Pattern recognition systems

The focus of this research is on images extracted from surveillance videos that
have a low resolution and are taken under low illumination. In recent years, great
advances have been made in face recognition and many studies mention results of 80%
and 90% of recognition efficiency, however, most of these studies reported results using
face images under controlled conditions. Current surveillance systems are equipped with
low resolution cameras and are located in places with changing illumination, as opposed
to a controlled environment. To be used in face recognition, images extracted from
videos need to be normalized, enlarged and preprocessed. There is a multitude of
processing algorithms for image enhancement, and each algorithm faces its advantages
and disadvantages. This thesis presents a novel method for image enlargement of human
faces applied to low quality video recordings. Results and comparison to traditional
methods are also presented.

For many years people have consider the possibility that brain activity might provide
a new channel for communication between a person's brain and the external world.
Brain Computer Interface allows humans to control electronic devices using only
their thoughts. The goal of this project is to provide the users with a basic control of a
prosthetic arm using the signal acquired by an Electroencephalogram (EEG). The
main objective of the research is to demonstrate and provide a system that allows
individuals to obtain control of the device with very little training and very few
electrodes. The research includes the development of an elaborate signal-processing
algorithm that uses an Artificial Neural Network to determine the intentions of the
user and their translation into commands to operate the prosthetic arm.

A variety of classifiers for solving classification problems is available from
the domain of machine learning. Commonly used classifiers include support vector
machines, decision trees and neural networks. These classifiers can be configured
by modifying internal parameters. The large number of available classifiers and
the different configuration possibilities result in a large number of combinatiorrs of
classifier and configuration settings, leaving the practitioner with the problem of
evaluating the performance of different classifiers. This problem can be solved by
using performance metrics. However, the large number of available metrics causes
difficulty in deciding which metrics to use and when comparing classifiers on the
basis of multiple metrics. This paper uses the statistical method of factor analysis
in order to investigate the relationships between several performance metrics and
introduces the concept of relative performance which has the potential to case the
process of comparing several classifiers. The relative performance metric is also
used to evaluate different support vector machine classifiers and to determine if the
default settings in the Weka data mining tool are reasonable.

Most of the human visual field falls in the periphery, and peripheral processing is
important for normal visual functioning. Yet, little is known about peripheral object
recognition in naturalistic scenes and factors that modulate this ability. We propose that
a critical function of scene and object memory is in order to facilitate visual object
recognition in the periphery. In the first experiment, participants identified objects in
scenes across different levels of familiarity and contextual information within the scene.
We found that familiarity with a scene resulted in a significant increase in the distance
that objects were recognized. Furthermore, we found that a semantically consistent scene
improved the distance that object recognition is possible, supporting the notion that
contextual facilitation is possible in the periphery. In the second experiment, the preview
duration of a scene was varied in order to examine how a scene representation is built and
how memory of that scene and the objects within it contributes to object recognition in
the periphery. We found that the closer participants fixated to the object in the preview,
the farther on average they recognized that target object in the periphery. However, only a preview duration of the scenes for 5000 ms produced significantly farther peripheral
object recognition compared to not previewing the scene. Overall, these experiments
introduce a novel research paradigm for object recognition in naturalistic scenes, and
demonstrates multiple factors that have systematic effects on peripheral object
recognition.

How does spatial organization of objects affect the perceptual processing of a scene? Surprisingly, little research has explored this topic. A few studies have reported that, when simple, homogenous stimuli (e.g., dots), are presented in a regular formation, they are judged to be more numerous than when presented in a random configuration (Ginsburg, 1976; 1978). However, these results may not apply to real-world objects. In the current study, fewer objects were believed to be on organized desks than their disorganized equivalents. Objects that are organized may be more likely to become integrated, due to classic Gestalt principles. Consequently, visual search may be more difficult. Such object integration may diminish saliency, making objects less apparent and more difficult to find. This could explain why, in the present study, objects on disorganized desks were found faster.

The present research is a targeted endeavor to study the underlying characteristics and novel applications of millimeter (mm) wave through terahertz (THz) spectrum of electromagnetic (EM) energy. Focused thereof are the following specific tasks broadly considered pertinent to the said EM spectral range: (i) To elucidate the material characteristics vis-à-vis the interaction with EM energy at the test frequencies; (ii) to identify biomedical applications based on the material characteristics studied and applied to biomedia; and (iii) to model the wireless communication channels supporting EM waves at the test frequency bands of interest. Commensurate with the scope as above, the objectives of the research are as follows:

Software simulations of a scaleable VLSI implementable architecture and algorithm for character recognition by a research group at Florida Atlantic University (FAU) have shown encouraging results. We address here hardware implementation issues pertinent to the classification phase of character recognition. Using the digit classification techniques developed at FAU as a foundation, we have designed and simulated general purpose building blocks useful for a possible implementation of a Digital & Analog CMOS VLSI chip that is suitable for a variety of artificial neural network (ANN) architectures. HSPICE was used to perform circuit-level simulations of the building blocks. We present here the details of implementation of the recognition chip including the architecture, circuit design and the simulation results.

In this thesis we report a VLSI design implementation of an application specific, full-frame architecture CCD image sensor for a handwritten Optical Character Recognition system. The design is targeted to the MOSIS 2mu, 2-poly/ 2-metal n-buried channel CCD/CMOS technology. The front side illuminated CCD image sensor uses a transparent polysilicon gate structure and is comprised of 84 (H) x 100 (V) pixels arranged in a hexagonal lattice structure. The sensor has unit pixel dimensions of 18 lambda (H) x 16 lambda (V). A second layer of metal is used for shielding certain areas from incident light, and the effective pixel photosite area is 8 lambda x 8 lambda. The imaging pixels use a 3-phase structure (with an innovative addressing scheme for the hexagonal lattice) for image sensing and horizontal charge shift. Columns of charge are shifted into the vertical 2-phase CCD shift registers, which shift the charge out serially at high speed. The chip has been laid out on the 'tinychip' (2250 mu m x 2220 (mu m) pad frame and fabrication through MOSIS is planned next.

This thesis deals with the recognition of digitized handprinting characters. Digitized character images are thresholded, binarized and converted into 32 x 32 matrices. The binarized character matrices are preprocessed to remove noise and thin down to one pixel per linewidth. For dominant features, namely, (1) number of loops, (2) number of end-pixels, (3) number of 3-branch-pixels, and (4) number of 4-branch-pixels, are used as criteria to pre-classify characters into 14 groups. Characters belonging to larger groups are encoded into chain code and compiled into a data base. Recognition of characters belonging to larger groups is achieved by data base look-up and or decision tree tests if ambiguities occur in the data base entries. Recognition of characters belonging to the smaller groups is doned by decision tree tests.

A study was made on the feasibility of the syntactic
approach to the problem of hand printed character
recognition. The characters are represented as postfix expressions in
Picture Description Language. By comparing them with the
prototype expressions, each character is classified as the
prototype that is closest to it. Programs written in the Pascal language, which generate
the postfix expressions for the characters, and recognize
the characters, are presented.