A Mandarin speech recognition system for addresses based on MFCC, hidden Markov model (HMM) and Viterbi algorithm is proposed in this thesis. HMM is a doubly stochastic process describing the ways of pronunciation by recording the state transitions according to the time-varing properties of the speech signal. In order to simplify the system design and reduce the computational cost, the mono-syllable structure information in Mandarin is used by incorporating both mono-syllable recognizor and HMM for our system. For the speaker-dependent case, Mandarin address inputting can be accomplished within 60 seconds and 98% correct identification rate can be achieved in the laboratory environment.

A Mandarin speech recognition system for addresses in Taiwan, based on end-point detection, MFCC and HMM, is proposed and implemented in this thesis. It includes both phrase and monosyllable recognition tasks. For the phrase recognition part, we select the initial candidates before the final recognition stage to tremendously reduce the computational time. On the other side, for the monosyllable recognition part, we further refine the recognition details to improve the correct rate under easily confused circumstances. The final system can achieve 85% correct identification rate, and the address recognition can be completed within 2 seconds in the laboratory environment for speaker-dependent case.

Human identification systems are commonly used for security issues. Most of them are based on ID card. However, using an ID card for identification may not be safe enough since people may not have any protection against the theft. Another solution to the identification problem is to use iris or fingerprints. However, systems based on the iris or fingerprints need close interaction to identification machine. Identifying someone from his photograph overcomes all these problems which can be called as face recognition. Common face recognition systems are based on the 2D image recognition but success rates of these methods are strictly depending on the environment. Variations on brightness and pose, complex background are the main problems for 2D image recognition systems. At this point, three dimensional face recognition techniques gain importance. Although there are a lot of methods developed for 3D face recognition, many of them assume that face is not rotated and there is not any destructive (i.e. beard, moustache, hair, hat, and eyeglasses) on the face. However, identification needs to be done though these destructives. Basic step for the face recognition is the determination of the anchor points (i.e. nose tip, inner eye points). In this study, the goal is to implement previously proposed four face recognition methods based on anchor point detection / &ldquo / Multimodal Facial Feature Extraction for Automatic 3D Face Recognition&rdquo / , &ldquo / Automatic Feature Extraction for Multiview 3D Face Recognition&rdquo / , &ldquo / Multiple Nose Region Matching for 3D Face Recognition under Varying Facial Expression&rdquo / , &ldquo / 3D face detection using curvature analysis&rdquo / , to compare the success rates of them for rotated and destructed images and finally to propose improvements on these methods.

For the last several decades, sequential change point problems have been studied in both the theoretical area (sequential analysis) and the application area (industrial SPC). In the conventional application, the baseline process is assumed to be stationary, and the shift pattern is a step function that is sustained after the shift. However, in biosurveillance, the underlying assumptions of problems are more complicated. This thesis investigates several issues in biosurveillance such as non-homogeneous populations, spatiotemporal surveillance methods, and correlated structures in regional data. The first part of the thesis discusses popular surveillance methods in sequential change point problems and off-line problems based on count data. For sequential change point problems, the CUSUM and the EWMA have been used in healthcare and public health surveillance to detect increases in the rates of diseases or symptoms. On the other hand, for off-line problems, scan statistics are widely used. In this chapter, we link the method for off-line problems to those for sequential change point problems. We investigate three methods--the CUSUM, the EWMA, and scan statistics--and compare them by conditional expected delay (CED). The second part of the thesis pertains to the on-line monitoring problem of detecting a change in the mean of Poisson count data with a non-homogeneous population size. The most common detection schemes are based on generalized likelihood ratio statistics, known as an optimal method under Lodern's criteria. We propose alternative detection schemes based on the weighted likelihood ratios and the adaptive threshold method, which perform better than generalized likelihood ratio statistics in an increasing population. The properties of these three detection schemes are investigated by both a theoretical approach and numerical simulation. The third part of the thesis investigates spatiotemporal surveillance based on likelihood ratios. This chapter proposes a general framework for spatiotemporal surveillance based on likelihood ratio statistics over time windows. We show that the CUSUM and other popular likelihood ratio statistics are the special cases under such a general framework. We compare the efficiency of these surveillance methods in spatiotemporal cases for detecting clusters of incidence using both Monte Carlo simulations and a real example. The fourth part proposes multivariate surveillance methods based on likelihood ratio tests in the presence of spatial correlations. By taking advantage of spatial correlations, the proposed methods can perform better than existing surveillance methods by providing the faster and more accurate detection. We illustrate the application of these methods with a breast cancer case in New Hampshire when observations are spatially correlated.

The distribution of asset returns may lead to structural breaks. Thesebreaks may result in changes of the optimal portfolio weights. For a port-folio investor, the ability of timely detection of any systematic changesin the optimal portfolio weights is of a great interest.In this master thesis work, the use of the Shewhart method, as amethod for detecting a sudden parameter change, the implied changein the multivariate portfolio weights and its performance is reviewed.

Electromyogram (EMG) is an established tool to study operation of neuromuscular systems. In analysing EMG signals, accurate detection of the movement-related events in the signal is frequently necessary. I explored the application of change-point detection algorithm proposed by Moskvina et. al., 2003 to EMG event detection, and evaluated the technique’s performance comparing it to two common threshold-based event detection methods and to the visual estimates of the EMG events performed by trained practitioners in the field. The algorithm was implemented in MATLAB and applied to EMG segments recorded from wrist and trunk muscles. The quality and frequency of successful detection were assessed for all methods, using the average visual estimate as the baseline, against which techniques were evaluated. The application showed that the change-point detection can successfully locate multiple changes in the EMG signal, but the maximum value of the detection statistic did not always identify the muscle activation onset.

Electromyogram (EMG) is an established tool to study operation of neuromuscular systems. In analysing EMG signals, accurate detection of the movement-related events in the signal is frequently necessary. I explored the application of change-point detection algorithm proposed by Moskvina et. al., 2003 to EMG event detection, and evaluated the technique’s performance comparing it to two common threshold-based event detection methods and to the visual estimates of the EMG events performed by trained practitioners in the field. The algorithm was implemented in MATLAB and applied to EMG segments recorded from wrist and trunk muscles. The quality and frequency of successful detection were assessed for all methods, using the average visual estimate as the baseline, against which techniques were evaluated. The application showed that the change-point detection can successfully locate multiple changes in the EMG signal, but the maximum value of the detection statistic did not always identify the muscle activation onset.

En orthodontie, le diagnostic et la planification d'un traitement reposent sur la connaissance de l'architecture dentaire du patient relevée, entre autre, par un moulage en plâtre. Aujourd’hui, des logiciels permettent de manipuler des modèles numériques des arcades dentaires obtenus après numérisation des moulages. Afin d’observer l’engrènement des dents, il est nécessaire de mettre en occlusion les deux arcades numérisées séparément. Cette étape est actuellement manuelle et l’objet de ces travaux de thèse est de proposer une chaîne robuste de traitements permettant un recalage automatique des deux arcades guidé par plusieurs photos "en bouche" du patient. L'approche proposée consiste à définir trois types de points singuliers et à mettre en place des méthodes robustes de détection automatique à la fois sur les modèles 3D et les images couleur s’appuyant sur la courbure et la texture. Une fois mis en correspondance, ces points homologues 2D/3D permettent d'estimer les matrices de projection puis la transformation rigide (6ddl) pour positionner au mieux la mandibule par rapport au maxillaire en minimisant les erreurs de reprojection dans plusieurs vues. Afin de s’affranchir du bruit de détection, les positions 2D et/ou 3D des points sont améliorées au cours du processus d’optimisation. De nombreux tests sur des données virtuelles et réelles valident l'approche choisie. L’occlusion finale obtenue par recalage automatique est proche de la référence de l’expert. Les résultats sont encourageants pour fournir une alternative automatique à intégrer dans un outil d'aide au diagnostic. / In orthodontics, the diagnosis and the planning of a treatment rest on the knowledge of the dental architecture of the patient using, among others, a dental cast in plaster. Today, dedicated software allow to manipulate digital models of the dental arches obtained after digitalization of the casts. To observe the contact of teeth, it is necessary to register both arches scanned separately. This stage is at present manual and the object of this thesis research is to propose a robust chain processing allowing an automatic registration of both arches guided by several photos of the patient mouth. The proposed approach consists in defining three types of singular points and in setting up strong methods of automatic detection at the same time on the 3D models and the color images leaning on the curvature and the texture. Once put in correspondence, these 2D / 3D equivalent points allow to estimate the projection matrices then the rigid transformation (6ddl) to position at best the mandible in relation to the maxillary by minimizing the reprojection errors in several views. To free itself from the noise of detection, the 2D and/or 3D positions of the singular points are improved during the optimization process. Numerous tests on virtual and real data validate the proposed approach. The final occlusion obtained on the real data by automatic registration is close to the reference of the expert. These are encouraging results to supply an automatic alternative to be integrated into a help tool for the diagnosis.

This thesis deals with the detection of change in the intercept in panel data re- gression model. We are interested in testing a null hypothesis that there was no change in the intercept during the observation period in case with no depen- dency between panels and with the number of panels and observations in each panel going to infinity. Based on the results for simplified case with no additional regressors we propose a statistical test and show its properties. We also derive a consistent estimate of the parameter of change based on the least squares me- thod. The main contribution of the thesis is the derivation of theoretical results of the proposed test while variances of errors are known and its modification for unknown variance parameters. A large simulation study is conducted to examine the results. Then we present an application to real data, particularly we use four factor CAPM model to detect change in monthly returns of US mutual funds during an observation period 2004-2011 and show a significant change during the sub-prime crisis in 2007-2008. This work expands existing results for de- tecting changes in the mean in panel data and offers many directions for further beneficial research. 1

Change-point detection aims to reveal sudden changes in sequences of data. Special attention has been paid to the detection of abrupt level shifts, and applications of such techniques can be found in a great variety of fields, such as monitoring of climate change, examination of gene expressions and quality control in the manufacturing industry. In this work, we compared the performance of two methods representing frequentist and Bayesian approaches, respectively. The frequentist approach involved a preliminary search for level shifts using a tree algorithm followed by a dynamic programming algorithm for optimizing the locations and sizes of the level shifts. The Bayesian approach involved an MCMC (Markov chain Monte Carlo) implementation of a method originally proposed by Barry and Hartigan. The two approaches were implemented in R and extensive simulations were carried out to assess both their computational efficiency and ability to detect abrupt level shifts. Our study showed that the overall performance regarding the estimated location and size of change-points was comparable for the Bayesian and frequentist approach. However, the Bayesian approach performed better when the number of change-points was small; whereas the frequentist became stronger when the change-point proportion increased. The latter method was also better at detecting simultaneous change-points in vector time series. Theoretically, the Bayesian approach has a lower computational complexity than the frequentist approach, but suitable settings for the combined tree and dynamic programming can greatly reduce the processing time.