech in the form of text. The primary application of SSR is assisting people with difficulties
in phonation such as laryngectomees, who have undergone a surgical removal of larynx for
treating laryngeal cancer. Articulatory movements (e.g., tongue, lips) are less affected by the
impairment of larynx compared to normal acoustic signals. In this study, I use the movement
signal of four sensors (upper lip, lower lip, tongue tip, tongue back) attached to speakers'
articulators in my experiments. In this talk, I will present the experiments I have done on silent
speech recognition including using articulatory movement data from healthy speakers,
laryngectomees, and whispered speech that were recently collected in our lab. Two speaker
normalization approaches, Procrustes matching, a physiological approach, and fMLLR, a
data-driven approach have been used to reduce the speaker articulation variation. The
experimental results indicated the effectiveness of these approaches. I am currently exploring
inverse mapping approaches (acoustic-to-articulatory mapping), which will be used for
generating articulatory data from a large acoustic dataset. Then we will have the possibility to
avoid collecting a large articulatory movement dataset, which is logistically difficult.
Abstract:
Beiming CaoOctober 27, 2016 │ 1:00–2:00 pm │ BSB 11.102E
PhD Advisor: Dr. Jun Wang
Tight Bounds for Compressed
Sensing Algorithms
The LASSO and the Elastic Net (EN) formulations are among the most popular approaches
for sparse regression and compressed sensing. We introduce a new optimization formulation
for sparse regression and compressed sensing, called CLOT (Combined L-One and Two),
wherein the regularizer is a convex combination of the L1 - and L2 -norms. This formulation
differs from the Elastic Net (EN) formulation, in which the regularizer is a convex combination
of the L1 – and L2 -norm squared. This seemingly simple modification has fairly significant
consequences. In particular, it is shown that the EN formulation does not achieve robust
recovery of sparse vectors in the context of compressed sensing, whereas the new CLOT
formulation does so. Also, like EN but unlike LASSO, the CLOT formulation achieves the
grouping effect, wherein coefficients of highly correlated columns of the measurement (or
design) matrix are assigned roughly comparable values. It is noteworthy that LASSO does not
have the grouping effect and EN does not achieve robust sparse recovery. Therefore the
CLOT formulation combines the best features of both LASSO (robust sparse recovery) and
EN (grouping effect).
Abstract:
Niharika ChallapalliOctober 12, 2016 │ 10:00–11:00 am │ ECSS 3.910
PhD Advisor: Dr. Mathukumalli Vidyasagar
Integrated Isolated Power Converter Using Active
Rectification and Closed-Loop CRM Control for
Secondary Side Regulation in E-Meters
An integrated isolated power converter is presented to mitigate efficiency, accuracy and
cost challenges on secondary side regulation. It employs time-multiplexing scheme to
seamlessly power the non-isolated primary and the isolated secondary output. With the
coordination of proposed primary state detection (PSD) circuit, the secondary side output
is accurately regulated and automatically synchronized to the primary side without the use
of bulky transformers or optocouplers. In collaboration with the PSD, an active rectifier is
adopted, which replaces conventional power diode and LDO for efficiency enhancement.
Designed in a 0.35µm BCD process, the converter achieves closed-loop regulations on
both the non-isolated and the isolated output at 5V, with an input source of 24V. Thanks to
the proposed circuits and control scheme, the peak efficiency at secondary side is
improved by 12.4%, with the output variation reduced to 0.6% in a load range of 500mA.
Abstract:
Yingping ChenNovember 28, 2016 │ 2:00-3:00pm │ECSN 3.804
PhD Advisor: Dr. Dongsheng Ma
Identification of Meaningful Skill
Assessment Metrics Using the
Wisdome of Crowd
Robotic surgery requires skills which differ from that of open and laparoscopic surgery.
Training surgical residents to master these techniques is an ongoing field of research. In this
study we aim to find the data metrics (motion and physiological data) which best correlates
to the crowd’s choices of word from the semantic labeling lexicon of surgical expertise. Three
experts, three intermediates and three novices participated in our study. They completed two
tasks on the da Vinci surgical simulator: a ring and rail task and a a suture sponge task.
Different data measurement were acquired using sensors which recorded limb (hand and
arm) acceleration and angular velocity and joint (shoulder, elbow, wrist) positions. Posture
videos of the subject performing the task as well as videos from the task being performed
were rated by crowd workers and the best correlations between the data metrics and crowd
assignments were found. Through this we were able to find the best measurement data
metric for the word choices in our lexicon.
Abstract:
Marzieh Ershad LangroodiOctober 13, 2016 │ 9:00–10:00 am │ ECSN 2.704
PhD Advisor: Dr. Ann Majewicz
A Method for Online Incipient
Fault Detection in SiC MOSFETs
This presentation introduces a comprehensive study on degradation monitoring of
SiC MOSFETs and propose a method to detect incipient faults for early warning in
power converters and smart gate drivers. During the accelerated ageing tests (power
cycling) several electrical parameters are recorded to analyze critical signatures and
precursors for early fault detection. Among those, gate leakage current is identified as
one of the most practical precursor which exhibit consistent changes throughout the
aging and relatively easy to monitor. The proposed method is experimentally justified
which can be integrated to a gate driver to monitor the FETs condition. This method
naturally fits to the applications which cannot tolerate interrupts caused by
unpredicted failures. Due to its simple scheme and low cost, it can potentially be
embedded into commercial gate drivers featuring improved reliability options.
Abstract:
Feyzullah ErturkOctober 12, 2016│ 9:30–10:30 am │ECSN 4.702
PhD Advisor: Dr. Bilal Akin
Robust Routing in Networks with
Probabilistic Failures
I will be presenting a journal paper from IEEE/ACM Transactions on Networking titled Robust
Routing in Networks with Probabilistic Failures, by Hyang-Won Lee, Eytan Modiano, Kayi Lee.
The abstract of the paper is mentioned below. I will also be including some part of my research
work towards the end of the presentation which is related to ensuring survivability for multi-
domain optical networks. Abstract of the journal paper—We develop diverse routing schemes
for dealing with multiple, possibly correlated, failures. While disjoint path protection can
effectively deal with isolated single link failures, recovering from multiple failures is not
guaranteed. In particular, events such as natural disasters or intentional attacks can lead to
multiple correlated failures, for which recovery mechanisms are not well understood. We take a
probabilistic view of network failures where multiple failure events can occur simultaneously,
and develop algorithms for finding diverse routes with minimum joint failure probability.
Moreover, we develop a novel Probabilistic Shared Risk Link Group (PSRLG) framework for
modeling correlated failures. In this context, we formulate the problem of finding two paths with
minimum joint failure probability as an integer nonlinear program (INLP) and develop
approximations and linear relaxations that can find nearly optimal solutions in most cases.
Abstract:
Riti GourNovember 16, 2016│ 1:00-2:00 pm │ECSS 4.910
PhD Advisor: Dr. Jason Jue
Nondeterministic Sound
Source Localization with
Smartphones in Crowdsensing
The proliferation of smartphones nowadays has enabled many crowd assisted applications including audio-
based sensing. In such applications, detected sound sources are meaningless without location information.
However, it is challenging to localize sound sources accurately in a crowd using only microphones
integrated in smartphones without existing infrastructures, such as dedicated microphone sensor systems.
The main reason is that a smartphone is a nondeterministic platform that produces large and unpredictable
variance in data measurements. Most existing localization methods are deterministic algorithms that are ill
suited or cannot be applied to sound source localization using only smartphones. In this paper, we propose
a distributed localization scheme using nondeterministic algorithms. We use the multiple possible
outcomes of nondeterministic algorithms to weed out the effect of outliers in data measurements and
improve the accuracy of sound localization. We then proposed to optimize the cost function using least
absolute deviations rather than ordinary least squares to lessen the influence of the outliers. To evaluate
our proposal, we conduct a testbed experiment with a set of 16 Android devices and 9 sound sources. The
experiment results show that our nondeterministic localization algorithm achieves a root mean square error
(RMSE) of 1.19 m, which is close to the Cramer-Rao bound (0.8 m). Meanwhile, the best RMSE of
compared deterministic algorithms is 2.64 m.
Abstract:
Yiya HaoOctober 28, 2016 │ 11:00 am-12:00pm │ ECSS 3.910
PhD Advisor: Dr. Issa Panahi
Analysis of Driver Visual Attention
Monitoring driver behavior is crucial in the design of advanced driver assistance systems (ADAS)
that can detect driver actions, providing necessary warnings when not attentive to driving tasks.
The visual attention of a driver is an important aspect to consider, as most driving tasks require
visual resources. Previous work has investigated algorithms to detect driver visual attention by
tracking the head or eye movement. While tracking pupil can give an accurate direction of visual
attention, estimating gaze on vehicle environment is a challenging problem due to changes in
illumination, head rotations, and occlusions (e.g. hand, glasses). Instead, this study investigates
the use of the head pose as a coarse estimate of the driver visual attention. The key challenge is
the non-trivial relation between head and eye movements while glancing to a target object, which
depends on the driver, the underlying cognitive and visual demand, and the environment. First, we
evaluate the performance of a state-of-the-art head pose detection algorithm over natural driving
recordings, which are compared with ground truth estimations derived from AprilTags attached to a
headband. Then, the study proposes regression models to estimate the drivers’ gaze based on the
head position and orientation, which are built with data from natural driving recordings. The
proposed system achieves high accuracy over the horizontal direction, but moderate/low
performance over the vertical direction. We compare results while our participants were driving,
and when the vehicle was parked.
Abstract:
Sumit JhaNovember 10, 2016 │ 2:00-3:00pm │ ECSN 4.702
PhD Advisor: Dr. Carlos Busso
Extremum Seeking Control for Multi-
Objective Optimization Problems
Many real world applications require multiple conflicting objectives to be optimized. However,
explicit functions mapping the system variables to outputs are often unknown, making traditional
optimization approaches impossible. Extremum Seeking Control (ESC) is one method to tackle
this problem by estimating the local gradient of the objective functions. Extensive literature on
ESC focuses on single objective optimization, which use sinusoids as the dither signal. An
inherent disadvantage with the sinusoidal nature of the dither is that after convergence the output
continues to move around in a neighborhood of the optimal set-point, with the size of the
neighborhood proportional to the amplitude of the dither signal. This presentation focuses on using
square waves as dithers and proves that the square wave can produce a constant output, in
contrast to the typical sinusoid where the undesirable dithering effect still exists at the output. The
basic ESC scheme is extended to execute Multiple Gradient Descent Algorithm (MGDA) to solve a
Multi Objective Optimization Problem (MOOP). A two-stage ESC estimates the local gradient of
each objective function, then estimates the optimal weighting of gradients to move to the Pareto
front. This eliminates the need for a decision maker as required in a priori scalarization solutions a
MOOP. Simulation results show that ESC using MGDA is able to find the Pareto optimal solutions,
starting from different initial conditions. An application of a simple ESC is illustrated on auto- tuning
the PD gains of the knee and the ankle joints for a prosthetic controller.
Abstract:
Saurav KumarOctober 27, 2016 │ 2:00-3:00 pm │ ECSN 4.728
PhD Advisor: Dr. Robert D Gregg
An Energy-Stored Quasi-Z-Source
Inverter for Application to
Photovoltaic Power System
This presentation starts with the introduction to the concept and control method of Z-
source converter. Then it comes to present a new topology of the energy-stored quasi-Z-
source inverter (qZSI) to overcome the shortcoming of the existing solutions in
photovoltaic (PV) power system. Two strategies are discussed with the related design
principles to control the new energy-stored qZSI. They can control the inverter output
power, track the PV panel’s maximum power point, and manage the battery power,
simultaneously. The voltage boost and inversion, and energy storage are integrated in a
single-stage inverter. An experimental prototype is built to test the proposed circuit and
the two discussed control methods. The obtained results verify the theoretical analysis
and prove the effectiveness of the proposed control of the inverter’s input and output
powers and battery power regardless of the charging or discharging situation. A real PV
panel is used in the grid-tie test of the proposed energy-stored qZSI, which demonstrates
three operational modes suitable for application in the PV power system.
Abstract:
Sen LiNovember 17, 2016 │ 10:30-11:30 am │ ECSN 4.728
PhD Advisor: Dr. Poras Balsara
Optical Antenna Enhanced Light
Emitting Devices
Small on-chip light emitting devices are desirable for potential applications, including optical
communication, biomedical sensors, high resolution imaging and spectroscopy. However,
diffraction limit and high threshold have hold back the miniaturization of such devices. To overcome
restrictions imposed by the diffraction limit in all-dielectric devices, plasmonics, which uses metals
for extreme light concentration and manipulation in nanostructures, is introduced. Recent studies
on plasmonic lasers have shown the possibility to squeeze the effective mode volume into
subwavelength range. As nano-emitters of small aperture typically suffer from impedance
mismatch, resulting in the detectable output power to be much less than the already minuscule
emitted power, the idea of applying the optical antenna concept to the nano-emitter design
becomes very attractive. Similar to its radio frequency and microwave counterparts, optical antenna
serves as a transducer to convert localized energy to the energy of free propagating radiation, and
vice versa. Thus, nano-antenna can bridge the impedance mismatch between nano-emitters and
free space radiation, and concurrently enhance electric field confinement to increase radiation
efficiency of the device. Furthermore, it is feasible for designers to make trade-offs between high
electric field localization, broad bandwidth, high directivity by selecting different antenna types. The
small size of optical antennas also enables on-chip phased array configuration for even more
complex functions such as beam steering.
Abstract:
Xi LiNovember 18, 2016 │ 3:30-5:00pm │ ECSN 4.728
PhD Advisor: Dr. Qing Gu
Noise Subspace-Based Iterative
Technique for Direction Finding
In the area of array signal processing, direction of arrival (DoA) estimation is a widely
studied topic. In this estimation process, the noise subspace of the received signal
covariance matrix is often utilized and obtained through numerical methods. We explicitly
derive an algebraic expression of the noise subspace when the number of signal sources
present is less than the number of elements of a uniform linear array (ULA). This expression
of the noise subspace is then used to formulate a constrained minimization problem to
obtain the DoAs of all the sources in a scene in the presence of spatially white noise of
identical power. This noise subspace-based estimation (NISE) algorithm iteratively solves
for each source’s DoA, potentially yielding (depending on the number of iterations) lower
complexity than existing DoA estimation algorithms, such as fast root-MUSIC (FRM), while
exhibiting performance advantages for a low number of time samples and low signal-to-
noise ratio (SNR). The convergence of NISE is then proven mathematically. In addition, it is
shown how NISE can readily incorporate prior knowledge into the DoA estimation process.
Abstract:
Jiawei LiuOctober 26, 2016 │ 10:00-11:00am │ ECSN 4.728
PhD Advisor: Dr. Mohammad Saquib
Prediction of Time to Tumor
Recurrence in Ovarian Cancer using
Sparse Regression Methods
Ovarian cancer is the most fatal and the most aggressive gynecological malignancy. It is also
considered to be the fourth commonest cause of cancer deaths worldwide and also the cause of half
of the deaths related to gynecological cancers. Therefore, cancer prognosis plays an important role in
the evaluation and the treatment of a cancer patient. Sparse regression seems to be a useful tool in
predicting the time to tumor recurrence in Ovarian cancer. By applying sparse regression methods
such as CLOT and LASSO to the gene expressions of Ovarian cancer samples, one would be able to
prognosticate the time of tumor recurrence and also to compute the concordance index which is a
prognostic factor in Ovarian cancer. Due to the high dimensionality of gene expression data,
recursive feature elimination method is also recommended for the regression methods in order to
decrease the number of features used. In this study, we have applied two sparse regression
methods, CLOT and LASSO to four different Ovarian cancer datasets and computed the
concordance index for these datasets. Results show that sparse regression methods are more
successful in the computation of the concordance index for Ovarian cancer than previous methods in
overall.
Abstract:
Mahsa LotfiOctober 12, 2016 │ 12:00-1:00 pm │ ECSS 3.910
PhD Advisor: Dr. Mathukumalli Vidyasagar
Self-Correcting Op-Amp Input Offset
Using Analog Floating Gate Memory
Keeping input offset voltage low is important in high precision Op-Amps. However, input offset
errors caused by mismatch in differential signal paths as a result of random variations are
unavoidable even with optimum layout techniques. Mismatch increases as transistor
geometries reduce, so various techniques have been developed to minimize this, including:
increasing the size of the input pair, auto-zeroing, chopper stabilization and digital trim
techniques using such as flash, fuse or EEPROM. A relatively new method, the use of Analog
Floating Gate (AFG) devices, to enable correction is being studied. AFG devices act as
analog storage and allow precise trimming of input offset. AFG devices have been included in
an operational transconductance amplifier, where they can be trimmed to correct for input
offset. The proposed methodology results in offset correction for continuous time operation,
provides low power operation, does not limit bandwidth and avoids discrete errors seen with
some correction methods. However, unlike some other methods AFG devices have a
tendency to lose charge over time, typically a few mV per year. As a result, we have
developed circuitry that automatically recalibrates the AFG charge and therefore retains Op-
amp offset targets.
Abstract:
Sai Govinda Rao NimmalapudiNovember 8, 2016 │ 10:00-11:00 am │ ECSS 3.503
PhD Advisor: Dr. Andrew Marshall
Concepts and Methods in Optimization
of Phase Noise in LC VCOs
Integrated LC voltage-controlled oscillators (VCOs) are common functional blocks in modern
radio frequency communication systems and are used as local oscillators to up-convert and
down-convert the signals. Circuit noise and device noise can disturb both the amplitude and
phase of an oscillator’s output. Because amplitude fluctuations are usually greatly attenuated,
as a result, phase noise generally dominates. Phase Noise in oscillators can be described as
random frequency fluctuations of a signal, and negatively impacts both the transmitter and
receiver chains of the system. So, it is highly important to design a low phase-noise oscillator. A
design strategy centered around an inductance selection scheme is executed to optimize the
phase noise subject to design constraints such as power dissipation, tank amplitude, tuning
range, startup condition, etc. Two modes of operation, named current-limited and voltage-limited
regions, are studied for a typical LC oscillator. Important concepts such as waste of inductance
and waste of power have been observed in these modes of oscillator operation. As a result, the
design strategy can be summarized as follows: for a given bias current, phase noise increases
with an increasing inductance in the voltage limited region, which corresponds to waste of
inductance. The phase noise also increases with the bias current in the voltage limited region,
which corresponds to waste of power.
Abstract:
Venkata Pavan Kumar YelleswarapuNovember 29, 2016 │ 10:00-11:00am │ ECSN 3.802
PhD Advisor: Dr. Kenneth O
Literature Review of Fast
Extremum Seeking Control for
Wiener-Hammerstein Plants
Extremum Seeking Control (ESC) has emerged as a class of model-free real
time optimization algorithm, which is of particular interest for control
applications where acquisition of accurate plant models for model based
approaches is cost or even prohibitive. One major drawback for the
conventional ESC scheme is the relatively slow convergence due to the
nature of time-scale separation for such framework. Remarkable research has
recently been conducted on fast dither based ESC algorithm for Wiener-
Hammerstein plants, with the primary results obtained based on semi-global
analysis. Substantial improvements have been observed. This review will
describe the relevant analysis in both continuous-time and discrete-time
domain, including a linear parameter-varying approach.
Abstract:
Zhongfan ZhaoNovember 28, 2016 │ 9:30-11:30am │ ECSN 4.702
PhD Advisor: Dr. Yaoyu Li