Toward the ultimate synthesis/recognition system.

This paper predicts speech synthesis, speech recognition, and speaker recognition technology for the year 2001, and it describes the most important research problems to be solved in order to arrive at these ultimate synthesis and recognition systems. The problems for speech synthesis include natural and intelligible voice production, prosody control based on meaning, capability of controlling synthesized voice quality and choosing individual speaking style, multilingual and multidialectal synthesis, choice of application-oriented speaking styles, capability of adding emotion, and synthesis from concepts. The problems for speech recognition include robust recognition against speech variations, adaptation/normalization to variations due to environmental conditions and speakers, automatic knowledge acquisition for acoustic and linguistic modeling, spontaneous speech recognition, naturalness and ease of human-machine interaction, and recognition of emotion. The problems for speaker recognition are similar to those for speech recognition. The research topics related to all these techniques include the use of articulatory and perceptual constraints and evaluation methods for measuring the quality of technology and systems.

Speech technology in 2001: new research directions.

Research in speech recognition and synthesis over the past several decades has brought speech technology to a point where it is being used in "real-world" applications. However, despite the progress, the perception remains that the current technology is not flexible enough to allow easy voice communication with machines. The focus of speech research is now on producing systems that are accurate and robust but that do not impose unnecessary constraints on the user. This chapter takes a critical look at the shortcomings of the current speech recognition and synthesis algorithms, discusses the technical challenges facing research, and examines the new directions that research in speech recognition and synthesis must take in order to form the basis of new solutions suitable for supporting a wide range of applications.

A corpus-based analysis of transfer effects and connected speech processes in Vietnamese English

This paper presents a corpus-based descriptive analysis of the most prevalent transfer effects and connected speech processes observed in a comparison of 11 Vietnamese English speakers (6 females, 5 males) and 12 Australian English speakers (6 males, 6 females) over 24 grammatical paraphrase items. The phonetic processes are segmentally labelled in terms of IPA diacritic features using the EMU speech database system with the aim of labelling departures from native-speaker pronunciation. An analysis of prosodic features was made using ToBI framework. The results show many phonetic and prosodic processes which make non-native speakers’ speech distinct from native ones. The corpusbased methodology of analysing foreign accent may have implications for the evaluation of non-native accent, accented speech recognition and computer assisted pronunciation- learning.

Evaluating the appropriateness of speech input in marine applications : a field evaluation

This paper discusses the first of three studies which collectively represent a convergence of two ongoing research agendas: (1) the empirically-based comparison of the effects of evaluation environment on mobile usability evaluation results; and (2) the effect of environment - in this case lobster fishing boats - on achievable speech-recognition accuracy. We describe, in detail, our study and outline our results to date based on preliminary analysis. Broadly speaking, the potential for effective use of speech for data collection and vessel control looks very promising - surprisingly so! We outline our ongoing analysis and further work.

Evaluating the appropriateness of speech input in marine applications:a field evaluation

This paper discusses the first of three studies which collectively represent a convergence of two ongoing research agendas: (1) the empirically-based comparison of the effects of evaluation environment on mobile usability evaluation results; and (2) the effect of environment - in this case lobster fishing boats - on achievable speech-recognition accuracy. We describe, in detail, our study and outline our results to date based on preliminary analysis. Broadly speaking, the potential for effective use of speech for data collection and vessel control looks very promising - surprisingly so! We outline our ongoing analysis and further work.

The resolution of complex spectral patterns by cochlear implant and normal-hearing listeners

The differences in spectral shape resolution abilities among cochlear implant ~CI! listeners, and between CI and normal-hearing ~NH! listeners, when listening with the same number of channels ~12!, was investigated. In addition, the effect of the number of channels on spectral shape resolution was examined. The stimuli were rippled noise signals with various ripple frequency-spacings. An adaptive 4IFC procedure was used to determine the threshold for resolvable ripple spacing, which was the spacing at which an interchange in peak and valley positions could be discriminated. The results showed poorer spectral shape resolution in CI compared to NH listeners ~average thresholds of approximately 3000 and 400 Hz, respectively!, and wide variability among CI listeners ~range of approximately 800 to 8000 Hz!. There was a significant relationship between spectral shape resolution and vowel recognition. The spectral shape resolution thresholds of NH listeners increased as the number of channels increased from 1 to 16, while the CI listeners showed a performance plateau at 4–6 channels, which is consistent with previous results using speech recognition measures. These results indicate that this test may provide a measure of CI performance which is time efficient and non-linguistic, and therefore, if verified, may provide a useful contribution to the prediction of speech perception in adults and children who use CIs.

Assessment of two central auditory tests in elderly patients without hearing complaints

Speech understanding disorders in the elderly may be due to peripheral or central auditory dysfunctions. Asymmetry of results in dichotic testing increases with age, and may reflect on a lack of inter-hemisphere transmission and cognitive decline. Aim: To investigate auditory processing of aged people with no hearing complaints. Study design: clinical prospective. Materials and Methods: Twenty-two voluntary individuals, aged between 55 and 75 years, were evaluated. They reported no hearing complaints and had maximal auditory thresholds of 40 dB HL until 4 KHz, 80% of minimal speech recognition scores and peripheral symmetry between the ears. We used two kinds of tests: speech in noise and dichotic alternated dissyllables (SSW). Results were compared between males and females, right and left ears and between age groups. Results: There were no significant differences between genders, in both tests. Their Left ears showed worse results, in the competitive condition of SSW. Individuals aged 65 or older had poorer performances than those aged 55 to 64. Conclusion: Central auditory tests showed worse performance with aging. The employment of a dichotic test in the auditory evaluation setting in the elderly may help in the early identification of degenerative processes, which are common among these patients.

Multi-criteria optimization manipulator trajectory planning

In the last twenty years genetic algorithms (GAs) were applied in a plethora of fields such as: control, system identification, robotics, planning and scheduling, image processing, and pattern and speech recognition (Bäck et al., 1997). In robotics the problems of trajectory planning, collision avoidance and manipulator structure design considering a single criteria has been solved using several techniques (Alander, 2003). Most engineering applications require the optimization of several criteria simultaneously. Often the problems are complex, include discrete and continuous variables and there is no prior knowledge about the search space. These kind of problems are very more complex, since they consider multiple design criteria simultaneously within the optimization procedure. This is known as a multi-criteria (or multiobjective) optimization, that has been addressed successfully through GAs (Deb, 2001). The overall aim of multi-criteria evolutionary algorithms is to achieve a set of non-dominated optimal solutions known as Pareto front. At the end of the optimization procedure, instead of a single optimal (or near optimal) solution, the decision maker can select a solution from the Pareto front. Some of the key issues in multi-criteria GAs are: i) the number of objectives, ii) to obtain a Pareto front as wide as possible and iii) to achieve a Pareto front uniformly spread. Indeed, multi-objective techniques using GAs have been increasing in relevance as a research area. In 1989, Goldberg suggested the use of a GA to solve multi-objective problems and since then other researchers have been developing new methods, such as the multi-objective genetic algorithm (MOGA) (Fonseca & Fleming, 1995), the non-dominated sorted genetic algorithm (NSGA) (Deb, 2001), and the niched Pareto genetic algorithm (NPGA) (Horn et al., 1994), among several other variants (Coello, 1998). In this work the trajectory planning problem considers: i) robots with 2 and 3 degrees of freedom (dof ), ii) the inclusion of obstacles in the workspace and iii) up to five criteria that are used to qualify the evolving trajectory, namely the: joint traveling distance, joint velocity, end effector / Cartesian distance, end effector / Cartesian velocity and energy involved. These criteria are used to minimize the joint and end effector traveled distance, trajectory ripple and energy required by the manipulator to reach at destination point. Bearing this ideas in mind, the paper addresses the planning of robot trajectories, meaning the development of an algorithm to find a continuous motion that takes the manipulator from a given starting configuration up to a desired end position without colliding with any obstacle in the workspace. The chapter is organized as follows. Section 2 describes the trajectory planning and several approaches proposed in the literature. Section 3 formulates the problem, namely the representation adopted to solve the trajectory planning and the objectives considered in the optimization. Section 4 studies the algorithm convergence. Section 5 studies a 2R manipulator (i.e., a robot with two rotational joints/links) when the optimization trajectory considers two and five objectives. Sections 6 and 7 show the results for the 3R redundant manipulator with five goals and for other complementary experiments are described, respectively. Finally, section 8 draws the main conclusions.

Análise de Desempenho e Recomendação para Soluções IVR

Mestrado em Engenharia Informática, Área de Especialização em Tecnologias do Conhecimento e da Decisão

Predicting market direction with hidden Markov models

This paper develops the model of Bicego, Grosso, and Otranto (2008) and applies Hidden Markov Models to predict market direction. The paper draws an analogy between financial markets and speech recognition, seeking inspiration from the latter to solve common issues in quantitative investing. Whereas previous works focus mostly on very complex modifications of the original hidden markov model algorithm, the current paper provides an innovative methodology by drawing inspiration from thoroughly tested, yet simple, speech recognition methodologies. By grouping returns into sequences, Hidden Markov Models can then predict market direction the same way they are used to identify phonemes in speech recognition. The model proves highly successful in identifying market direction but fails to consistently identify whether a trend is in place. All in all, the current paper seeks to bridge the gap between speech recognition and quantitative finance and, even though the model is not fully successful, several refinements are suggested and the room for improvement is significant.

Sparse nonlinear discriminants

Kernel-Functions, Machine Learning, Least Squares, Speech Recognition, Classification, Regression

Aplicacions de reconeixement de la parla: sistemes de respeaking III

En aquest projecte es fa una introducció als reconeixedors de la parla, el seu funcionament i la seva base matemàtica. Un cop tots els conceptes han quedat clars, es mostra el mètode de creació que hem seguit per obtenir el nostre propi reconeixedor de la parla, utilitzant les eines HTK, en català. S’avaluen les seves virtuts i els seus defectes a través de diferents proves realitzades als seus components. A més a més, el projecte arrodoneix la feina implementant un sistema de dictat automàtic que explota el reconeixedor de la parla utilitzant Julius.

Intégration de méthodes de data mining dans le renseignement criminel : analyse par des structures issues de la théorie des graphes dans le profilage des stupéfiants

Data mining can be defined as the extraction of previously unknown and potentially useful information from large datasets. The main principle is to devise computer programs that run through databases and automatically seek deterministic patterns. It is applied in different fields of application, e.g., remote sensing, biometry, speech recognition, but has seldom been applied to forensic case data. The intrinsic difficulty related to the use of such data lies in its heterogeneity, which comes from the many different sources of information. The aim of this study is to highlight potential uses of pattern recognition that would provide relevant results from a criminal intelligence point of view. The role of data mining within a global crime analysis methodology is to detect all types of structures in a dataset. Once filtered and interpreted, those structures can point to previously unseen criminal activities. The interpretation of patterns for intelligence purposes is the final stage of the process. It allows the researcher to validate the whole methodology and to refine each step if necessary. An application to cutting agents found in illicit drug seizures was performed. A combinatorial approach was done, using the presence and the absence of products. Methods coming from the graph theory field were used to extract patterns in data constituted by links between products and place and date of seizure. A data mining process completed using graphing techniques is called ``graph mining''. Patterns were detected that had to be interpreted and compared with preliminary knowledge to establish their relevancy. The illicit drug profiling process is actually an intelligence process that uses preliminary illicit drug classes to classify new samples. Methods proposed in this study could be used \textit{a priori} to compare structures from preliminary and post-detection patterns. This new knowledge of a repeated structure may provide valuable complementary information to profiling and become a source of intelligence.

Reconeixement de veu mitjançant xarxes neuronals

La interacció home-màquina per mitjà de la veu cobreix moltes àrees d’investigació. Es destaquen entre altres, el reconeixement de la parla, la síntesis i identificació de discurs, la verificació i identificació de locutor i l’activació per veu (ordres) de sistemes robòtics. Reconèixer la parla és natural i simple per a les persones, però és un treball complex per a les màquines, pel qual existeixen diverses metodologies i tècniques, entre elles les Xarxes Neuronals. L’objectiu d’aquest treball és desenvolupar una eina en Matlab per al reconeixement i identificació de paraules pronunciades per un locutor, entre un conjunt de paraules possibles, i amb una bona fiabilitat dins d’uns marges preestablerts. El sistema és independent del locutor que pronuncia la paraula, és a dir, aquest locutor no haurà intervingut en el procés d’entrenament del sistema. S’ha dissenyat una interfície que permet l’adquisició del senyal de veu i el seu processament mitjançant xarxes neuronals i altres tècniques. Adaptant una part de control al sistema, es podria utilitzar per donar ordres a un robot com l’Alfa6Uvic o qualsevol altre dispositiu.

ÄÄNEEN PERUSTUVA KÄYTTÄJÄN TODENTAMINEN PUHELINVERKON LISÄARVOPALVELUISSA

Tässä diplomityössä perehdytään puhujantunnistukseen ja sen käyttökelpoisuuteen käyttäjän henkilöllisyyden todentamisessa osana puhelinverkon lisäarvopalveluja. Puhelimitse ohjattavat palvelut ovat yleensä perustuneet puhelimen näppäimillä lähetettäviin äänitaajuusvalintoihin. Käyttäjän henkilöllisyydestä on voitu varmistua esimerkiksi käyttäjätunnuksen ja salaisen tunnusluvun perusteella. Tulevaisuudessa palvelut voivat perustua puheentunnistukseen, jolloin myös käyttäjän todentaminen äänen perusteella vaikuttaa järkevältä. Työssä esitellään aluksi erilaisia biometrisiä tunnistamismenetelmiä. Työssä perehdytään tarkemmin äänen perusteella tapahtuvaan puhujan todentamiseen. Työn käytännön osuudessa toteutettiin puhelinverkon palveluihin soveltuva puhujantodennussovelluksen prototyyppi. Työn tarkoituksena oli selvittää teknologian käyttömahdollisuuksia sekä kerätä kokemusta puhujantodennuspalvelun toteuttamisesta tulevaisuutta silmällä pitäen. Prototyypin toteutuksessa ohjelmointikielenä käytettiin Javaa.