Improving automatic detection of obstructive sleep apnea through nonlinear analysis of sustained speech

We present a novel approach for the detection of severe obstructive sleep apnea (OSA) based on patients' voices introducing nonlinear measures to describe sustained speech dynamics. Nonlinear features were combined with state-of-the-art speech recognition systems using statistical modeling techniques (Gaussian mixture models, GMMs) over cepstral parameterization (MFCC) for both continuous and sustained speech. Tests were performed on a database including speech records from both severe OSA and control speakers. A 10 % relative reduction in classification error was obtained for sustained speech when combining MFCC-GMM and nonlinear features, and 33 % when fusing nonlinear features with both sustained and continuous MFCC-GMM. Accuracy reached 88.5 % allowing the system to be used in OSA early detection. Tests showed that nonlinear features and MFCCs are lightly correlated on sustained speech, but uncorrelated on continuous speech. Results also suggest the existence of nonlinear effects in OSA patients' voices, which should be found in continuous speech.

Evaluation of a transplantation algorithm for expressive speech synthesis

When designing human-machine interfaces it is important to consider not only the bare bones functionality but also the ease of use and accessibility it provides. When talking about voice-based inter- faces, it has been proven that imbuing expressiveness into the synthetic voices increases signi?cantly its perceived naturalness, which in the end is very helpful when building user friendly interfaces. This paper proposes an adaptation based expressiveness transplantation system capable of copying the emotions of a source speaker into any desired target speaker with just a few minutes of read speech and without requiring the record- ing of additional expressive data. This system was evaluated through a perceptual test for 3 speakers showing up to an average of 52% emotion recognition rates relative to the natural voice recognition rates, while at the same time keeping good scores in similarity and naturality.

Categorization of Anti-Patterns in Software Project Management

In the last decades, software systems have become an intrinsic element in our daily lives. Software exists in our computers, in our cars, and even in our refrigerators. Today’s world has become heavily dependent on software and yet, we still struggle to deliver quality software products, on-time and within budget. When searching for the causes of such alarming scenario, we find concurrent voices pointing to the role of the project manager. But what is project management and what makes it so challenging? Part of the answer to this question requires a deeper analysis of why software project managers have been largely ineffective. Answering this question might assist current and future software project managers in avoiding, or at least effectively mitigating, problematic scenarios that, if unresolved, will eventually lead to additional failures. This is where anti-patterns come into play and where they can be a useful tool in identifying and addressing software project management failure. Unfortunately, anti-patterns are still a fairly recent concept, and thus, available information is still scarce and loosely organized. This thesis will attempt to help remedy this scenario. The objective of this work is to help organize existing, documented software project management anti-patterns by answering our two research questions: · What are the different anti-patterns in software project management? · How can these anti-patterns be categorized?

New methodology for the integration of biometric features in speaker recognition systems applied to security environments

La cuestión principal abordada en esta tesis doctoral es la mejora de los sistemas biométricos de reconocimiento de personas a partir de la voz, proponiendo el uso de una nueva parametrización, que hemos denominado parametrización biométrica extendida dependiente de género (GDEBP en sus siglas en inglés). No se propone una ruptura completa respecto a los parámetros clásicos sino una nueva forma de utilizarlos y complementarlos. En concreto, proponemos el uso de parámetros diferentes dependiendo del género del locutor, ya que como es bien sabido, la voz masculina y femenina presentan características diferentes que deberán modelarse, por tanto, de diferente manera. Además complementamos los parámetros clásicos utilizados (MFFC extraídos de la señal de voz), con un nuevo conjunto de parámetros extraídos a partir de la deconstrucción de la señal de voz en sus componentes de fuente glótica (más relacionada con el proceso y órganos de fonación y por tanto con características físicas del locutor) y de tracto vocal (más relacionada con la articulación acústica y por tanto con el mensaje emitido). Para verificar la validez de esta propuesta se plantean diversos escenarios, utilizando diferentes bases de datos, para validar que la GDEBP permite generar una descripción más precisa de los locutores que los parámetros MFCC clásicos independientes del género. En concreto se plantean diferentes escenarios de identificación sobre texto restringido y texto independiente utilizando las bases de datos de HESPERIA y ALBAYZIN. El trabajo también se completa con la participación en dos competiciones internacionales de reconocimiento de locutor, NIST SRE (2010 y 2012) y MOBIO 2013. En el primer caso debido a la naturaleza de las bases de datos utilizadas se obtuvieron resultados cercanos al estado del arte, mientras que en el segundo de los casos el sistema presentado obtuvo la mejor tasa de reconocimiento para locutores femeninos. A pesar de que el objetivo principal de esta tesis no es el estudio de sistemas de clasificación, sí ha sido necesario analizar el rendimiento de diferentes sistemas de clasificación, para ver el rendimiento de la parametrización propuesta. En concreto, se ha abordado el uso de sistemas de reconocimiento basados en el paradigma GMM-UBM, supervectores e i-vectors. Los resultados que se presentan confirman que la utilización de características que permitan describir los locutores de manera más precisa es en cierto modo más importante que la elección del sistema de clasificación utilizado por el sistema. En este sentido la parametrización propuesta supone un paso adelante en la mejora de los sistemas de reconocimiento biométrico de personas por la voz, ya que incluso con sistemas de clasificación relativamente simples se consiguen tasas de reconocimiento realmente competitivas. ABSTRACT The main question addressed in this thesis is the improvement of automatic speaker recognition systems, by the introduction of a new front-end module that we have called Gender Dependent Extended Biometric Parameterisation (GDEBP). This front-end do not constitute a complete break with respect to classical parameterisation techniques used in speaker recognition but a new way to obtain these parameters while introducing some complementary ones. Specifically, we propose a gender-dependent parameterisation, since as it is well known male and female voices have different characteristic, and therefore the use of different parameters to model these distinguishing characteristics should provide a better characterisation of speakers. Additionally, we propose the introduction of a new set of biometric parameters extracted from the components which result from the deconstruction of the voice into its glottal source estimate (close related to the phonation process and the involved organs, and therefore the physical characteristics of the speaker) and vocal tract estimate (close related to acoustic articulation and therefore to the spoken message). These biometric parameters constitute a complement to the classical MFCC extracted from the power spectral density of speech as a whole. In order to check the validity of this proposal we establish different practical scenarios, using different databases, so we can conclude that a GDEBP generates a more accurate description of speakers than classical approaches based on gender-independent MFCC. Specifically, we propose scenarios based on text-constrain and text-independent test using HESPERIA and ALBAYZIN databases. This work is also completed with the participation in two international speaker recognition evaluations: NIST SRE (2010 and 2012) and MOBIO 2013, with diverse results. In the first case, due to the nature of the NIST databases, we obtain results closed to state-of-the-art although confirming our hypothesis, whereas in the MOBIO SRE we obtain the best simple system performance for female speakers. Although the study of classification systems is beyond the scope of this thesis, we found it necessary to analise the performance of different classification systems, in order to verify the effect of them on the propose parameterisation. In particular, we have addressed the use of speaker recognition systems based on the GMM-UBM paradigm, supervectors and i-vectors. The presented results confirm that the selection of a set of parameters that allows for a more accurate description of the speakers is as important as the selection of the classification method used by the biometric system. In this sense, the proposed parameterisation constitutes a step forward in improving speaker recognition systems, since even when using relatively simple classification systems, really competitive recognition rates are achieved.

What can we learn from voice therapy of the speaking voice for singers's voice?

Voice therapies of muscle tension dysphonia in Germany need to be increased in effectiveness by applying intensive, manualized procedures and standardized assessment protocols. The same holds true for therapies of disturbed singer's voices. According to a Cochrane review on the effectiveness of therapies of functional dysphonia neither direct nor indirect voice therapies alone but combinations of both elements are effective (Ruotsalainen et al., 2007).

A comparison of open-source segmentation architectures for dealing with imperfect data from the media in speech synthesis

Traditional Text-To-Speech (TTS) systems have been developed using especially-designed non-expressive scripted recordings. In order to develop a new generation of expressive TTS systems in the Simple4All project, real recordings from the media should be used for training new voices with a whole new range of speaking styles. However, for processing this more spontaneous material, the new systems must be able to deal with imperfect data (multi-speaker recordings, background and foreground music and noise), filtering out low-quality audio segments and creating mono-speaker clusters. In this paper we compare several architectures for combining speaker diarization and music and noise detection which improve the precision and overall quality of the segmentation.

Development of a genre-dependent TTS system with cross-speaker speaking-style transplantation

One of the biggest challenges in speech synthesis is the production of contextually-appropriate naturally sounding synthetic voices. This means that a Text-To-Speech system must be able to analyze a text beyond the sentence limits in order to select, or even modulate, the speaking style according to a broader context. Our current architecture is based on a two-step approach: text genre identification and speaking style synthesis according to the detected discourse genre. For the final implementation, a set of four genres and their corresponding speaking styles were considered: broadcast news, live sport commentaries, interviews and political speeches. In the final TTS evaluation, the four speaking styles were transplanted to the neutral voices of other speakers not included in the training database. When the transplanted styles were compared to the neutral voices, transplantation was significantly preferred and the similarity to the target speaker was as high as 78%.

Broccoli sprouts: An exceptionally rich source of inducers of enzymes that protect against chemical carcinogens

Induction of phase 2 detoxication enzymes [e.g., glutathione transferases, epoxide hydrolase, NAD(P)H: quinone reductase, and glucuronosyltransferases] is a powerful strategy for achieving protection against carcinogenesis, mutagenesis, and other forms of toxicity of electrophiles and reactive forms of oxygen. Since consumption of large quantities of fruit and vegetables is associated with a striking reduction in the risk of developing a variety of malignancies, it is of interest that a number of edible plants contain substantial quantities of compounds that regulate mammalian enzymes of xenobiotic metabolism. Thus, edible plants belonging to the family Cruciferae and genus Brassica (e.g., broccoli and cauliflower) contain substantial quantities of isothiocyanates (mostly in the form of their glucosinolate precursors) some of which (e.g., sulforaphane or 4-methylsulfinylbutyl isothiocyanate) are very potent inducers of phase 2 enzymes. Unexpectedly, 3-day-old sprouts of cultivars of certain crucifers including broccoli and cauliflower contain 10–100 times higher levels of glucoraphanin (the glucosinolate of sulforaphane) than do the corresponding mature plants. Glucosinolates and isothiocyanates can be efficiently extracted from plants, without hydrolysis of glucosinolates by myrosinase, by homogenization in a mixture of equal volumes of dimethyl sulfoxide, dimethylformamide, and acetonitrile at −50°C. Extracts of 3-day-old broccoli sprouts (containing either glucoraphanin or sulforaphane as the principal enzyme inducer) were highly effective in reducing the incidence, multiplicity, and rate of development of mammary tumors in dimethylbenz(a)anthracene-treated rats. Notably, sprouts of many broccoli cultivars contain negligible quantities of indole glucosinolates, which predominate in the mature vegetable and may give rise to degradation products (e.g., indole-3-carbinol) that can enhance tumorigenesis. Hence, small quantities of crucifer sprouts may protect against the risk of cancer as effectively as much larger quantities of mature vegetables of the same variety.

A subset of skin tumor modifier loci determines survival time of tumor-bearing mice

Studies of mouse models of human cancer have established the existence of multiple tumor modifiers that influence parameters of cancer susceptibility such as tumor multiplicity, tumor size, or the probability of malignant progression. We have carried out an analysis of skin tumor susceptibility in interspecific Mus musculus/Mus spretus hybrid mice and have identified another seven loci showing either significant (six loci) or suggestive (one locus) linkage to tumor susceptibility or resistance. A specific search was carried out for skin tumor modifier loci associated with time of survival after development of a malignant tumor. A combination of resistance alleles at three markers [D6Mit15 (Skts12), D7Mit12 (Skts2), and D17Mit7 (Skts10)], all of which are close to or the same as loci associated with carcinoma incidence and/or papilloma multiplicity, is significantly associated with increased survival of mice with carcinomas, whereas the reverse combination of susceptibility alleles is significantly linked to early mortality caused by rapid carcinoma growth (χ2 = 25.22; P = 5.1 × 10−8). These data indicate that host genetic factors may be used to predict carcinoma growth rate and/or survival of individual backcross mice exposed to the same carcinogenic stimulus and suggest that mouse models may provide an approach to the identification of genetic modifiers of cancer survival in humans.

Identification of a third distinct estrogen receptor and reclassification of estrogen receptors in teleosts

This paper describes three distinct estrogen receptor (ER) subtypes: ERα, ERβ, and a unique type, ERγ, cloned from a teleost fish, the Atlantic croaker Micropogonias undulatus; the first identification of a third type of classical ER in vertebrate species. Phylogenetic analysis shows that ERγ arose through gene duplication from ERβ early in the teleost lineage and indicates that ERγ is present in other teleosts, although it has not been recognized as such. The Atlantic croaker ERγ shows amino acid differences in regions important for ligand binding and receptor activation that are conserved in all other ERγs. The three ER subtypes are genetically distinct and have different distribution patterns in Atlantic croaker tissues. In addition, ERβ and ERγ fusion proteins can each bind estradiol-17β with high affinity. The presence of three functional ERs in one species expands the role of ER multiplicity in estrogen signaling systems and provides a unique opportunity to investigate the dynamics and mechanisms of ER evolution.

A resistant genetic background leading to incomplete penetrance of intestinal neoplasia and reduced loss of heterozygosity in ApcMin/+ mice

Previous studies of Min/+ (multiple intestinal neoplasia) mice on a sensitive genetic background, C57BL/6 (B6), showed that adenomas have lost heterozygosity for the germ-line ApcMin mutation in the Apc (adenomatous polyposis coli) gene. We now report that on a strongly resistant genetic background, AKR/J (AKR), Min-induced adenoma multiplicity is reduced by about two orders of magnitude compared with that observed on the B6 background. Somatic treatment with a strong mutagen increases tumor number in AKR Min/+ mice in an age-dependent manner, similar to results previously reported for B6 Min/+ mice. Immunohistochemical analyses indicate that Apc expression is suppressed in all intestinal tumors from both untreated and treated AKR Min/+ mice. However, the mechanism of Apc inactivation in AKR Min/+ mice often differs from that observed for B6 Min/+ mice. Although loss of heterozygosity is observed in some tumors, a significant percentage of tumors showed neither loss of heterozygosity nor Apc truncation mutations. These results extend our understanding of the effects of genetic background on Min-induced tumorigenesis in several ways. First, the AKR strain carries modifiers of Min in addition to Mom1. This combination of AKR modifiers can almost completely suppress spontaneous intestinal tumorigenesis associated with the Min mutation. Second, even on such a highly resistant genetic background, tumor formation continues to involve an absence of Apc function. The means by which Apc function is inactivated is affected by genetic background. Possible scenarios are discussed.

Murine cytomegalovirus M78 protein, a G protein-coupled receptor homologue, is a constituent of the virion and facilitates accumulation of immediate-early viral mRNA

The M78 protein of murine cytomegalovirus exhibits sequence features of a G protein-coupled receptor. It is synthesized with early kinetics, it becomes partially colocalized with Golgi markers, and it is incorporated into viral particles. We have constructed a viral substitution mutant, SMsubM78, which lacks most of the M78 ORF. The mutant produces a reduced yield in cultured 10.1 fibroblast and IC21 macrophage cell lines. The defect is multiplicity dependent and greater in the macrophage cell line. Consistent with its growth defect in cultured cells, the mutant exhibits reduced pathogenicity in mice, generating less infectious progeny than wild-type virus in all organs assayed. SMsubM78 fails to efficiently activate accumulation of the viral m123 immediate-early mRNA in infected macrophages. M78 facilitates the accumulation of the immediate-early mRNA in cycloheximide-treated cells, arguing that it acts in the absence of de novo protein synthesis. We conclude that the M78 G protein-coupled receptor homologue is delivered to cells as a constituent of the virion, and it acts to facilitate the accumulation of immediate-early mRNA.

Two genes abrogate the inhibition of murine hepatocarcinogenesis by ovarian hormones.

Hormonal and genetic factors strongly influence the susceptibility of inbred mice to hepatocarcinogenesis. Female C57BR/cdJ (BR) mice are extremely susceptible to liver tumor induction relative to other strains because they are genetically insensitive to the inhibition of hepatocarcinogenesis by ovarian hormones. To determine the genetic basis for the sensitivity of BR mice relative to resistant C57BL/6J (B6) mice, we treated 12-day-old B6BRF1 x B6 and B6BRF1 x B6BRF1 (F2) animals with N,N-diethylnitrosamine (0.1 micromol/g of body weight) and enumerated liver tumors at 32 weeks of age in males and at 50 weeks in females. Genomic DNA samples from backcross and F2 mice were analyzed for 70 informative simple sequence length polymorphism markers. Genetic markers on chromosome 17 (D17Mit21) and chromosome 1 (D1Mit33) cosegregated with high tumor multiplicity in both sexes. Together, these loci [designated Hcf1 and Hcf2 (Hepatocarcinogenesis in females), respectively] account for virtually all of the difference in sensitivity between BR and B6 mice. The Hcf1 locus accounts for a majority of the higher susceptibility of BR mice of both sexes. Backcross female mice heterozygous at both loci (33 +/- 23 tumors per mouse) and at Hcf1 only (17 +/- 18) were 15- and 8-fold more sensitive, respectively, than mice homozygous for the B6 alleles at Hcf1 and Hcf2 (2.2 +/- 3.9). In backcross male mice, the double heterozygotes (35 +/- 22) and Hcf1 heterozygotes (28 +/- 12) were 5.4- and 4.3-fold more sensitive than mice homozygous for B6 alleles at both loci (6.5 +/- 5.4).

Models of speech synthesis.

The term "speech synthesis" has been used for diverse technical approaches. In this paper, some of the approaches used to generate synthetic speech in a text-to-speech system are reviewed, and some of the basic motivations for choosing one method over another are discussed. It is important to keep in mind, however, that speech synthesis models are needed not just for speech generation but to help us understand how speech is created, or even how articulation can explain language structure. General issues such as the synthesis of different voices, accents, and multiple languages are discussed as special challenges facing the speech synthesis community.