Search

This paper compares three unsupervised projection methods: Principal Component Analysis (PCA), which is linear, Self-Organizing Map (SOM) and Curvilinear Component Analysis (CCA), which are both nonlinear. Performance comparison of the three methods is made on a set of seismic data recorded on Stromboli that includes three classes of signals: explosion-quakes, landslides, and microtremors. The unsupervised analysis of the signals is able to discover the nature of the seismic events. Our analysis shows that the SOM algorithm discriminates better than CCA and PCA on the data under examination. [ABSTRACT FROM AUTHOR]

Chapter 2 of the book "The Corporate Assault on Youth: Commercialism, Exploitation and the End of Innocence," edited by Deron Boyles is presented. It analyzes the role of hip hop music as a medium for sex education specifically on the sexual desires of females. It likewise cites the song "Ill Na Na," by female rapper Foxy Brown as an example of a female's expression of sexual desire.

A bass reflex (or vented or ported) loudspeaker system (BRS) is a particular type of loudspeaker enclosure that makes use of the combination of two second-order mechanic/acoustic devices, i.e., the driver and a Helmotz resonator, in order to create a new system with reinforced emission in the low frequency region. The resonator is composed by the box itself in which one or more ports are present with suitable shapes and dimensions. This category of loudspeaker presents several advantages compared to closed-box systems such as higher efficiency and power, smaller dimensions and reduced distortion at lower frequencies. Notwithstanding these advantages, they present some drawbacks like more complexity and unloading of the cone below the tuning frequency. Moreover, at high power levels the airflow in the port(s) may generate unwanted noises due to turbulence as well as distortion and acoustic compression. In this work we will present and compare a series of experiments conducted on two different bass reflex ports designs to assess their performance in terms of flow turbulence and sound-level compression at high input power levels. These issues are quite important in professional sound systems, where increasing power levels and sound clarity require exponentially growing cost and weight. For these reasons it is vital to optimize port design. To the knowledge of the authors there does not exist an accurate, nonintrusive experimental full-field study of air flows emitting from reflex ports in operating conditions. In this work, the experimental fluid dynamic investigation has been conducted by means of PIV and LDA techniques. [ABSTRACT FROM AUTHOR]

Laser metrology provides techniques that have been successfully applied in industrial structural diagnostic fields but have not yet been refined and optimised for the special investigative requirements found in cultural heritage applications. A major impediment is the partial applicability of various optical coherent techniques, each one narrowing its use down to a specific application. This characteristic is not well suited for a field that encounters a great variety of diagnostic problems ranging from movable, multiple-composition museum objects, to immovable multi-layered wall paintings, statues and wood carvings, to monumental constructions and outdoor cultural heritage sites. Various diagnostic techniques have been suggested and are uniquely suited for each of the mentioned problems but it is this fragmented suitability that obstructs the technology transfer. Since optical coherent techniques for metrology are based on fundamental principles and take advantage of similar procedures for generation of informative signals for data collection, then the imposed limits elevate our aim to identify complementary capabilities to accomplish the needed functionality. [ABSTRACT FROM AUTHOR]

A large-scale survey (~700m2) of frescoes and wall paintings was undertaken in the US Capitol Building in Washington, DC to identify regions that may need structural repair due to detachment, delamination, or other defects. A common approach for post-processing time series called Proper Orthogonal Decomposition, or POD, was adapted to frequency-domain data in order to extract the essential features of the structure. We present a POD analysis for one of these panels, pinpointing regions that have experienced severe substructural degradation. [ABSTRACT FROM AUTHOR]

Service-oriented Architectures (SOA) introduce a major shift of perspective in software engineering: in contrast to components, services are used instead of being physically integrated. This leaves the user with no control over changes that can happen in the service itself. When the service evolves, the user may not be aware of the changes, and this can entail unexpected system failures. When a system integrator discovers a service and starts to use it, she/he may need to periodically re-test it to build confidence that (i) the service delivers over the time the desired functionality and (ii) at the same time it is able to meet Quality of Service requirements. Test cases can be used as a form of contract between a provider and the system integrators. This chapter describes an approach and a tool to allow users to run a test suite against a service to discover if functional and non-functional expectations are maintained over time. [ABSTRACT FROM AUTHOR]

The geomorphological interpretation of the high resolution bathymetry of the Albano lake (central Italy), together with conventional geological and geomorphological investigations for the subaerial slope, allowed us to identify several subaerial and submerged morphologies due to slope failures of different size and presumably age. Two main landslide categories will be decribed in this paper: totally submerged, combined subaerial-submerged landslides. Furthermore a detailed description of two past large slope failures (volume of 106m3) and the 1997 subaerial and submerged debris flow are presented. The wave induced by the 1997 debris flow testifies also the tsunamigenic potential of these phenomena which is still more serious if the presence of coastal settlements is taken into account. [ABSTRACT FROM AUTHOR]

Sulfate reduction can be used in a large number of environmental technologies. Methanogenic bioreactors treating organic wastewater containing sulfate can be negatively affected by the sulfide produced; however, it is possible to combine methanogenesis and sulfate reduction when adequate measures are applied. For the treatment of inorganic wastewaters containing sulfate, organic substrates or H2/CO2 are added as electron donors. Alternatively synthesis gas or methane can be used; however, the sulfate reduction rates with methane are still extremely low. Heavy metals such as Cu, Zn, Cd, Pb, Ni and Fe can be removed from waste streams by precipitation with biogenic sulfide. Because of differences in solubility products the metals can be selectively precipitated. The insoluble metal sulfides formed can be recovered in order to be reused. [ABSTRACT FROM AUTHOR]

Recent developments in sensitive genome characterization and quantitative gene expression analyses that permit precise molecular genetic fingerprinting of tumoral tissue are having a huge impact on cancer diagnostics. However, the significance of the data obtained with these techniques strictly depends on the composition of the biological sample to be analyzed and is greatly enhanced by including a preprocessing step that allows the researcher to distinguish and isolate selected cell populations from surrounding undesired material. This may represent a remarkable problem: indeed, genomic and proteomic analysis in the context of cancer investigation is susceptible to contamination by nonneoplastic cells, which can mask some tumor-specific alterations. Moreover, the heterogeneity of the tissues of a histo-logical section, in which the cell population of interest may constitute only a small fraction, can represent an insurmountable difficulty for the use of quantitative techniques that absolutely depend on genomic material stricdy derived from the cells that require analysis. This is obviously not possible if DNA or RNA is extracted from entire biopsies. In the past, this obstacle was partially overcome by manual dissection from slides with a needle or scalpel; however, this method is feasible only if there is a clear demarcation between the tissue under consideration and its surroundings and moreover, allows only an approximate separation of tissues. The recent development of microdissection systems based on laser technology has largely solved this important problem. Laser microdissection is a powerful tool for the isolation of specific cell populations (or single cells) from stained sections of both formalin-fixed, paraffin-embedded and frozen tissues, from cell cultures and even of a single chromosome within a metaphase cell. Resulting material is suitable for a wide range of downstream assays such LOH (loss of heterozygosity) studies, gene expression analysis at the mRNA level and a variety of proteomic approaches such as 2D gel analysis, reverse phase protein array and SELDI protein profiling. This chapter describes the characteristics of the most widely utilized laser microdissection systems and their current applications. [ABSTRACT FROM AUTHOR]

This paper presents an integrated navigation tool developed in the framework of an advanced study on navigation of Unmanned Aerial Vehicles. The study aimed at testing innovative navigation sensor configurations to support fully autonomous flight even during landings and other critical mission phases. The tool is composed of sensor simulation and data fusion software. The most important navigation sensors that are installed onboard an unmanned aircraft have been modeled: i.e. inertial, GPS, air data, high accuracy altimeter, and magnetometer. Their model included every non negligible error source that has been documented in the literature. Moreover, a specific sensor data fusion algorithm has been developed that integrates inertial sensor measurements with GPS and radar altimeter measurements. The paper reports on numerical testing of sensor simulator and data fusion algorithm. The algorithm was coded for real time implementation to perform hardware-in-the-loop validation and in flight tests onboard a small Unmanned Aerial Vehicle. [ABSTRACT FROM AUTHOR]

All the confirmed Soft Gamma-ray Repeaters have been observed with the EPIC instrument on the XMM-Newton satellite. We review the results obtained in these observations, providing the most accurate spectra on the persistent X-ray emission in the 1-10 keV range for these objects, and discuss them in the context of the magnetar interpretation. [ABSTRACT FROM AUTHOR]

We present a systematic analysis of all the BeppoSAX data of SGR1900+14. The observations spanning five years show that the source was brighter than usual on two occasions: ∼20 days after the August 1998 giant flare and during the 105 s long X-ray afterglow following the April 2001 intermediate flare. In the latter case, we explore the possibility of describing the observed short term spectral evolution only with a change of the temperature of the blackbody component. In the only BeppoSAX observation performed before the giant flare, the spectrum of the SGR1900+14 persistent emission was significantly harder and detected also above 10 keV with the PDS instrument. In the last BeppoSAX observation (April 2002) the flux was at least a factor 1.2 below the historical level, suggesting that the source was entering a quiescent period. [ABSTRACT FROM AUTHOR]

This review focuses on the realisation of optical sensors able to monitor the effect of complex space radiation on biological components, based on the biosensor concept. A biosensor is a device that can reveal a biochemical variable using a biological component interfaced with a transducer. It issues an electric signal which is easy to process, depending on the analysed variable. Biosensors are useful to study the effect of stress conditions on living organisms. One of the goals of this research was to develop two types of biosensors able to monitor directly the response of oxygenic photosynthetic organisms to radiation present in space in view of their importance for future space colonization. In ground experiments and in balloon stratosphere flights, the photosynthetic process has been analysed at the level of photosystem II (PSII), the supramolecular pigment-protein complex in the chloroplast which catalyses the light-induced transfer of electrons from water to plastoquinone; PSII splits water into molecular oxygen, protons and electrons, thereby sustaining an aerobic atmosphere on Earth and providing the reducing equivalents necessary to fix carbon dioxide to organic molecules, creating biomass, food and fuel. The results indicated that presence of space radiation in the dark has a synergistic effect on photosystem II activity, suggesting that PSII D1 protein turnover may be involved in resistance to space stress. The resistance of the tested microorganisms to space stress seems to be related to their position on the evolutive scale of photosynthesis. The present studies allow to establish a regular and reliable correlation between measured physical characteristics of space radiation and biological radiation effect. [ABSTRACT FROM AUTHOR]

Many chromophore molecules, such as bacteriochlorophylls, bacteriopheophytins and quinones, are arranged in Reaction Centers with a relevant distance and energy status such as to ensure unidirectional electron transfer. Therefore even a single Reaction Center is a sophisticated molecular device suitable for technological approaches. The structures and functions of the photosynthetic proteins differ in photosynthetic organisms at various evolutionary stages, allowing their exploitation in various technological applications. [ABSTRACT FROM AUTHOR]

In this paper we present the application of the inductive database approach to two practical analytical case studies: Web usage mining in Web logs and financial data. As far as concerns the Web domain, we have considered the enriched XML Web logs, that we call conceptual logs, produced by specific Web applications. These ones have been built by using a conceptual model, namely WebML, and its accompanying CASE tool, WebRatio. The Web conceptual logs integrate the usual information about user requests with meta-data concerning the Web site structure. As far as concerns the analysis of financial data, we have considered the trade stock exchange index Dow Jones and studied its component stocks from 1997 to 2002 using the so-called technical analysis. Technical analysis consists in the identification of the relevant (graphical) patterns that occur in the plot of evolution of a stock quote as time proceeds, often adopting different time granularities. On the plots the correlations between distinctive variables of the stocks quote are pointed out, such as the quote trend, the percentage variation and the volume of the stocks exchanged. In particular we adopted candle-sticks, a figurative pattern representing in a condensed diagram the evolution of the stock quotes in a daily stock exchange. In technical analysis, candle-sticks have been frequently used by practitioners to predict the trend of the stocks quotes in the market. We then apply a data mining language, namely MINE RULE, to these data in order to identify different types of patterns. As far as Web data is concerned, recurrent navigation paths, page contents most frequently visited, and anomalies such as intrusion attempts or a harmful usage of the resources are among the most important patterns. As far as concerns the financial domain, we searched for the sets of stocks which frequently exhibited a positive daily exchange in the same days, so as to constitute a collection of quotes for the constitution of the customers' portfolio, or the candle-sticks frequently associated to certain stocks, or finally the most similar stocks, in the sense that they mostly presented in the same dates the same typology of candle-stick, that is the same behaviour in time. The purpose of this paper is to show that the exploitation of the nuggets of information embedded in the data and of the specialised mining constructs provided by the query languages, enables the rapid customization of the mining procedures following to the users' need. Given our experience, we also claim that the use of queries in advanced languages, as opposed to ad-hoc heuristics, eases the specification and the discovery of a large spectrum of patterns. [ABSTRACT FROM AUTHOR]

Constraints-based mining languages are widely exploited to enhance the KDD process. In this paper we propose a novel incremental approach to extract itemsets and association rules from large databases. Here incremental is used to emphasize that the mining engine does not start from scratch. Instead, it exploits the result set of previously executed queries in order to simplify the mining process. Incremental algorithms show several beneficial features. First of all they exploit previous results in the pruning of the itemset lattice. Second, they are able to exploit the mining constraints of the current query in order to prune the search space even more. In this paper we propose two incremental algorithms that are able to deal with two — recently identified — types of constraints, namely item dependent and context dependent ones. Moreover, we describe an algorithm that can be used to extract association rules from scratch in presence of context dependent constraints. [ABSTRACT FROM AUTHOR]

Voice pathologies have become a social concern, as voice and speech play an important role in certain professions, and in the general population quality of life. In these last years emphasis has been placed in early pathology detection, for which classical perturbation measurements (jitter, shimmer, HNR, etc.) have been used. Going one step ahead the present work is aimed to estimate the values of the biomechanical parameters of the vocal fold system, as mass, stiffness and losses by the inversion of the vocal fold structure, which could help non only in pathology detection, but in classifying the specific patient's pathology as well. The model structure of the vocal cord will be presented, and a method to estimate the biomechanical parameters of the cord body structure will be described. From these, deviations from normophonic cases, and unbalance between cords may be extracted to serve as pathology correlates. The relevance of deviations and unbalance in Pathology Detection is shown through Principal Component Analysis. Results for normal and pathological cases will be presented and discussed. [ABSTRACT FROM AUTHOR]

The new method of segmented wavelet transform (SegWT) makes it possible to compute the discrete-time wavelet transform of a signal segment-by-segment. This means that the method could be utilized for wavelet-type processing of a signal in "real time", or in case we need to process a long signal (not necessarily in real time), but there is insufficient computational memory capacity for it (for example in the signal processors). Then it is possible to process the signal part-by-part with low memory costs by the new method. The method is suitable also for the speech processing, e.g. denoising the speech signal via thresholding the wavelet coefficients or speech coding. In the paper, the principle of the segmented forward wavelet transform is explained and the algorithm is described in detail. [ABSTRACT FROM AUTHOR]

In this paper we present a method for blind deconvolution of linear channels based on source separation techniques, for real word signals. This technique applied to blind deconvolution problems is based in exploiting not the spatial independence between signals but the temporal independence between samples of the signal. Our objective is to minimize the mutual information between samples of the output in order to retrieve the original signal. In order to make use of use this idea the input signal must be a non-Gaussian i.i.d. signal. Because most real world signals do not have this i.i.d. nature, we will need to preprocess the original signal before the transmission into the channel. Likewise we should assure that the transmitted signal has non-Gaussian statistics in order to achieve the correct function of the algorithm. The strategy used for this preprocessing will be presented in this paper. If the receiver has the inverse of the preprocess, the original signal can be reconstructed without the convolutive distortion. [ABSTRACT FROM AUTHOR]

In current speech technology, linear prediction dominates. The linear vocal tract model is well justified biomechanically, and linear prediction is a simple and well understood signal processing task. However, it has been established that, in voiced sounds, the vocal folds exhibit a high degree of nonlinearity. Hence there exists the need for an approach to modelling the behaviour of the vocal folds. This paper presents a simple, nonlinear, biophysical vocal fold model. A complementary discrete model is derived that reflects accurately the energy dynamics in the continuous model. This model can be implemented easily on standard digital signal processing hardware, and it is formulated in such a way that a simple form of nonlinear prediction can be carried out on vocal fold signals. This model could be of utility in many speech technological applications where low computational complexity synthesis and analysis of vocal fold dynamics is required. [ABSTRACT FROM AUTHOR]

For the modeling of the speech production system linear models are widely used. However, not all features of speech can be covered by linear systems. Therefore nonlinear systems are interesting in speech processing. In this contribution a time variable recursive Volterra system is used to model the fluctuations of the voiced excitation while a linear system models the resonances of the speech production system. The estimation of the Volterra system is performed by a prediction algorithm. The prediction problem is solved with the aid of an approximation by series expansion. Speech examples show that the use of a time variable Volterra system improves the naturalness of the synthetic speech. [ABSTRACT FROM AUTHOR]

In this work, we develop and compare noise suppression filtering systems based on maximum a posterior probability (MAP) and cumulative distribution function equalization (CDFE) estimation of speech spectrum. In these systems, we use a double-gamma modeling for both the speech and noise spectral components, in which the distributions are adapted to the actual parameters in each frequency bin. The performances of the proposed systems are tested using the Aurora database they are shown to be better than conventional systems derived from the MMSE method. Whereas the MAP-based method performed best in the SNR improvement, the CDFE-based system provides a lower musical noise level and shows a higher recognition rate. [ABSTRACT FROM AUTHOR]

In this paper, new Wiener filtering based binaural sub-band schemes are proposed for adaptive speech-enhancement. The proposed architectures combine a Multi-Microphone Sub-band Adaptive (MMSBA) system with Wiener filtering in order to further reduce the in-coherent noise components resulting from application of conventional MMSBA noise cancellers. A human cochlear model resulting in a non-linear distribution of the sub-band filters is also employed in the developed schemes. Preliminary comparative results achieved in simulation experiments using anechoic speech corrupted with real automobile noise show that the proposed structures are capable of significantly outperforming the conventional MMSBA scheme without Wiener filtering. [ABSTRACT FROM AUTHOR]

Multilingual speech recognition system is required for tasks that use several languages in one speech recognition application. In this paper, we propose an approach for multilingual speech recognition by spotting consonant-vowel (CV) units. The important features of spotting approach are that there is no need for automatic segmentation of speech and it is not necessary to use models for higher level units to recognise the CV units. The main issues in spotting multilingual CV units are the location of anchor points and labeling the regions around these anchor points using suitable classifiers. The vowel onset points (VOPs) have been used as anchor points. The distribution capturing ability of autoassociative neural network (AANN) models is explored for detection of VOPs in continuous speech. We explore classification models such as support vector machines (SVMs) which are capable of discriminating confusable classes of CV units and generalisation from limited amount of training data. The data for similar CV units across languages are shared to train the classifiers for recognition of CV units of speech in multiple languages. We study the spotting approach for recognition of a large number of CV units in the broadcast news corpus of three Indian languages. [ABSTRACT FROM AUTHOR]

An important topic in Automatic Speech Recognition (ASR) is to reduce the effect of noise, in particular when mismatch exists between the training and application conditions. Many noise robutness schemes within the feature processing domain use as a prerequisite a noise estimate prior to the appearance of the speech signal which require noise robust voice activity detection and assumptions of stationary noise. However, both of these requirements are often not met and it is therefore of particular interest to investigate methods like the Quantile Based Noise Estimation (QBNE) mehtod which estimates the noise during speech and non-speech sections without the use of a voice activity detector. While the standard QBNE-method uses a fixed pre-defined quantile accross all frequency bands, this paper suggests adaptive QBNE (AQBNE) which adapts the quantile individually to each frequency band. Furthermore the paper investigates an alternative to the standard mel frequency cepstral coefficient filter bank (MFCC), an empirically chosen Speech Band Emphasizing filter bank (SBE), which improves the resolution in the speech band. The combinations of AQBNE and SBE are tested on the Danish SpeechDat-Car database and compared to the performance achieved by the standards presented by the Aurora consortium (Aurora Baseline and Aurora Advanced Fronted). For the High Mismatch (HM) condition, the AQBNE achieves significantly better performance compared to the Aurora Baseline, both when combined with SBE and standard MFCC. AQBNE also outperforms the Aurora Baseline for the Medium Mismatch (MM) and Well Matched (WM) conditions. Though for all three conditions, the Aurora Advanced Frontend achieves superior performance, the AQBNE is still a relevant method to consider for small foot print applications. [ABSTRACT FROM AUTHOR]

Speech feature extraction methods are commonly based on time and frequency processing approaches. In this paper, we propose a new framework based on sub-band processing and non-linear prediction. The key idea is to pre-process the speech signal by a filter bank. From the resulting signals, non-linear predictors are computed. The feature extraction method involves the association of different Neural Predictive Coding (NPC) models. We apply this new framework to phoneme classification and experiments carried out with the NTIMIT database show an improvement of the classification rates in comparison with the full-band approach. The new method is also shown to give better performance than the traditional Linear Predictive Coding (LPC), Mel Frequency Cepstral Coding (MFCC) and Perceptual Linear Prediction (PLP) methods. [ABSTRACT FROM AUTHOR]

Novel speech features calculated from third-order statistics of subband-filtered speech signals are introduced and studied for robust speech recognition. These features have the potential to capture nonlinear information not represented by cepstral coefficients. Also, because the features presented in this paper are based on the third-order moments, they may be more immune to Gaussian noise than cepstrals, as Gaussian distributions have zero third-order moments. Experiments on the AURORA2 database studying these features in combination with Mel-frequency cepstral coefficients (MFCC's) are presented, and some improvement over the MFCC-only baseline is shown when clean speech is used for training, though the same improvement is not seen when multi-condition training data is used. [ABSTRACT FROM AUTHOR]

The segment boundaries produced by the Synface low latency phoneme recogniser are analysed. The precision in placing the boundaries is an important factor in the Synface system as the aim is to drive the lip movements of a synthetic face for lip-reading support. The recogniser is based on a hybrid of recurrent neural networks and hidden Markov models. In this paper we analyse the look-ahead length in the Viterbi-like decoder affects the precision of boundary placement. The properties of the entropy of the posterior probabilities estimated by the neural network are also investigated in relation to the distance of the frame from a phonetic transition. [ABSTRACT FROM AUTHOR]

Automatic Speech Recognition (ASR) is essentially a problem of pattern classification, however, the time dimension of the speech signal has prevented to pose ASR as a simple static classification problem. Support Vector Machine (SVM) classifiers could provide an appropriate solution, since they are very well adapted to high-dimensional classification problems. Nevertheless, the use of SVMs for ASR is by no means straightforward, mainly because SVM classifiers require an input of fixed-dimension. In this paper we study the use of a HMM-based segmentation as a mean to get the fixed-dimension input vectors required by SVMs, in a problem of isolated-digit recognition. Different configurations for all the parameters involved have been tested. Also, we deal with the problem of multi-class classification (as SVMs are initially binary classifers), studying two of the most popular approaches: 1-vs-all and 1-vs-1. [ABSTRACT FROM AUTHOR]

Support Vector Machines (SVMs) have become a popular tool for discriminative classification. An exciting area of recent application of SVMs is in speech processing. In this paper discriminatively trained SVMs have been introduced as a novel approach for the automatic detection of voice impairments. SVMs have a distinctly different modelling strategy in the detection of voice impairments problem, compared to other methods found in the literature (such a Gaussian Mixture or Hidden Markov Models): the SVM models the boundary between the classes instead of modelling the probability density of each class. In this paper it is shown that the scheme proposed fed with short-term cepstral and noise parameters can be applied for the detection of voice impairments with a good performance. [ABSTRACT FROM AUTHOR]

The presentation concerns the simulation of disordered voices. The phonatory excitation model is based on shaping functions, which are nonlinear memoryless input-output characteristics that transform a trigonometric driving function into a synthetic phonatory excitation signal. The shaping function model enables controlling the instantaneous frequency and spectral brilliance of the phonatory excitation via two separate parameters. The presentation demonstrates the synthesis of different types of dysperiodicities via a modulation of the amplitude and instantaneous frequency of the harmonic driving function. The voice disorders that are simulated are short- and long-term perturbations of the vocal frequency and cycle amplitude, biphonation, diplophonia and raucity. Acoustic noise due to turbulent airflow is modeled by means of additive white noise. [ABSTRACT FROM AUTHOR]

The article concerns methodological problems posed by multi-step predictive analysis of speech, carried out with a view to estimating vocal dysperiodicities. Problems that are discussed are the following. First, the stability of the multi-step predictive synthesis filter; second, the decrease of quantization noise by means of multiple prediction coefficients; third, the implementation of multi-step predictive analyses via lattice filters; fourth, the adequacy per se of the multi-step predictive analysis paradigm for estimating vocal dysperiodicities. Results suggest that implementations of multi-step predictive analyses that are considered to be optimal for speech coding are sub-optimal for clinical applications and vice versa. Also, multi-step predictive analyses as such do not appear to be under all circumstances a paradigm adequate for analysing vocal dysperiodicities unambiguously. An alternative is discussed, which is based on a generalized variogram of the speech signal. [ABSTRACT FROM AUTHOR]

In this paper we propose the use of nonlinear speech features to improve the voice quality measurement. We have tested a couple of features from the Dynamical System Theory, namely: the Correlation Dimension and the largest Lyapunov Exponent. In particular, we have studied the optimal size of time window for this type of analysis in the field of the characterization of the voice quality. Two systems of automatic detection of laryngeal pathologies, one of them including these features, have been implemented with the purpose of validating the usefulness of the suggested nonlinear features. We obtain slight improvements with respect to a classical system. [ABSTRACT FROM AUTHOR]

A new Bispectra Analysis application is presented is this paper. A set of bispectrum estimators for robust and effective voice activity detection (VAD) algorithm are proposed for improving speech recognition performance in noisy environments. The approach is based on filtering the input channel to avoid high energy noisy components and then the determination of the speech/non-speech bispectra by means of third order auto-cumulants. This algorithm differs from many others in the way the decision rule is formulated (detection tests) and the domain used in this approach. Clear improvements in speech/non-speech discrimination accuracy demonstrate the effectiveness of the proposed VAD. It is shown that application of statistical detection test leads to a better separation of the speech and noise distributions, thus allowing a more effective discrimination and a tradeoff between complexity and performance. The algorithm also incorporates a previous noise reduction block improving the accuracy in detecting speech and non-speech. The experimental analysis carried out on the AURORA databases and tasks provides an extensive performance evaluation together with an exhaustive comparison to the standard VADs such as ITU G.729, GSM AMR and ETSI AFE for distributed speech recognition (DSR), and other recently reported VADs. [ABSTRACT FROM AUTHOR]

Electropalatography is a well established technique for recording information on the patterns of contact between the tongue and the hard palate during speech. It leads to a stream of binary vectors, called electropalatograms. We are interested in the mapping from the acoustic signal to electropalatographic information. We present results on experiments using Support Vector Classification and a combination of Principal Component Analysis and Support Vector Regression. [ABSTRACT FROM AUTHOR]

Real generalized cepstral analysis is introduced and applied to speech deconvolution. Real pseudo cepstrum of the vocal tract model impulse response is defined and applied to the analysis of Czech vowels. The energy concentration measure of the real pseudo cepstrum of the vocal tract model impulse response is introduced and evaluated for Czech vowels pronounced by male and female speakers. The goal of this investigation is to find a robust and more reliable method of vocal tract modeling also for voices with high fundamental frequency, i.e. for female and child voices. From the investigation follows that vowel and speaker dependent generalized cepstral analysis can be found which is more robust in speech modeling than cepstral and LPC analysis. [ABSTRACT FROM AUTHOR]

Cepstral analysis is used to estimate the harmonics-to-noise ratio (HNR) in speech signals. The inverse Fourier transformed liftered cepstrum approximates a noise baseline from which the harmonics-to-noise ratio is estimated. The present study highlights the cepstrum-based noise baseline estimation process; it is shown to analogous to the action of a moving average filter applied to the power spectrum of voiced speech. The noise baseline, which is taken to approximate the noise excited vocal tract is influenced by the window length and the shape of the glottal source spectrum. Two existing estimation techniques are tested systematically using synthetically generated glottal flow and voiced speech signals with a priori knowledge of the HNR. The source influence is removed using a novel harmonic pre-emphasis technique. The results indicate accurate HNR estimation using the present approach. A preliminary investigation of the method with a set of normal/ pathological data is investigated. [ABSTRACT FROM AUTHOR]

The transmission protocol of voiced speech is hypothesized to be based on a funda mental drive process, which synchronizes the vocal tract excitation on the trans mitter side and evokes the pitch perception on the receiver side. A band limited fundamental drive is extrac ted from a voice specific subband decom position of the speech signal. When the near periodic drive is used as fun damental drive of a two-level drive-response model, a more or less aperiodic voiced excitation can be recon struc ted as a more or less aperiodic trajectory on a low dimensional continuous syn chro nization manifold (surface) described by speaker and phoneme specific coupling functions. In the case of vowels and nasals the excitation can be described by a univariate coupling function, which depends on the momentary phase of the funda mental drive. In the case of other voiced consonants the coupling function may as well depend on a delayed funda mental phase with a phoneme speci fic time delay. The delay may exceed the length of the analysis window. The resulting long range correlation cannot be analysed or synthesized by models assuming stationary excitation. [ABSTRACT FROM AUTHOR]