|
| Electronics and Information Processing |
| |
| (Faculty:
S. Bhansali, D. Hilbelink,
K. Muffly,
N.
Ranganathan,
R. Sankar,
R. Schlaf,
T. Weller,
P.
Wiley) |
| |
| Our vision is to produce inexpensive sensing
devices by means of sensor integration and electronic
miniaturization using MEMS. These devices will monitor a wide
variety of physiological parameters and will be equipped with
significant processing and memory (i.e., sensor integration with DSP).
They also will provide wireless communication capabilities where the
sensors themselves become nodes in an ad-hoc network (i.e., sensor
integration with RF communication). |
| |
| Characterization of ultrasonic probe
discrimination ability in the investigation of skin biomechanical
and anatomical properties |
| |
| Development of skin tissue models: Several models
will be developed to assist in the experimental investigations. This
will include natural tissues such as human cadaver skin and pigskin
and other materials such as gelatin-based “spiked” samples. All
tissues will be fully characterized anatomically using appropriate
optical microscope and other techniques. We also intend to
incorporate skin with known diseases and other conditions of
interest, including various forms of skin cancer. In addition, age,
gender, and other compositional influences will be addressed. Longer
term, informed human subjects will be included as part of the
clinical validation process, again including normal skin and other
medical conditions of interest (cancer, traumatized tissue, wound
healing, aging, etc.). All human subject testing will be approved by
USF’s Institutional Review Board (IRB). |
| |
| Characterization of ultrasound devices (low and
high frequency): Using the skin models, the performance of 5, 10,
20, and 50 MHz devices will be explored, with an emphasis on
comparing the relative abilities of the various frequencies, using
both A-mode and B-mode. In addition to the pure frequency effects,
consideration of operator factors such as skin surface pressure will
also be addressed. In human subject testing, Doppler ultrasound will
also be included since skin microcirculation will be functional. |
| |
| Understanding ultrasound information for tissues
under mechanical loading: The model tissues will be subjected to
various mechanical loading (compressive, shear, torsional) and the
ultrasound responses analyzed for extractable information. In
parallel, the biomechanical properties of the specimens will be
characterized via stress-strain curves (except as related to human
subject testing). Real-time ultrasound response during dynamic
loading (for example, cyclic loading) will also be studied. |
| |
| VLSI and information processing for bio-sensor
systems |
| |
| Image reconstruction and processing algorithms: We
will develop efficient algorithms for 3-D image reconstruction from
the impedance measurements done by the microsensors. We will
investigate adapting existing approaches as well as new algorithms
for obtaining a 3-D image of the cell distribution under the skin
surface. Further, we will develop algorithms for processing the
images for tissue characterization. |
| |
| Image and data compression: We plan to develop
efficient lossless and medically lossless compression and
decompression of biomedical images. We have extensive expertise in
the design of data compression algorithms.Most existing lossy image
compression algorithms cannot be applied for biomedical images since
we do not want to loose “medically valuable” information and the
lossless ones designed for general purpose images will not exploit
specific properties in biomedical images. We intend to specifically
investigate skin and other biomedical images for lossy and the
lossless compression. |
| |
| Application Specific Integrated Circuits: We will
develop hardware algorithms and architectures for control of sensors
as well as decision making systems for applications such as alarm
triggering/monitoring and drug delivery. The ASICs will be realized
as either field programmable gate arrays or custom VLSI circuitry.
These circuits will be required to be area and power efficient while
yielding high speed computations. |
| |
| Synthesis and low power design techniques: We have
developed a new method for low power VLSI design called dynamic
frequency clocking which when combined with voltage scaling allows a
circuit to operate at different modes in terms of performance and
power consumption by allowing the chip to vary in speed dynamically
based on the operation being performed at a given time. We have also
developed new synthesis techniques for implementing basic datapath
components which will aid in design of low power circuits for
integration with the skin sensors. |
| |
| Intelligent decision making system: In the recent
past, we have developed an intelligent decision making system which
integrates a neural network for learning the dynamically changing
environment and an expert system which has rules coded based on
apriori knowledge about the system. Such a system results in a more
accurate and less expensive system compared to independent neural
network-based or expert system-based approaches. We have
successfully demonstrated several real-time applications such as
traffic control, autonomous vehicle navigation and failure control
where such an integrated system performs much better. For drug
delivery systems, precision decision making and failure control will
be critical and will beinvestigated in the context of skin sensor
control and response monitoring. |
| |
| Communications and information/signal processing
for biosensors |
| |
| Development of methods and algorithms for signal
processing, detection and classification using Statistical Hidden
Markov Models, Wavelet transforms, and Neural Network Approaches:
Biomedical signal processing is complex since (i) involves nonlinear
processing methods for better analysis of biomedical signals and
representation of signals as non-stationary random processes for
better accuracy. (ii) bulk volume of data need to be measured and
analyzed to extract any useful information (iii) exhibit low signal
to noise ratio or high degree of artifacts. Three well established
methods – that is – statistical hidden Markov models and
discriminant analysis, wavelet transforms, and neural networks will
be investigated for feature extraction, detection and
classification. We have extensive expertise in the development of
signal processing algorithms, which will aid in adapting exiting
techniques and as well as developing new methods suitable for this
application. |
| |
| Communications and Telemetry - Wireless Sensors
Networking: We propose to use low powered RF communication using
Bluetooth technology as a possible solution for communications and
telemetry among sensor devices. The sensors will permit remote
monitoring and tracking of physiological parameters from each of the
signal profiles such as bio-impedance, temperature, heart rate,
blood pressure, etc. by placing the sensors or electrodes on the
skin surface or in its proximity. Bluetooth provides low power (1 mW),
communications capability (peak data transfer rate of 1 Mbps) using
the unlicensed ISM band at 2.4 GHz. The sensor devices will be
networked to form piconets (small networks of master and slave
devices). The proposed work will involve tailoring the Bluetooth
protocols for the RF communications and telemetry from the sensors
to the external devices for processing and analysis. The cable
replacement protocol, RFCOMM will be implemented to emulate the
serial communications interface between the devices. In the next
phase, we plan on extending the piconets to form ad-hoc networks
where the sensor network will have distributed control for sharing
information among them and as well as to any external unit. We have
successfully developed intelligent techniques for mobility
management and protocols for wireless networks that will aid in the
design of wireless sensor network. |
| |
| RF electronics for integrated bio-sensors |
| |
| Low Voltage RF MEM Switches with Thermal &
Electrostatic Actuation: One of the challenges in MEM switch design,
which is especially important in the SKINS micro-sensor project, is
minimizing the required actuation force. Electro-static actuation is
most commonly used, and while it requires essentially no power for
operation the required voltages can be quite high (typically 10 Vdc
or greater). As actuation voltages are reduced, the switch can be
susceptible to vibration and RF performance can be affected. In this
project, a combination of electro-static and thermal actuation will
be utilized. The objectives will be to optimize the balance between
thermal and electro-static actuation to minimize power dissipation
and actuation voltage. The goal is to achieve RF MEM switches with
1.5V actuation voltage that remain stable in the vibration
environment expected to exist with the SKINS micro-sensors. |
| |
| RF MEM Switch Modulators for Low Data Rate
Communications: In communications applications requiring only low
data rates (a few KHz at the most) and close proximity (e.g., within
a room or laboratory, or less), application specific telemetry
systems can be designed that require little to no power to operate.
In this project, integrated architectures, built around the RF MEM
switches described above, will be investigated. In operation, a
continuous wave (CW) signal delivered by a source is received at the
sensor by a small, planar antenna, modulated using on-off keying by
the MEM switch, and subsequently re-transmitted in an orthogonal
polarization back to the source. Prototype systems that implement
the technique (passive modulating reflectenna, or PMR) are currently
being studied at USF. The focus of the IGERT-related project will be
to expand this work toward the development of higher frequency,
retrodirective PMR arrays. The retrodirective capability will allow
semi-arbitrary orientation between the sensor and the
transmitter/receiver. |
| |
| Multi-Layer Integration of High Performance RF and
Analog CMOS Circuitry: The goal of co-fabrication and co-location of
RF circuitry with the sensor-related structures will be addressed
through the IGERT project. Conventional and emerging design
methodologies for microwave passives (beyond a few GHz) are almost
exclusively based on the assumption of available thick, low-loss
substrate material. The IGERT project will address basic research in
design and fabrication of passive microwave circuitry that is
integrated on semi-conducting silicon substrates. Results of this
work will advance a capability to imbed microwave/wireless
connectivity on essentially any (potentially lossy) surface such as
silicon integrated circuits, solar cells and plastics. |
| |
| Design of Optimum Integrated Micro-Sensor Antennas
using Genetic Algorithm Techniques: A proposed IGERT project is to
develop efficient design and optimization techniques for integrated
antennas embedded in complex electromagnetic (EM) environments. The
antennas will be developed for the SKINS multi-layer sensor
architecture. The objective is to represent the entire sensor
architecture and surrounding medium within the numerical EM
simulation, and be able to efficiently identify semi-arbitrary,
planar antenna geometries that meet performance specifications.
GA-MOM is an elegant formulation that requires the computationally
expensive impedance matrix to be determined only once, with the
majority of the optimization process falling to efficient GA
routines. |
| |
