Search

Common Voice és un projecte que convida a la gent a donar la seva veu en la seva llengua materna amb el propòsit de reunir corpus de veus. Aquests corpus, disponibles en més de noranta idiomes, estan formats per una gran quantitat d’àudios curts, en els quals els usuaris es graven llegint una frase, i es poden utilitzar per millorar els sistemes de reconeixement i síntesi de parla. Precisament en català la majoria d’aquests sistemes actualment necessiten millores en els seus models i la base de dades de Common Voice pot arribar a ser de gran utilitat. Un dels seus avantatges és l'accessibilitat per a tothom, tant per la gent que vol desenvolupar sistemes de reconeixement com per la que vol proporcionar la seva veu. Per aquest motiu, aquest projecte utilitza el corpus català amb la intenció de validar-lo, intentant generar una segmentació fonètica dels àudios i posteriorment analitzant la precisió de les etiquetes, amb l’objectiu de poder contribuir en el desenvolupament de les tecnologies de la parla, a més d’adquirir coneixements sobre el funcionament de gestió de dades i entrenament de models amb xarxes neuronals. Inicialment, es pretenia processar la base de dades completa, formada per 25 GB de dades i més de 1000 hores gravades, però això requereix una gran quantitat de temps i de recursos computacionals, així que s’ha optat per seleccionar un conjunt reduït de dades i treballar a menor escala. El procediment ha sigut realitzar un alineament forçat amb els àudios i les seves transcripcions amb els programes Festival i Montreal Forced Aligner. Per aquest últim s’ha necessitat entrenar un model acústic, utilitzant el mateix corpus de Common Voice però amb diferents arxius. Finalment s’ha analitzat la precisió temporal de les etiquetes resultants, en les que s’ha vist que no eren del tot perfectes., Common Voice es un proyecto que invita a la gente a donar su voz en su lengua materna con el propósito de reunir corpus de voz. Estos corpus, disponibles en más de noventa idiomas, estan formados por una gran cantidad de audios cortos, en los que los usuarios se graban leyendo una frase, y se pueden utilizar para mejorar los sistemes de reconocimiento y síntesi del habla. Precisamente en catalán la mayoría de estos sistemas actualmente necesitan mejoras en sus models y la base de datos de Common Voice puede llegar a ser de gran utilidad. Una de sus ventajas es la accesibilidad para todo el mundo, tanto para la gente que quiere desarrollar sistemas de reconocimiento como para la que quiere proporcionar su voz. Por eso, este proyecto utiliza el corpus catalán con la intención de validarlo, intentando generar una segmentación fonética de los audios y posteriormente analizando la precisión de las etiquetas, con el objetivo de poder contribuir con el desarrollo de las tecnologías del habla, además de adquirir conocimientos sobre el funcionamiento de gestión de datos y entrenamiento de modelos con redes neuronales. Inicialmente, se pretendía procesar la base de datos completa, formada por 25 GB de datos y más de 1000 horas grabadas, pero esto requiere una gran cantidad de tiempo y de recursos computacionales, así que se ha optado por seleccionar un conjunto reducido de datos y trabajar a menor escala. El procedimiento ha sido realizar un alineamiento forzado con los audios y sus transcripciones con los programas Festival y Montreal Forced Aligner. Para este último se ha necesitado entrenar un modelo acústico, usando el mismo corpus de Common Voice pero con diferentes archivos. Finalmente se ha analizado la precisión temporal de las etiquetas resultantes, en las que se ha visto que no eran del todo perfectas., Common Voice is a project that invites people to donate their voice in their mother tongue in order to collect speech corpora. These corpora, available in more than ninety languages, consist of a large quantity of short audio files, in which users record themselves reading a sentence, and can be used to improve speech recognition and synthesis systems. In catalan, most of these systems currently need improvement in their models and the Common Voice dataset may be very helpful. One of its advantages is the accessibility for everyone, both for people who want to develop speech recognition systems and for people who want to provide their voice data. For this reason, this project uses the catalan corpus with the purpose of validating it, generating a phonetic segmentation of the audios and subsequently analyzing the labeling precision, with the aim of being able to contribute in the development of speech technologies, as well as gaining knowledge about data management and model training. The initial idea was processing the whole dataset, consisting of 25 GB of data and more than 1000 hours of recording, but this requires a large amount of time and computing resources, so a later choice was selecting a reduced set of audios and work on a smaller scale. The procedure was performing forced alignment on the audios and their orthographic transcriptions with Festival and Montreal Forced Aligner. For the latter, the training of a new acoustic model was needed, using the same Common Voice corpus but with different files. Finally the accuracy of the labels was analyzed, but the results showed that they were not entirely accurate.

Les îlots linguistiques se constituent normalement à la faveur d'obstacles géographiques qui les préservent des contacts extérieurs, et par conséquent de l'hybridation linguistique. Tel n'est pas le cas de la ville de Sfax qui est, au contraire, une ville de commerce ouverte sur la mer, en tant que deuxième port de la Tunisie, et sur les terres, puisqu'elle est édifiée dans les plaines de la région, côtière. Tout en étant ouverte, le parler de ses habitants tranche avec celui des Tunisiens. Nous en décrirons certaines spécificités linguistiques et nous focaliserons particulièrement sur les aspects culturels., Els illots lingüístics es constitueixen normalment aprofitant els obstacles geogràfics que els preserven dels contacts exteriors i, per tant, de la hibridació lingüística. No és el cas de la ciutat de Sfax que és, per contra, una ciutat comercial oberta al mar com a segon port de Tunísia i al territori atès que està edificada en les planes de la regió costanera. Tot i ser oberta, la parla dels seus habitants contrasta amb la dels tunisians. En describim algunes de les especificitats lingüístiques i ens centrarem en particular en els aspectes culturals., Los islotes lingüísticos se constituyen normalmente al calor de obstáculos geográficos que los preservan de los contactos exteriores y, por consiguiente, de la hibridación lingüística. No es el caso de la ciudad de Sfax que es, por el contrario, una ciudad comercial abierta al mar como segundo puerto de Túnez y al territorio puesto que está edificada en las llanuras de la región costera. Aun siendo abierta, el habla de sus habitantes contrasta con la de los tunecinos. Describimos algunas de sus especificidades lingüísticas y nos centraremos en particular en los aspectos culturales., This article describes the linguistic specificities of Sfax, a Tunisian coastal city. Open to the Mediterranean, it paradoxically represents a linguistic space of which specificities stand out with the rest of the standard Tunisian dialect, despite the development of means of communication. Beyond linguistic specificities, cultural dimensions emerge.

A The Descent of Man (1871), Darwin fonamenta la idea que la nostra espècie va originar-se per selecció natural, la ment inclosa. Consegüentment, enfronta el repte de Max Müller segons el qual la capacitat de llenguatge en particular no pot explicar-se per l’acció de la selecció natural. El repte el supera Darwin amb una conjectura sobre l’evolució del llenguatge que és menys coneguda però més suggeridora i poderosa, completa i integrada que d’altres ara vigents; a més, en centrar-se en la parla articulada, és més plausible biològicament. La potència de la seva proposta neix d’un coneixement profund del llenguatge. També aquí Darwin estudia la filogènia amb la vista posada en l’ontogènia. No se li escapa tampoc la glossogènia, és a dir, el canvi lingüístic. Filogènia, ontogènia i glossogènia constitueixen, per aquest ordre, les tres parts d’aquest article.Paraules clau: Darwin, parla, vocal production learning, evolució del llenguatge, adquisició del llenguatge, glossogènia., In The Descent of Man (1871), Darwin underpins the idea that our species was evolved by natural selection, which includes the human mind. Consequently, he has to face Max Müller’s challenge that our capacity for language cannot be explained through natural selection. Such a challenge is overcome by Darwin by means of a conjecture which is more suggestive and powerful, complete and integrative than others now current; besides, by being centered in speech, it is more plausible biologically. The power of Darwin’s proposal arises from a deep knowledge of language. Here, as elsewhere in the Darwinian enterprise, phylogeny is studied with an eye on ontogeny. Glossogeny, it is to say, linguistic change is also contemplated. Phylogeny, ontogeny and glossogeny, in this order, are discussed in the three parts of the present essay.Keywords: Darwin, speech, vocal production learning, evolution of language, language acquisition, glossogeny.

A The Descent of Man (1871), Darwin fonamenta la idea que la nostra espècie va originar-se per selecció natural, la ment inclosa. Consegüentment, enfronta el repte de Max Müller segons el qual la capacitat de llenguatge en particular no pot explicar-se per l’acció de la selecció natural. El repte el supera Darwin amb una conjectura sobre l’evolució del llenguatge que és menys coneguda però més suggeridora i poderosa, completa i integrada que d’altres ara vigents; a més, en centrar-se en la parla articulada, és més plausible biològicament. La potència de la seva proposta neix d’un coneixement profund del llenguatge. També aquí Darwin estudia la filogènia amb la vista posada en l’ontogènia. No se li escapa tampoc la glossogènia, és a dir, el canvi lingüístic. Filogènia, ontogènia i glossogènia constitueixen, per aquest ordre, les tres parts d’aquest article.Paraules clau: Darwin, parla, vocal production learning, evolució del llenguatge, adquisició del llenguatge, glossogènia., In The Descent of Man (1871), Darwin underpins the idea that our species was evolved by natural selection, which includes the human mind. Consequently, he has to face Max Müller’s challenge that our capacity for language cannot be explained through natural selection. Such a challenge is overcome by Darwin by means of a conjecture which is more suggestive and powerful, complete and integrative than others now current; besides, by being centered in speech, it is more plausible biologically. The power of Darwin’s proposal arises from a deep knowledge of language. Here, as elsewhere in the Darwinian enterprise, phylogeny is studied with an eye on ontogeny. Glossogeny, it is to say, linguistic change is also contemplated. Phylogeny, ontogeny and glossogeny, in this order, are discussed in the three parts of the present essay.Keywords: Darwin, speech, vocal production learning, evolution of language, language acquisition, glossogeny.

The use of private layers is explored for a multi-task learning approach to speech emotion recognition. The methods used include LSTMs, multi-task learning, high level representation, and "in the wild" testing.

©2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works., Speech enhancement deep learning systems usually require large amounts of training data to operate in broad conditions or real applications. This makes the adaptability of those systems into new, low resource environments an important topic. In this work, we present the results of adapting a speech enhancement generative adversarial network by fine-tuning the generator with small amounts of data. We investigate the minimum requirements to obtain a stable behavior in terms of several objective metrics in two very different languages: Catalan and Korean. We also study the variability of test performance to unseen noise as a function of the amount of different types of noise available for training. Results show that adapting a pre-trained English model with 10 min of data already achieves a comparable performance to having two orders of magnitude more data. They also demonstrate the relative stability in test performance with respect to the number of training noise types., Peer Reviewed, Postprint (published version)

This study aims to demonstrate the potential for understanding first millennium glass beads not as individual representatives of types, but as collections of objects brought together and curated by owners. It uses the author’s experience as a skilled bead maker to investigate processes of bead production and mechanics of bead collection current in Scandinavia and Anglo-Saxon England in the period of 6th to 9th century AD. In the study the bead collections of seven graves are examined from the perspective of their production techniques, materials, and damage from wear and cremation. The results point to beads being acquired in different numbers and often worn for long periods of time before being buried.

This study aims to demonstrate the potential for understanding first millennium glass beads not as individual representatives of types, but as collections of objects brought together and curated by owners. It uses the author’s experience as a skilled bead maker to investigate processes of bead production and mechanics of bead collection current in Scandinavia and Anglo-Saxon England in the period of 6th to 9th century AD. In the study the bead collections of seven graves are examined from the perspective of their production techniques, materials, and damage from wear and cremation. The results point to beads being acquired in different numbers and often worn for long periods of time before being buried.

This work introduces new results on the modeling of early-vocal development using artificial intelligent cognitive architectures and a simulated vocal tract. The problem is addressed using intrinsically-motivated learning algorithms for autonomous sensorimotor exploration, a kind of algorithm belonging to the active learning architectures family. The artificial agent is able to autonomously select goals to explore its own sensorimotor system in regions where its competence to execute intended goals is improved. We propose to include a somatosensory system to provide a proprioceptive feedback signal to reinforce learning through the autonomous discovery of motor constraints. Constraints are represented by a somatosensory model which is unknown beforehand to the learner. Both the sensorimotor and somatosensory system are modeled using Gaussian mixture models. We argue that using an architecture which includes a somatosensory model would reduce redundancy in the sensorimotor model and drive the learning process more efficiently than algorithms taking into account only auditory feedback. The role of this proposed system is to predict whether an undesired collision within the vocal tract under a certain motor configuration is likely to occur. Thus, compromised motor configurations are rejected, guaranteeing that the agent is less prone to violate its own constraints., Peer Reviewed, Postprint (author's final draft)

This is an experimental study of a few glass objects from the Birka Excavations 1987-1989. A number of beads and waste from bead production have been studied. A group of these objects have also been examined with a scanning electron microscope (SEM) equipped with energy-dispersive detectors (EDS) to comparatively analyze the materials composition. Experiments to recreate the technique in which they were made have been attempted with various results. The techniques found in the beads include the use of murrini, stringer, dotting, and blown-drawn. However the waste material available to this study only suggests that in Birka, beads where made with the winding technique and ornamented with stringer and possibly dotting. The glass used was heated in clay crucibles and some rods where premade.

L'avaluació de la comunicació oral ha continuat evolucionant durant les darreres dècades. S'ha ampliat el constructe de comunicació oral i ara inclou la competència interaccional, exercint també la tecnologia un paper important en el tipus de tasques que són més populars en l'actualitat. En aquest article comentem els factors que afecten el procés de l'avaluació de la comunicació oral, les conceptualitzacions actuals del constructe que s'ha d'avaluar, i cinc tasques que s'usen per avaluar aquest constructe. Aquestes tasques inclouen entrevistes de domini oral, tasques de diàleg en parella/grup, tasques simulades, tasques de comunicació oral integrades i tasques d'imitació suscitades. Avaluem aquestes tasques basant-nos en conceptualitzacions actuals del constructe de comunicació oral, i extraiem la conclusió de que no avaluen un constructe ample d'habilitat oral amb igualtat. Segons la nostra avaluació, aconsellem els desenvolupadors de tasques que consideren els aspectes de la comunicació oral que volen incloure o excloure en llur avaluació quan seleccionen un d'aquests tipus de tasca., The assessment of oral communication has continued to evolve over the past few decades. The construct being assessed has broadened to include interactional competence, and technology has played a role in the types of tasks that are currently popular. In this paper, we discuss the factors that affect the process of oral communication assessment, current conceptualizations of the construct to be assessed, and five tasks that are used to assess this construct. These tasks include oral proficiency interviews, paired/group oral discussion tasks, simulated tasks, integrated oral communication tasks, and elicited imitation tasks. We evaluate these tasks based on current conceptualizations of the construct of oral communication, and conclude that they do not assess a broad construct of oral communication equally. Based on our evaluation, we advise test developers to consider the aspects of oral communication that they aim to include or exclude in their assessment when they select one of these task types., La evaluación de la comunicación oral ha venido evolucionando en las últimas décadas. Se ha ampliado el constructo de comunicación oral para que incluya la competencia interaccional, desempeñando también la tecnología un papel importante en el tipo de tareas que son más populares en la actualidad. En este artículo comentamos los factores que afectan al proceso de la evaluación de la comunicación oral, las conceptualizaciones actuales del constructo que debe evaluarse, y cinco tareas que se utilizan para evaluar este constructo. Estas tareas incluyen entrevistas para el dominio de la expresión oral, tareas de discusiones orales en parejas y en grupo, tareas de simulaciones, tareas de comunicación oral integradas y tareas de imitación suscitada. Evaluamos estas tareas en base a conceptualizaciones actuales del constructo de comunicación oral, y concluimos que no evalúan en la misma medida un constructo amplio de comunicación oral. Según nuestra evaluación, aconsejamos a los diseñadores de pruebas que consideren los aspectos de la comunicación oral que quieren incluir o excluir en su evaluación cuando seleccionan uno de estos tipos de tarea.

This is an experimental study of a few glass objects from the Birka Excavations 1987-1989. A number of beads and waste from bead production have been studied. A group of these objects have also been examined with a scanning electron microscope (SEM) equipped with energy-dispersive detectors (EDS) to comparatively analyze the materials composition. Experiments to recreate the technique in which they were made have been attempted with various results. The techniques found in the beads include the use of murrini, stringer, dotting, and blown-drawn. However the waste material available to this study only suggests that in Birka, beads where made with the winding technique and ornamented with stringer and possibly dotting. The glass used was heated in clay crucibles and some rods where premade.

L'avaluació de la comunicació oral ha continuat evolucionant durant les darreres dècades. S'ha ampliat el constructe de comunicació oral i ara inclou la competència interaccional, exercint també la tecnologia un paper important en el tipus de tasques que són més populars en l'actualitat. En aquest article comentem els factors que afecten el procés de l'avaluació de la comunicació oral, les conceptualitzacions actuals del constructe que s'ha d'avaluar, i cinc tasques que s'usen per avaluar aquest constructe. Aquestes tasques inclouen entrevistes de domini oral, tasques de diàleg en parella/grup, tasques simulades, tasques de comunicació oral integrades i tasques d'imitació suscitades. Avaluem aquestes tasques basant-nos en conceptualitzacions actuals del constructe de comunicació oral, i extraiem la conclusió de que no avaluen un constructe ample d'habilitat oral amb igualtat. Segons la nostra avaluació, aconsellem els desenvolupadors de tasques que consideren els aspectes de la comunicació oral que volen incloure o excloure en llur avaluació quan seleccionen un d'aquests tipus de tasca., The assessment of oral communication has continued to evolve over the past few decades. The construct being assessed has broadened to include interactional competence, and technology has played a role in the types of tasks that are currently popular. In this paper, we discuss the factors that affect the process of oral communication assessment, current conceptualizations of the construct to be assessed, and five tasks that are used to assess this construct. These tasks include oral proficiency interviews, paired/group oral discussion tasks, simulated tasks, integrated oral communication tasks, and elicited imitation tasks. We evaluate these tasks based on current conceptualizations of the construct of oral communication, and conclude that they do not assess a broad construct of oral communication equally. Based on our evaluation, we advise test developers to consider the aspects of oral communication that they aim to include or exclude in their assessment when they select one of these task types., La evaluación de la comunicación oral ha venido evolucionando en las últimas décadas. Se ha ampliado el constructo de comunicación oral para que incluya la competencia interaccional, desempeñando también la tecnología un papel importante en el tipo de tareas que son más populares en la actualidad. En este artículo comentamos los factores que afectan al proceso de la evaluación de la comunicación oral, las conceptualizaciones actuales del constructo que debe evaluarse, y cinco tareas que se utilizan para evaluar este constructo. Estas tareas incluyen entrevistas para el dominio de la expresión oral, tareas de discusiones orales en parejas y en grupo, tareas de simulaciones, tareas de comunicación oral integradas y tareas de imitación suscitada. Evaluamos estas tareas en base a conceptualizaciones actuales del constructo de comunicación oral, y concluimos que no evalúan en la misma medida un constructo amplio de comunicación oral. Según nuestra evaluación, aconsejamos a los diseñadores de pruebas que consideren los aspectos de la comunicación oral que quieren incluir o excluir en su evaluación cuando seleccionan uno de estos tipos de tarea.

This is an experimental study of a few glass objects from the Birka Excavations 1987-1989. A number of beads and waste from bead production have been studied. A group of these objects have also been examined with a scanning electron microscope (SEM) equipped with energy-dispersive detectors (EDS) to comparatively analyze the materials composition. Experiments to recreate the technique in which they were made have been attempted with various results. The techniques found in the beads include the use of murrini, stringer, dotting, and blown-drawn. However the waste material available to this study only suggests that in Birka, beads where made with the winding technique and ornamented with stringer and possibly dotting. The glass used was heated in clay crucibles and some rods where premade.

This is an experimental study of a few glass objects from the Birka Excavations 1987-1989. A number of beads and waste from bead production have been studied. A group of these objects have also been examined with a scanning electron microscope (SEM) equipped with energy-dispersive detectors (EDS) to comparatively analyze the materials composition. Experiments to recreate the technique in which they were made have been attempted with various results. The techniques found in the beads include the use of murrini, stringer, dotting, and blown-drawn. However the waste material available to this study only suggests that in Birka, beads where made with the winding technique and ornamented with stringer and possibly dotting. The glass used was heated in clay crucibles and some rods where premade.

[CATALÀ] En aquest projecte s'intenta esbrinar si l'entropia de bloc permet descriure el desenvolupament de la parla dels infants. Per a fer-ho, s'ha desenvolupat programari adient per estudiar l'entropia de bloc en tot tipus de textos., [ANGLÈS] This project tries to find out if block entropy allows the description of the development of children's speech. In order to do this, appropriate software to study the block entropy in all kinds of texts has been developed.

This work aims to improve anN-gram-based statistical machine translation system between the Catalan and Spanish languages, trained with an aligned Spanish– Catalan parallel corpus consisting of 1.7 million sentences taken from El Periódico newspaper. Starting from a linguistic error analysis above this baseline system, orthographic, morphological, lexical, semantic and syntactic problems are approached using a set of techniques. The proposed solutions include the development and application of additional statistical techniques, text pre- and post-processing tasks, and rules based on the use of grammatical categories, as well as lexical categorization. The performance of the improved system is clearly increased, as is shown in both human and automatic evaluations of the system, with a gain of about 1.1 points BLEU observed in the Spanish-to-Catalan direction of translation, and a gain of about 0.5 points in the reverse direction. The final system is freely available online as a linguistic resource, Postprint (published version)

This article presents a multimodal approach to head pose estimation of individuals in environments equipped with multiple cameras and microphones, such as SmartRooms or automatic video conferencing. Determining the individuals head orientation is the basis for many forms of more sophisticated interactions between humans and technical devices and can also be used for automatic sensor selection (camera, microphone) in communications or video surveillance systems. The use of particle filters as a unified framework for the estimation of the head orientation for both monomodal and multimodal cases is proposed. In video, we estimate head orientation from color information by exploiting spatial redundancy among cameras. Audio information is processed to estimate the direction of the voice produced by a speaker making use of the directivity characteristics of the head radiation pattern. Furthermore, two different particle filter multimodal information fusion schemes for combining the audio and video streams are analyzed in terms of accuracy and robustness. In the first one, fusion is performed at a decision level by combining each monomodal head pose estimation, while the second one uses a joint estimation system combining information at data level. Experimental results conducted over the CLEAR 2006 evaluation database are reported and the comparison of the proposed multimodal head pose estimation algorithms with the reference monomodal approaches proves the effectiveness of the proposed approach., Postprint (published version)

Fase previa para el desarrollo de un conversor texto a voz con soporte completo al idioma Catalán. En concreto, se trata de una adaptación del sistema "Festival", que fue desarrollado por la Universidad de Edimburgo con licencia de libre distribución. En el proyecto se definen los pasos para la creación del módulo de voz con soporte Catalán, se centrará en el estudio y aprendizaje del funcionamiento del sistema, y en los pasos para la construcción de la nueva voz, incluye una primera prueba de adaptación utilizando los módulos de soporte del idioma Español, para la creación de una voz en seudo-catalán, así como una prueba de definición de características y grabación de un nuevo corpus de voz.

In this paper is made a comparison between sorne quantizers applied in a Multipulse Speech coder. Vector quantizer is compared against scalar quantization in the LPC parameter in the short predictor. Adaptive Multistage Vector Quantizer gives the best results. Long predictor coefficient obtained either in open or closed loop is quantized and results are compared. Finally, excitation signa! is quantized scalar and vectorially and results are compared. Any of the different VQ squemes tested gives a better performance in SNR terms than the scalar squeme, Peer Reviewed, Postprint (published version)

Peer Reviewed, Postprint (published version)

In this paper we present a speech recogmuon system based on Hidden Markov Models, which uses as a quantifier a fonotropic map. This kind of arquitecture has been used for tasks like discriminating the phonemes of the Japanese language /1/. We will use a similar arquitecture in order to construct . a system that might be able to recognize isolated words (digits) independently of the speaker and we will compare the results with those of a classical system (RAM SES) /71 where the codebook is trained using the LBG algorithm /6/., Peer Reviewed, Postprint (published version)

Peer Reviewed, Postprint (published version)

Vector quantization (VQ) is a simultaneous quantization of a sequence of samples or vector. This process allows to malee effective use of the interrelations among the different vector components and performance arbitrarily close to the ultimate rate-distortion can be achieved by VQ if the vector dimension is high enough. Nevertheless, the exponential growth in complexity forces the use of low -dimensionality VQ in practica! systems. In this case sorne kind of adaptation is necessary to obtain adequate performance. Adaptive Multistage VQ (AMSVQ) [1] is a simple method that allows to redesign the quantizer as it is used, thus increasing its robustness and performance. In this paper, it is shown the close relation between the proposed adaptation algorithm and the methods usually employed to design the vector quantizers. Then sorne results are presented in which the AMSVQ reduce the complexity and improve the performance of previous VQ-based schemes of quantizing the LPC parameters., Peer Reviewed, Postprint (published version)

The use of an orthogonal expansion for feature selection and data compression in isolated word recognition is presented. Assuming that the spectral evolution, given by a LPC analysis, is a noisy measure in the sense that it is a linear combination of a set of real features, the objective of the orthogonal expansion is to find these real features. The new feature vectors are used to perform the recognition process. The recognition results as well as the meaning of the orthogonal expansion are discussed, Peer Reviewed, Postprint (published version)

This paper presents an algorithm that provides a syllabic segmentation of speech following the syllabification rules of Spanish language. The implemented algorithm is divided into two parts. First, an initial segmentation is made based on energy contour, sonority and duration. Second, a fine adjustement of syllable boundaries and final segmentation is made by applying syllabic rules., Peer Reviewed, Postprint (published version)

One of the most popular algorithms for connected word (or subword phonetic unit) recognition is the one-stage dynamic programming algorithm. In its available formulation, this algorithm is not designed to provide multiple hypothesis; such a limitation is currently becoming a drawback, since the need of alternative recognitions in the systems presently under research is being acknowledged. This paper introduces a modified version of the one-stage dynamic programming algorithm tailored to afford multiple hypothesis., Peer Reviewed, Postprint (published version)

Envelope and instantaneous phase of a real signal are two concepts mainly associated with modulation systems. In this paper is shown how envelope and lnstantaneous phase can be applied in coding systems. In particular, the envelope of t he residual provides informatton ahnut pitch ilnd energy of the signal and can be user to properly codify it. A parametric version of the envelope is taken from the residual of a s ignal and this informat ion ls used ln two different coders: an APC corleranda stochastlcally excited coder. In APC coder the residual envelope is used to quantize the prediction error providing step size and dynamlc bU assignment information. In stochastically excitec coders, the input to the synthesls filter is composecl by the product of the parametric envelope of the residual multiplied by an instantaneous phase taken from a codebook of random phases ., Peer Reviewed, Postprint (published version)

The combination of Vector Quantization (VQ) and predictive techniques -named AVPC systems- has been shown as interesting systems for speech waveform coding at medium-high rates (ref. 1 to 3). In this work we present a such system including a previous four subbbands splitting. This allows good quality of speech at low-medium rates (1-1.25 bits/sample). A comparative study show that the AVPC-SBC outperforms the simple AVPC coder., Peer Reviewed, Postprint (published version)

The authors present a low-bit-rate coding system where the envelope and instantaneous phase of the residual are used. A time-varying filter (short-delay filter) is excited by a signal composed by a parametric version of the residual, multiplied by a sequence from a codebook. Two alternatives are studied for the design of the codebook: sequences formed by random pulses and sequences formed by random phases to simulate the instantaneous phase of the residual, Peer Reviewed, Postprint (published version)

This communication reports the use of demisyllables for continuous speech recognition in a specific application: the recognition of Spanish numbers. After a brief outline of the recognition system, a description of demisyllable syntactic constraints and one-speaker reference generation is provided. Finally, the recognition performance is assessed by means of two experiments: the recognition of integer numbers from zero to one thousand and telephone numbers uttered in a Spanish way (strings of integers from zero to ninety nine), in both applications the results that the system yielded were excellent., Peer Reviewed, Postprint (published version)