Biometric identification

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.

In this study, we investigated what informational aspects of faces could account
for the ability to match an individual’s face to their voice, using only static images. In
each of the first six experiments, we simultaneously presented one voice recording along
with two manipulated images of faces (e.g. top half of the face, bottom half of the face,
etc.), a target face and distractor face. The participant’s task was to choose which of the
images they thought belonged to the same individual as the voice recording. The voices
remained un-manipulated. In Experiment 7 we used eye tracking in order to determine
which informational aspects of the model’s faces people are fixating while performing
the matching task, as compared to where they fixate when there are no immediate task
demands. We presented a voice recording followed by two static images, a target and
distractor face. The participant’s task was to choose which of the images they thought
belonged to the same individual as the voice recording, while we tracked their total
fixation duration. In the no-task, passive viewing condition, we presented a male’s voice
recording followed sequentially by two static images of female models, or vice versa, counterbalanced across participants. Participant’s results revealed significantly better
than chance performance in the matching task when the images presented were the
bottom half of the face, the top half of the face, the images inverted upside down, when
presented with a low pass filtered image of the face, and when the inner face was
completely blurred out. In Experiment 7 we found that when completing the matching
task, the time spent looking at the outer area of the face increased, as compared to when
the images and voice recordings were passively viewed. When the images were passively
viewed, the time spend looking at the inner area of the face increased. We concluded that
the inner facial features (i.e. eyes, nose, and mouth) are not necessary informational
aspects of the face which allow for the matching ability. The ability likely relies on global
features such as the face shape and size.

Appealing features of cloud services such as elasticity, scalability, universal access, low entry cost, and flexible billing motivate consumers to migrate their core businesses into the cloud. However, there are challenges about security, privacy, and compliance. Building compliant systems is difficult because of the complex nature of regulations and cloud systems. In addition, the lack of complete, precise, vendor neutral, and platform independent software architectures makes compliance even harder. We have attempted to make regulations clearer and more precise with patterns and reference architectures (RAs). We have analyzed regulation policies, identified overlaps, and abstracted them as patterns to build compliant RAs. RAs should be complete, precise, abstract, vendor neutral, platform independent, and with no implementation details; however, their levels of detail and abstraction are still debatable and there is no commonly accepted definition about what an RA should contain. Existing approaches to build RAs lack structured templates and systematic procedures. In addition, most approaches do not take full advantage of patterns and best practices that promote architectural quality. We have developed a five-step approach by analyzing features from available approaches but refined and combined them in a new way. We consider an RA as a big compound pattern that can improve the quality of the concrete architectures derived from it and from which we can derive more specialized RAs for cloud systems. We have built an RA for HIPAA, a compliance RA (CRA), and a specialized compliance and security RA (CSRA) for cloud systems. These RAs take advantage of patterns and best practices that promote software quality. We evaluated the architecture by creating profiles. The proposed approach can be used to build RAs from scratch or to build new RAs by abstracting real RAs for a given context. We have also described an RA itself as a compound pattern by using a modified POSA template. Finally, we have built a concrete deployment and availability architecture derived from CSRA that can be used as a foundation to build compliance systems in the cloud.

Underwater vehicles often use acoustics or dead reckoning for global positioning, which is impractical for low cost, high proximity applications. An optical based positional feedback system for a wave tank operated biomimetic station-keeping vehicle was made using an extended Kalman filter and a model of a nearby light source. After physical light model verification, the filter estimated surge, sway, and heading with 6 irradiance sensors and a low cost inertial measurement unit (~$15). Physical testing with video feedback suggests an average error of ~2cm in surge and sway, and ~3deg in yaw, over a 1200 cm2 operational area. This is 2-3 times better, and more consistent, than adaptations of prior art tested alongside the extended Kalman filter feedback system. The physical performance of the biomimetic platform was also tested. It has a repeatable forward velocity response with a max of 0.3 m/s and fair stability in surface testing conditions.

Continuous changes in the software development community require challenging conventional approaches resulting in techniques that allow for early decisions at the design level. This project is a demonstration of the use of design patterns as a common way to organize objects to make practical design decisions helping to generate flexible, manageable and agile software architectures. Due to the continuity and unpredictability of its requirements, the Biometric Industry is appropriate to illustrate of the use of design patterns and object oriented analysis. First, the conceptual model of an Electronic Fingerprint Service establishes the vocabulary for discussing how a system is constructed. Since good design decisions eventually result in a good design model, this model is used to leverage the object reused when requirements change. The Electronic Biometric Services model demonstrates how by applying design patterns, the system can gain the flexibility and agility required to grow and change according to new requirements.

Two experiments were conducted to investigate the impact of poser and perceiver gender on the Happiness/Anger Superiority effect and the Female Advantage in facial expression recognition. Happy, neutral, and angry facial expressions were presented on male and female faces under Continuous Flash Suppression (CFS). Participants of both genders indicated when the presented faces broke through the suppression. In the second experiment, angry and happy expressions were reduced to 50% intensity. At full intensity, there was no difference in the reaction time for female neutral and angry faces, but male faces showed a difference in detection between all expressions. Across experiments, male faces were detected later than female faces for all facial expressions. Happiness was generally detected faster than anger, except when on female faces at 50% intensity. No main effect for perceiver gender emerged. It was concluded that happiness is superior to anger in CFS, and that poser gender affects facial expression recognition.

The literature regarding biological motion suggests that people may accurately identify and recognize the gender of others using movement cues in the absence of typical identifiers. This study compared identification and gender judgments of traditional point-light stimuli to skeleton stimuli. Controlling for previous experience and execution of actions, the frequency and familiarity of movements was also considered. Watching action clips, participants learned to identify 4 male and 4 female actors. Participants then identified the corresponding point-light or skeleton displays. Although results indicate higher than chance performance, no difference was observed between stimuli conditions. Analyses did show better gender recognition for common as well as previously viewed actions. This suggests that visual experience influences extraction and application of biological motion. Thus insufficient practice in relying on movement cues for identification could explain the significant yet poor performance in biological motion point-light research.

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.

Video identification or copy detection is a challenging problem and is becoming increasingly important with the popularity of online video services. The problem addressed in this thesis is the identification of a given video clip in a given set of videos. For a given query video, the system returns all the instance of the video in the data set. This identification system uses video signatures based on video tomography. A robust and low complexity video signature is designed and implemented. The nature of the signature makes it independent to the most commonly video transformations. The signatures are generated for video shots and not individual frames, resulting in a compact signature of 64 bytes per video shot. The signatures are matched using simple Euclidean distance metric. The results show that videos can be identified with 100% recall and over 93% precision. The experiments included several transformations on videos.

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.