Speech Enhancement using Intra-frame Dependency in DCT Domain

In this paper, we present a new speech enhancement approach, that is based on exploiting the intra-frame dependency of discrete cosine transform (DCT) domain coefficients. It can be noted that the existing enhancement techniques treat the transformdomain coefficients independently. Instead of this traditional approach of independently processing the scalars, we split the DCT domain noisy speech vector into sub-vectors and each sub-vector is enhanced independently. Through this sub-vector based approach, the higher dimensional enhancement advantage, viz. non-linear dependency, is exploited. In the developed method, each clean speech sub-vector is modeled using a Gaussian mixture (GM) density. We show that the proposed Gaussian mixture model (GMM) based DCT domain method, using sub-vector processing approach, provides better performance than the conventional approach of enhancing the transform domain scalar components independently. Performance improvement over the recently proposed GMM based time domain approach is also shown.

GMM basedbayesian approach to speech enhancement in signal/transform domain

Considering a general linear model of signal degradation, by modeling the probability density function (PDF) of the clean signal using a Gaussian mixture model (GMM) and additive noise by a Gaussian PDF, we derive the minimum mean square error (MMSE) estimator.The derived MMSE estimator is non-linear and the linear MMSE estimator is shown to be a special case. For speech signal corrupted by independent additive noise, by modeling the joint PDF of time-domain speech samples of a speech frame using a GMM, we propose a speech enhancement method based on the derived MMSE estimator. We also show that the same estimator can be used for transform-domain speech enhancement.

Semi-supervised SVMs for classification with unknown class proportions and a small labeled dataset

In the design of practical web page classification systems one often encounters a situation in which the labeled training set is created by choosing some examples from each class; but, the class proportions in this set are not the same as those in the test distribution to which the classifier will be actually applied. The problem is made worse when the amount of training data is also small. In this paper we explore and adapt binary SVM methods that make use of unlabeled data from the test distribution, viz., Transductive SVMs (TSVMs) and expectation regularization/constraint (ER/EC) methods to deal with this situation. We empirically show that when the labeled training data is small, TSVM designed using the class ratio tuned by minimizing the loss on the labeled set yields the best performance; its performance is good even when the deviation between the class ratios of the labeled training set and the test set is quite large. When the labeled training data is sufficiently large, an unsupervised Gaussian mixture model can be used to get a very good estimate of the class ratio in the test set; also, when this estimate is used, both TSVM and EC/ER give their best possible performance, with TSVM coming out superior. The ideas in the paper can be easily extended to multi-class SVMs and MaxEnt models.

Voice source characterization using pitch synchronous discrete cosine transform for speaker identification

A characterization of the voice source (VS) signal by the pitch synchronous (PS) discrete cosine transform (DCT) is proposed. With the integrated linear prediction residual (ILPR) as the VS estimate, the PS DCT of the ILPR is evaluated as a feature vector for speaker identification (SID). On TIMIT and YOHO databases, using a Gaussian mixture model (GMM)-based classifier, it performs on par with existing VS-based features. On the NIST 2003 database, fusion with a GMM-based classifier using MFCC features improves the identification accuracy by 12% in absolute terms, proving that the proposed characterization has good promise as a feature for SID studies. (C) 2015 Acoustical Society of America

Structured precision modelling with Cholesky basis superposition for speech recognition

Structured precision modelling is an important approach to improve the intra-frame correlation modelling of the standard HMM, where Gaussian mixture model with diagonal covariance are used. Previous work has all been focused on direct structured representation of the precision matrices. In this paper, a new framework is proposed, where the structure of the Cholesky square root of the precision matrix is investigated, referred to as Cholesky Basis Superposition (CBS). Each Cholesky matrix associated with a particular Gaussian distribution is represented as a linear combination of a set of Gaussian independent basis upper-triangular matrices. Efficient optimization methods are derived for both combination weights and basis matrices. Experiments on a Chinese dictation task showed that the proposed approach can significantly outperformed the direct structured precision modelling with similar number of parameters as well as full covariance modelling. © 2011 IEEE.

Single camera pose estimation using Bayesian filtering and Kinect motion priors

Traditional approaches to upper body pose estimation using monocular vision rely on complex body models and a large variety of geometric constraints. We argue that this is not ideal and somewhat inelegant as it results in large processing burdens, and instead attempt to incorporate these constraints through priors obtained directly from training data. A prior distribution covering the probability of a human pose occurring is used to incorporate likely human poses. This distribution is obtained offline, by fitting a Gaussian mixture model to a large dataset of recorded human body poses, tracked using a Kinect sensor. We combine this prior information with a random walk transition model to obtain an upper body model, suitable for use within a recursive Bayesian filtering framework. Our model can be viewed as a mixture of discrete Ornstein-Uhlenbeck processes, in that states behave as random walks, but drift towards a set of typically observed poses. This model is combined with measurements of the human head and hand positions, using recursive Bayesian estimation to incorporate temporal information. Measurements are obtained using face detection and a simple skin colour hand detector, trained using the detected face. The suggested model is designed with analytical tractability in mind and we show that the pose tracking can be Rao-Blackwellised using the mixture Kalman filter, allowing for computational efficiency while still incorporating bio-mechanical properties of the upper body. In addition, the use of the proposed upper body model allows reliable three-dimensional pose estimates to be obtained indirectly for a number of joints that are often difficult to detect using traditional object recognition strategies. Comparisons with Kinect sensor results and the state of the art in 2D pose estimation highlight the efficacy of the proposed approach.

Partially Supervised Neighbor Embedding for Example-Based Image Super-Resolution

Neighbor embedding algorithm has been widely used in example-based super-resolution reconstruction from a single frame, which makes the assumption that neighbor patches embedded are contained in a single manifold. However, it is not always true for complicated texture structure. In this paper, we believe that textures may be contained in multiple manifolds, corresponding to classes. Under this assumption, we present a novel example-based image super-resolution reconstruction algorithm with clustering and supervised neighbor embedding (CSNE). First, a class predictor for low-resolution (LR) patches is learnt by an unsupervised Gaussian mixture model. Then by utilizing class label information of each patch, a supervised neighbor embedding is used to estimate high-resolution (HR) patches corresponding to LR patches. The experimental results show that the proposed method can achieve a better recovery of LR comparing with other simple schemes using neighbor embedding.

Investigation into DCT Feature Selection for Visual Lip-Based Biometric Authentication

This paper investigated using lip movements as a behavioural biometric for person authentication. The system was trained, evaluated and tested using the XM2VTS dataset, following the Lausanne Protocol configuration II. Features were selected from the DCT coefficients of the greyscale lip image. This paper investigated the number of DCT coefficients selected, the selection process, and static and dynamic feature combinations. Using a Gaussian Mixture Model - Universal Background Model framework an Equal Error Rate of 2.20% was achieved during evaluation and on an unseen test set a False Acceptance Rate of 1.7% and False Rejection Rate of 3.0% was achieved. This compares favourably with face authentication results on the same dataset whilst not being susceptible to spoofing attacks.

Agrupamento multi-orador

Actualmente tem-se observado um aumento do volume de sinais de fala em diversas aplicações, que reforçam a necessidade de um processamento automático dos ficheiros. No campo do processamento automático destacam-se as aplicações de “diarização de orador”, que permitem catalogar os ficheiros de fala com a identidade de oradores e limites temporais de fala de cada um, através de um processo de segmentação e agrupamento. No contexto de agrupamento, este trabalho visa dar continuidade ao trabalho intitulado “Detecção do Orador”, com o desenvolvimento de um algoritmo de “agrupamento multi-orador” capaz de identificar e agrupar correctamente os oradores, sem conhecimento prévio do número ou da identidade dos oradores presentes no ficheiro de fala. O sistema utiliza os coeficientes “Mel Line Spectrum Frequencies” (MLSF) como característica acústica de fala, uma segmentação de fala baseada na energia e uma estrutura do tipo “Universal Background Model - Gaussian Mixture Model” (UBM-GMM) adaptado com o classificador “Support Vector Machine” (SVM). No trabalho foram analisadas três métricas de discriminação dos modelos SVM e a avaliação dos resultados foi feita através da taxa de erro “Speaker Error Rate” (SER), que quantifica percentualmente o número de segmentos “fala” mal classificados. O algoritmo implementado foi ajustado às características da língua portuguesa através de um corpus com 14 ficheiros de treino e 30 ficheiros de teste. Os ficheiros de treino dos modelos e classificação final, enquanto os ficheiros de foram utilizados para avaliar o desempenho do algoritmo. A interacção com o algoritmo foi dinamizada com a criação de uma interface gráfica que permite receber o ficheiro de teste, processá-lo, listar os resultados ou gerar um vídeo para o utilizador confrontar o sinal de fala com os resultados de classificação.

Vidéosurveillance intelligente pour la détection de chutes chez les personnes âgées

Les pays industrialisés comme le Canada doivent faire face au vieillissement de leur population. En particulier, la majorité des personnes âgées, vivant à domicile et souvent seules, font face à des situations à risques telles que des chutes. Dans ce contexte, la vidéosurveillance est une solution innovante qui peut leur permettre de vivre normalement dans un environnement sécurisé. L’idée serait de placer un réseau de caméras dans l’appartement de la personne pour détecter automatiquement une chute. En cas de problème, un message pourrait être envoyé suivant l’urgence aux secours ou à la famille via une connexion internet sécurisée. Pour un système bas coût, nous avons limité le nombre de caméras à une seule par pièce ce qui nous a poussé à explorer les méthodes monoculaires de détection de chutes. Nous avons d’abord exploré le problème d’un point de vue 2D (image) en nous intéressant aux changements importants de la silhouette de la personne lors d’une chute. Les données d’activités normales d’une personne âgée ont été modélisées par un mélange de gaussiennes nous permettant de détecter tout événement anormal. Notre méthode a été validée à l’aide d’une vidéothèque de chutes simulées et d’activités normales réalistes. Cependant, une information 3D telle que la localisation de la personne par rapport à son environnement peut être très intéressante pour un système d’analyse de comportement. Bien qu’il soit préférable d’utiliser un système multi-caméras pour obtenir une information 3D, nous avons prouvé qu’avec une seule caméra calibrée, il était possible de localiser une personne dans son environnement grâce à sa tête. Concrêtement, la tête de la personne, modélisée par une ellipsoide, est suivie dans la séquence d’images à l’aide d’un filtre à particules. La précision de la localisation 3D de la tête a été évaluée avec une bibliothèque de séquence vidéos contenant les vraies localisations 3D obtenues par un système de capture de mouvement (Motion Capture). Un exemple d’application utilisant la trajectoire 3D de la tête est proposée dans le cadre de la détection de chutes. En conclusion, un système de vidéosurveillance pour la détection de chutes avec une seule caméra par pièce est parfaitement envisageable. Pour réduire au maximum les risques de fausses alarmes, une méthode hybride combinant des informations 2D et 3D pourrait être envisagée.

Développement d’un modèle de classification probabiliste pour la cartographie du couvert nival dans les bassins versants d’Hydro-Québec à l’aide de données de micro-ondes passives

Chaque jour, des décisions doivent être prises quant à la quantité d'hydroélectricité produite au Québec. Ces décisions reposent sur la prévision des apports en eau dans les bassins versants produite à l'aide de modèles hydrologiques. Ces modèles prennent en compte plusieurs facteurs, dont notamment la présence ou l'absence de neige au sol. Cette information est primordiale durant la fonte printanière pour anticiper les apports à venir, puisqu'entre 30 et 40% du volume de crue peut provenir de la fonte du couvert nival. Il est donc nécessaire pour les prévisionnistes de pouvoir suivre l'évolution du couvert de neige de façon quotidienne afin d'ajuster leurs prévisions selon le phénomène de fonte. Des méthodes pour cartographier la neige au sol sont actuellement utilisées à l'Institut de recherche d'Hydro-Québec (IREQ), mais elles présentent quelques lacunes. Ce mémoire a pour objectif d'utiliser des données de télédétection en micro-ondes passives (le gradient de températures de brillance en position verticale (GTV)) à l'aide d'une approche statistique afin de produire des cartes neige/non-neige et d'en quantifier l'incertitude de classification. Pour ce faire, le GTV a été utilisé afin de calculer une probabilité de neige quotidienne via les mélanges de lois normales selon la statistique bayésienne. Par la suite, ces probabilités ont été modélisées à l'aide de la régression linéaire sur les logits et des cartographies du couvert nival ont été produites. Les résultats des modèles ont été validés qualitativement et quantitativement, puis leur intégration à Hydro-Québec a été discutée.

Generative topographic mapping applied to clustering and visualization of motor unit action potentials

The identification and visualization of clusters formed by motor unit action potentials (MUAPs) is an essential step in investigations seeking to explain the control of the neuromuscular system. This work introduces the generative topographic mapping (GTM), a novel machine learning tool, for clustering of MUAPs, and also it extends the GTM technique to provide a way of visualizing MUAPs. The performance of GTM was compared to that of three other clustering methods: the self-organizing map (SOM), a Gaussian mixture model (GMM), and the neural-gas network (NGN). The results, based on the study of experimental MUAPs, showed that the rate of success of both GTM and SOM outperformed that of GMM and NGN, and also that GTM may in practice be used as a principled alternative to the SOM in the study of MUAPs. A visualization tool, which we called GTM grid, was devised for visualization of MUAPs lying in a high-dimensional space. The visualization provided by the GTM grid was compared to that obtained from principal component analysis (PCA). (c) 2005 Elsevier Ireland Ltd. All rights reserved.

Corpus-Based Speech Enhancement With Uncertainty Modeling and Cepstral Smoothing

We present a new approach for corpus-based speech enhancement that significantly improves over a method published by Xiao and Nickel in 2010. Corpus-based enhancement systems do not merely filter an incoming noisy signal, but resynthesize its speech content via an inventory of pre-recorded clean signals. The goal of the procedure is to perceptually improve the sound of speech signals in background noise. The proposed new method modifies Xiao's method in four significant ways. Firstly, it employs a Gaussian mixture model (GMM) instead of a vector quantizer in the phoneme recognition front-end. Secondly, the state decoding of the recognition stage is supported with an uncertainty modeling technique. With the GMM and the uncertainty modeling it is possible to eliminate the need for noise dependent system training. Thirdly, the post-processing of the original method via sinusoidal modeling is replaced with a powerful cepstral smoothing operation. And lastly, due to the improvements of these modifications, it is possible to extend the operational bandwidth of the procedure from 4 kHz to 8 kHz. The performance of the proposed method was evaluated across different noise types and different signal-to-noise ratios. The new method was able to significantly outperform traditional methods, including the one by Xiao and Nickel, in terms of PESQ scores and other objective quality measures. Results of subjective CMOS tests over a smaller set of test samples support our claims.

A Pseudolikelihood Approach for Simultaneous Analysis of Array Comparative Genomic Hybridizations (aCGH)

DNA sequence copy number has been shown to be associated with cancer development and progression. Array-based Comparative Genomic Hybridization (aCGH) is a recent development that seeks to identify the copy number ratio at large numbers of markers across the genome. Due to experimental and biological variations across chromosomes and across hybridizations, current methods are limited to analyses of single chromosomes. We propose a more powerful approach that borrows strength across chromosomes and across hybridizations. We assume a Gaussian mixture model, with a hidden Markov dependence structure, and with random effects to allow for intertumoral variation, as well as intratumoral clonal variation. For ease of computation, we base estimation on a pseudolikelihood function. The method produces quantitative assessments of the likelihood of genetic alterations at each clone, along with a graphical display for simple visual interpretation. We assess the characteristics of the method through simulation studies and through analysis of a brain tumor aCGH data set. We show that the pseudolikelihood approach is superior to existing methods both in detecting small regions of copy number alteration and in accurately classifying regions of change when intratumoral clonal variation is present.

First year qualifying report: neural networks for extracting wind vectors from satellite scatterometer data

The ERS-1 Satellite was launched in July 1991 by the European Space Agency into a polar orbit at about km800, carrying a C-band scatterometer. A scatterometer measures the amount of radar back scatter generated by small ripples on the ocean surface induced by instantaneous local winds. Operational methods that extract wind vectors from satellite scatterometer data are based on the local inversion of a forward model, mapping scatterometer observations to wind vectors, by the minimisation of a cost function in the scatterometer measurement space.par This report uses mixture density networks, a principled method for modelling conditional probability density functions, to model the joint probability distribution of the wind vectors given the satellite scatterometer measurements in a single cell (the `inverse' problem). The complexity of the mapping and the structure of the conditional probability density function are investigated by varying the number of units in the hidden layer of the multi-layer perceptron and the number of kernels in the Gaussian mixture model of the mixture density network respectively. The optimal model for networks trained per trace has twenty hidden units and four kernels. Further investigation shows that models trained with incidence angle as an input have results comparable to those models trained by trace. A hybrid mixture density network that incorporates geophysical knowledge of the problem confirms other results that the conditional probability distribution is dominantly bimodal.par The wind retrieval results improve on previous work at Aston, but do not match other neural network techniques that use spatial information in the inputs, which is to be expected given the ambiguity of the inverse problem. Current work uses the local inverse model for autonomous ambiguity removal in a principled Bayesian framework. Future directions in which these models may be improved are given.