Pattern recognition systems

A social network is a structure of individuals and organizations, which are connected by one or more types of interdependency, such as friendship, affinity, common interests or knowledge. Social networks use Web 2.0 technology, which is mostly based on a service-oriented architecture. We are studying patterns for social networks in this environment. A pattern is an encapsulated solution to a software problem in a given context, secure threats are possible in this context. We present a collection of patterns associated with the most important aspects of social networks, with emphasis on controlling the actions of the users of these networks.

This dissertation introduces our work on face recognition using a novel approach based on creating 3D face model from 2D face images. Together with the pose angle estimation and illumination compensation, this method can be used successfully to recognize 2D faces with 3D recognition algorithms. The results reported here were obtained partially with our own face image database, which had 2D and 3D face images of 50 subjects, with 9 different pose angles. It is shown that by applying even the simple PCA algorithm, this new approach can yield successful recognition rates using 2D probing images and 3D gallery images. The insight gained from the 2D/3D face recognition study was also extended to the case of involving 2D probing and 2D gallery images, which offers a more flexible approach since it is much easier and practical to acquire 2D photos for recognition. To test the effectiveness of the proposed approach, the public AT&T face database, which had 2D only face photos of 40 subjects, with 10 different images each, was utilized in the experimental study. The results from this investigation show that with our approach, the 3D recognition algorithm can be successfully applied to 2D only images. The performance of the proposed approach was further compared with some of the existing face recognition techniques. Studies on imperfect conditions such as domain and pose/illumination variations were also carried out. Additionally, the performance of the algorithms on noisy photos was evaluated. Pros and cons of the proposed face recognition technique along with suggestions for future studies are also given in the dissertation.

Sustained resonance in a linear oscillator is achievable with a drive whose constant frequency matches the resonant frequency of the oscillator. In oscillators with nonlinear restoring forces, i.e., Dung-type oscillators, resonant frequency changes with amplitude, so a constant frequency drive generates a beat oscillation instead of sustained resonance. Dung-type oscillators can be driven into sustained resonance, called autoresonance (AR), when drive frequency is swept in time to match the changing resonant frequency of the oscillator. It is found that near-optimal drive linear sweep rates for autoresonance can be estimated from the beat oscillation resulting from constant frequency excitation. Specically, a least squares estimate of the slope of the Teager-Kaiser instantaneous frequency versus time plot for the rising half-cycle of the beat response to a stationary drive provides a near-optimal estimate of the linear drive sweep rate that sustains resonance in the pendulum, Dung and Dung-Van der Pol oscillators. These predictions are confirmed with model-based numerical simulations. A closed-form approximation to the AM-FM nonlinear resonance beat response of a Dung oscillator driven at its low-amplitude oscillator frequency is obtained from a solution to an associated Mathieu equation. AR time responses are found to evolve along a Mathieu equation primary resonance stability boundary. AR breakdown occurs at sweep rates just past optimal and map to a single stable point just off the Mathieu equation primary resonance stability boundary. Optimal AR sweep rates produce oscillating phase dierences with extrema near 90 degrees, allowing extended time in resonance. AR breakdown occurs when phase difference equals 180 degrees. Nonlinear resonance of the van der Pol type may play a role in the extraordinary sensitivity of the human ear.

During the Leukocytapheresis (LCAP) process used to treat patients suffering from acute Ulcerative Colitis, medical practitioners have to continuously monitor the Hematocrit (Ht) level in the blood to ensure it is within the acceptable range. The work done, as a part of this thesis, attempts to create an early warning system that can be used to predict if and when the Ht values will deviate from the acceptable range. To do this we have developed an algorithm based on the Group Method of Data Handling (GMDH) and compared it to other Neural Network algorithms, in particular the Multi Layer Perceptron (MLP). The standard GMDH algorithm captures the fluctuation very well but there is a time lag that produces larger errors when compared to MLP. To address this drawback we modified the GMDH algorithm to reduce the prediction error and produce more accurate results.

Design of intelligent systems that can learn from the environment and adapt to the change in the environment has been pursued by many researchers in this age of information technology. The Group Method of Data Handling (GMDH) algorithm to be implemented is a multilayered neural network. Neural network consists of neurons which use information acquired in training to deduce relationships in order to predict future responses. Most software tool during the simulation of the neural network based algorithms in a sequential, single processor machine like Pascal, C or C++ takes several hours or even days. But in this thesis, the GMDH algorithm was modified and implemented into a software tool written in Verilog HDL and tested with specific application (XOR) to make the simulation faster. The purpose of the development of this tool is also to keep it general enough so that it can have a wide range of uses, but robust enough that it can give accurate results for all of those uses. Most of the applications of neural networks are basically software simulations of the algorithms only but in this thesis the hardware design is also developed of the algorithm so that it can be easily implemented on hardware using Field Programmable Gate Array (FPGA) type devices. The design is small enough to require a minimum amount of memory, circuit space, and propagation delay.

Video signature techniques based on tomography images address the problem of video identification. This method relies on temporal segmentation and sampling strategies to build and determine the unique elements that will form the signature. In this thesis an extension for these methods is presented; first a new feature extraction method, derived from the previously proposed sampling pattern, is implemented and tested, resulting in a highly distinctive set of signature elements, second a robust temporal video segmentation system is used to replace the original method applied to determine shot changes more accurately. Under a very exhaustive set of tests the system was able to achieve 99.58% of recall, 100% of precision and 99.35% of prediction precision.

In recent years there has been great interest in implementing object recognition frame work on mobile phones. This has stemmed from the fact the advances in object recognition algorithm and mobile phone capabilities have built a congenial ecosystem. Application developers on mobile platforms are trying to utilize the object recognition technology to build better human computer interfaces. This approach is in the nascent phase and proper application framework is required. In this thesis, we propose a framework to overcome design challenges and provide an evaluation methodology to assess the system performance. We use the emerging Android mobile platform to implement and test the framework. We performed a case study using the proposal and reported the test result. This assessment will help developers make wise decisions about their application design. Furthermore, the Android API developers could use this information to provide better interfaces to the third party developers. The design and evaluation methodology could be extended to other mobile platforms for a wider consumer base.

This study examines whether forming a single identity is crucial to learning to bind faces and voices, or if people are equally able to do so without tying this information to an identity. To test this, individuals learned paired faces and voices that were in one of three different conditions: True voice, Gender Matched, or Gender Mismatched conditions. Performance was measured in a training phase as well as a test phase, and results show that participants were able to learn more quickly and have higher overall performance for learning in the True Voice and Gender Matched conditions. During the test phase, performance was almost at chance in the Gender Mismatched condition which may mean that learning in the training phase was simply memorization of the pairings for this condition. Results support the hypothesis that learning to bind faces and voices is a process that involves forming a supramodal identity from multisensory learning.

This work explores the process of model-based classification of speech audio signals using low-level feature vectors. The process of extracting low-level features from audio signals is described along with a discussion of established techniques for training and testing mixture model-based classifiers and using these models in conjunction with feature selection algorithms to select optimal feature subsets. The results of a number of classification experiments using a publicly available speech database, the Berlin Database of Emotional Speech, are presented. This includes experiments in optimizing feature extraction parameters and comparing different feature selection results from over 700 candidate feature vectors for the tasks of classifying speaker gender, identity, and emotion. In the experiments, final classification accuracies of 99.5%, 98.0% and 79% were achieved for the gender, identity and emotion tasks respectively.

Biometrics is the science and technology of measuring and analyzing biological data for authentication purposes. Its progress has brought in a large number of civilian and government applications. The candidate modalities used in biometrics include retinas, fingerprints, signatures, audio, faces, etc. There are two types of biometric system: single modal systems and multiple modal systems. Single modal systems perform person recognition based on a single biometric modality and are affected by problems like noisy sensor data, intra-class variations, distinctiveness and non-universality. Applying multiple modal systems that consolidate evidence from multiple biometric modalities can alleviate those problems of single modal ones. Integration of evidence obtained from multiple cues, also known as fusion, is a critical part in multiple modal systems, and it may be consolidated at several levels like feature fusion level, matching score fusion level and decision fusion level. Among biometric modalities, both audio and face modalities are easy to use and generally acceptable by users. Furthermore, the increasing availability and the low cost of audio and visual instruments make it feasible to apply such Audio-Visual (AV) systems for security applications. Therefore, this dissertation proposes an algorithm of face recognition. In addition, it has developed some novel algorithms of fusion in different levels for multiple modal biometrics, which have been tested by a virtual database and proved to be more reliable and robust than systems that rely on a single modality.