Spectral Zero-Crossings: Localization Properties and Applications

We present an analysis of the rate of sign changes in the discrete Fourier spectrum of a sequence. The sign changes of either the real or imaginary parts of the spectrum are considered, and the rate of sign changes is termed as the spectral zero-crossing rate (SZCR). We show that SZCR carries information pertaining to the locations of transients within the temporal observation window. We show duality with temporal zero-crossing rate analysis by expressing the spectrum of a signal as a sum of sinusoids with random phases. This extension leads to spectral-domain iterative filtering approaches to stabilize the spectral zero-crossing rate and to improve upon the location estimates. The localization properties are compared with group-delay-based localization metrics in a stylized signal setting well-known in speech processing literature. We show applications to epoch estimation in voiced speech signals using the SZCR on the integrated linear prediction residue. The performance of the SZCR-based epoch localization technique is competitive with the state-of-the-art epoch estimation techniques that are based on average pitch period.

IPLR: An online resource for Greek word-level and sublexical information

We present a new online psycholinguistic resource for Greek based on analyses of written corpora combined with text processing technologies developed at the Institute for Language & Speech Processing (ILSP), Greece. The "ILSP PsychoLinguistic Resource" (IPLR) is a freely accessible service via a dedicated web page, at http://speech.ilsp.gr/iplr. IPLR provides analyses of user-submitted letter strings (words and nonwords) as well as frequency tables for important units and conditions such as syllables, bigrams, and neighbors, calculated over two word lists based on printed text corpora and their phonetic transcription. Online tools allow retrieval of words matching user-specified orthographic or phonetic patterns. All results and processing code (in the Python programming language) are freely available for noncommercial educational or research use. © 2010 Springer Science+Business Media B.V.

State estimation schemes for independent component coupled hidden Markov models

Conventional Hidden Markov models generally consist of a Markov chain observed through a linear map corrupted by additive noise. This general class of model has enjoyed a huge and diverse range of applications, for example, speech processing, biomedical signal processing and more recently quantitative finance. However, a lesser known extension of this general class of model is the so-called Factorial Hidden Markov Model (FHMM). FHMMs also have diverse applications, notably in machine learning, artificial intelligence and speech recognition [13, 17]. FHMMs extend the usual class of HMMs, by supposing the partially observed state process is a finite collection of distinct Markov chains, either statistically independent or dependent. There is also considerable current activity in applying collections of partially observed Markov chains to complex action recognition problems, see, for example, [6]. In this article we consider the Maximum Likelihood (ML) parameter estimation problem for FHMMs. Much of the extant literature concerning this problem presents parameter estimation schemes based on full data log-likelihood EM algorithms. This approach can be slow to converge and often imposes heavy demands on computer memory. The latter point is particularly relevant for the class of FHMMs where state space dimensions are relatively large. The contribution in this article is to develop new recursive formulae for a filter-based EM algorithm that can be implemented online. Our new formulae are equivalent ML estimators, however, these formulae are purely recursive and so, significantly reduce numerical complexity and memory requirements. A computer simulation is included to demonstrate the performance of our results. © Taylor & Francis Group, LLC.

Investigation of multilingual deep neural networks for spoken term detection

The development of high-performance speech processing systems for low-resource languages is a challenging area. One approach to address the lack of resources is to make use of data from multiple languages. A popular direction in recent years is to use bottleneck features, or hybrid systems, trained on multilingual data for speech-to-text (STT) systems. This paper presents an investigation into the application of these multilingual approaches to spoken term detection. Experiments were run using the IARPA Babel limited language pack corpora (∼10 hours/language) with 4 languages for initial multilingual system development and an additional held-out target language. STT gains achieved through using multilingual bottleneck features in a Tandem configuration are shown to also apply to keyword search (KWS). Further improvements in both STT and KWS were observed by incorporating language questions into the Tandem GMM-HMM decision trees for the training set languages. Adapted hybrid systems performed slightly worse on average than the adapted Tandem systems. A language independent acoustic model test on the target language showed that retraining or adapting of the acoustic models to the target language is currently minimally needed to achieve reasonable performance. © 2013 IEEE.

Stream segregation on a single electrode as a function of pulse rate in cochlear implant listeners.

While cochlear implants (CIs) usually provide high levels of speech recognition in quiet, speech recognition in noise remains challenging. To overcome these difficulties, it is important to understand how implanted listeners separate a target signal from interferers. Stream segregation has been studied extensively in both normal and electric hearing, as a function of place of stimulation. However, the effects of pulse rate, independent of place, on the perceptual grouping of sequential sounds in electric hearing have not yet been investigated. A rhythm detection task was used to measure stream segregation. The results of this study suggest that while CI listeners can segregate streams based on differences in pulse rate alone, the amount of stream segregation observed decreases as the base pulse rate increases. Further investigation of the perceptual dimensions encoded by the pulse rate and the effect of sequential presentation of different stimulation rates on perception could be beneficial for the future development of speech processing strategies for CIs.

Analysis of information leakage from encrypted Skype conversations

Voice over IP (VoIP) has experienced a tremendous growth over the last few years and is now widely used among the population and for business purposes. The security of such VoIP systems is often assumed, creating a false sense of privacy. This paper investigates in detail the leakage of information from Skype, a widely used and protected VoIP application. Experiments have shown that isolated phonemes can be classified and given sentences identified. By using the dynamic time warping (DTW) algorithm, frequently used in speech processing, an accuracy of 60% can be reached. The results can be further improved by choosing specific training data and reach an accuracy of 83% under specific conditions. The initial results being speaker dependent, an approach involving the Kalman filter is proposed to extract the kernel of all training signals.

A rule-based grapheme-to-phone converter for TTS systems in european portuguese

In this paper, a linguistically rule-based grapheme-to-phone (G2P) transcription algorithm is described for European Portuguese. A complete set of phonological and phonetic transcription rules regarding the European Portuguese standard variety is presented. This algorithm was implemented and tested by using online newspaper articles. The obtained experimental results gave rise to 98.80% of accuracy rate. Future developments in order to increase this value are foreseen. Our purpose with this work is to develop a module/ tool that can improve synthetic speech naturalness in European Portuguese. Other applications of this system can be expected like language teaching/learning. These results, together with our perspectives of future improvements, have proved the dramatic importance of linguistic knowledge on the development of Text-to-Speech systems (TTS).

Le chunking perceptif de la parole : sur la nature du groupement temporel et son effet sur la mémoire immédiate

Dans de nombreux comportements qui reposent sur le rappel et la production de séquences, des groupements temporels émergent spontanément, créés par des délais ou des allongements. Ce « chunking » a été observé tant chez les humains que chez certains animaux et plusieurs auteurs l’attribuent à un processus général de chunking perceptif qui est conforme à la capacité de la mémoire à court terme. Cependant, aucune étude n’a établi comment ce chunking perceptif s’applique à la parole. Nous présentons une recension de la littérature qui fait ressortir certains problèmes critiques qui ont nui à la recherche sur cette question. C’est en revoyant ces problèmes qu’on propose une démonstration spécifique du chunking perceptif de la parole et de l’effet de ce processus sur la mémoire immédiate (ou mémoire de travail). Ces deux thèmes de notre thèse sont présentés séparément dans deux articles. Article 1 : The perceptual chunking of speech: a demonstration using ERPs Afin d’observer le chunking de la parole en temps réel, nous avons utilisé un paradigme de potentiels évoqués (PÉ) propice à susciter la Closure Positive Shift (CPS), une composante associée, entre autres, au traitement de marques de groupes prosodiques. Nos stimuli consistaient en des énoncés et des séries de syllabes sans sens comprenant des groupes intonatifs et des marques de groupements temporels qui pouvaient concorder, ou non, avec les marques de groupes intonatifs. Les analyses démontrent que la CPS est suscitée spécifiquement par les allongements marquant la fin des groupes temporels, indépendamment des autres variables. Notons que ces marques d’allongement, qui apparaissent universellement dans la langue parlée, créent le même type de chunking que celui qui émerge lors de l’apprentissage de séquences par des humains et des animaux. Nos résultats appuient donc l’idée que l’auditeur chunk la parole en groupes temporels et que ce chunking perceptif opère de façon similaire avec des comportements verbaux et non verbaux. Par ailleurs, les observations de l’Article 1 remettent en question des études où on associe la CPS au traitement de syntagmes intonatifs sans considérer les effets de marques temporels. Article 2 : Perceptual chunking and its effect on memory in speech processing:ERP and behavioral evidence Nous avons aussi observé comment le chunking perceptif d’énoncés en groupes temporels de différentes tailles influence la mémoire immédiate d’éléments entendus. Afin d’observer ces effets, nous avons utilisé des mesures comportementales et des PÉ, dont la composante N400 qui permettait d’évaluer la qualité de la trace mnésique d’éléments cibles étendus dans des groupes temporels. La modulation de l’amplitude relative de la N400 montre que les cibles présentées dans des groupes de 3 syllabes ont bénéficié d’une meilleure mise en mémoire immédiate que celles présentées dans des groupes plus longs. D’autres mesures comportementales et une analyse de la composante P300 ont aussi permis d’isoler l’effet de la position du groupe temporel (dans l’énoncé) sur les processus de mise en mémoire. Les études ci-dessus sont les premières à démontrer le chunking perceptif de la parole en temps réel et ses effets sur la mémoire immédiate d’éléments entendus. Dans l’ensemble, nos résultats suggèrent qu’un processus général de chunking perceptif favorise la mise en mémoire d’information séquentielle et une interprétation de la parole « chunk par chunk ».

Apprentissage des réseaux de neurones profonds et applications en traitement automatique de la langue naturelle

En apprentissage automatique, domaine qui consiste à utiliser des données pour apprendre une solution aux problèmes que nous voulons confier à la machine, le modèle des Réseaux de Neurones Artificiels (ANN) est un outil précieux. Il a été inventé voilà maintenant près de soixante ans, et pourtant, il est encore de nos jours le sujet d'une recherche active. Récemment, avec l'apprentissage profond, il a en effet permis d'améliorer l'état de l'art dans de nombreux champs d'applications comme la vision par ordinateur, le traitement de la parole et le traitement des langues naturelles. La quantité toujours grandissante de données disponibles et les améliorations du matériel informatique ont permis de faciliter l'apprentissage de modèles à haute capacité comme les ANNs profonds. Cependant, des difficultés inhérentes à l'entraînement de tels modèles, comme les minima locaux, ont encore un impact important. L'apprentissage profond vise donc à trouver des solutions, en régularisant ou en facilitant l'optimisation. Le pré-entraînnement non-supervisé, ou la technique du ``Dropout'', en sont des exemples. Les deux premiers travaux présentés dans cette thèse suivent cette ligne de recherche. Le premier étudie les problèmes de gradients diminuants/explosants dans les architectures profondes. Il montre que des choix simples, comme la fonction d'activation ou l'initialisation des poids du réseaux, ont une grande influence. Nous proposons l'initialisation normalisée pour faciliter l'apprentissage. Le second se focalise sur le choix de la fonction d'activation et présente le rectifieur, ou unité rectificatrice linéaire. Cette étude a été la première à mettre l'accent sur les fonctions d'activations linéaires par morceaux pour les réseaux de neurones profonds en apprentissage supervisé. Aujourd'hui, ce type de fonction d'activation est une composante essentielle des réseaux de neurones profonds. Les deux derniers travaux présentés se concentrent sur les applications des ANNs en traitement des langues naturelles. Le premier aborde le sujet de l'adaptation de domaine pour l'analyse de sentiment, en utilisant des Auto-Encodeurs Débruitants. Celui-ci est encore l'état de l'art de nos jours. Le second traite de l'apprentissage de données multi-relationnelles avec un modèle à base d'énergie, pouvant être utilisé pour la tâche de désambiguation de sens.

Median filtering: structure analysis and application

Median filtering is a simple digital non—linear signal smoothing operation in which median of the samples in a sliding window replaces the sample at the middle of the window. The resulting filtered sequence tends to follow polynomial trends in the original sample sequence. Median filter preserves signal edges while filtering out impulses. Due to this property, median filtering is finding applications in many areas of image and speech processing. Though median filtering is simple to realise digitally, its properties are not easily analysed with standard analysis techniques,

Minimalist perspectives for psycholinguistic research

A particular approach to the study of psycholinguistic processes is presented, which is based on a minimalist conception of language. A procedural model of language acquisition is sketched, which reconciles infants speech processing and the idea of innately guided learning with a minimalist view of the initial state of language acquisition in an account of the bootstrapping problem. Language acquisition is viewed as proceeding with the progressive specification of formal features of the functional categories of the lexicon. The possibility of a minimalist derivation to be incorporated in a sentence production and/or comprehension model is discussed. Possible sources of language impairment, as manifested in SLI (Specific Language Impairment) syndrome, are considered in the light of this integrative approach. Reference is made to experimental work carried out in Brazilian Portuguese, with some extension to European Portuguese and River-Plate Spanish.

Self-organising neural networks for adaptive control

Self-organizing neural networks have been implemented in a wide range of application areas such as speech processing, image processing, optimization and robotics. Recent variations to the basic model proposed by the authors enable it to order state space using a subset of the input vector and to apply a local adaptation procedure that does not rely on a predefined test duration limit. Both these variations have been incorporated into a new feature map architecture that forms an integral part of an Hybrid Learning System (HLS) based on a genetic-based classifier system. Problems are represented within HLS as objects characterized by environmental features. Objects controlled by the system have preset targets set against a subset of their features. The system's objective is to achieve these targets by evolving a behavioural repertoire that efficiently explores and exploits the problem environment. Feature maps encode two types of knowledge within HLS — long-term memory traces of useful regularities within the environment and the classifier performance data calibrated against an object's feature states and targets. Self-organization of these networks constitutes non-genetic-based (experience-driven) learning within HLS. This paper presents a description of the HLS architecture and an analysis of the modified feature map implementing associative memory. Initial results are presented that demonstrate the behaviour of the system on a simple control task.