This diploma thesis deals with digital sound synthesis. The main task was to design and implement new sound synthesizer. Created tool uses different approaches to the sound synthesis, so it can be described as a hybrid. Instrument design was inspired by existing audio synthesizers. For implementation, C++ language and VST technology from Steinberg are used. As an extension, a module, that can process voice or text input and then build a MIDI file with melody (which can be interpreted with using any synthesizer) was designed and implemented. For this module, Python language is used. For the synthesizer, a simple graphical user interface was created.

Thesis (M. Sc. (Computer Science)) -- University of Limpopo, 2021 / The task of automatic speaker recognition, wherein a system verifies or identifies speakers from a recording of their voices, has been researched for several decades. However, research in this area has been carried out largely on freely accessible speaker datasets built on languages that are well-resourced like English. This study undertakes automatic speaker recognition research focused on a low-resourced language, Sepedi. As one of the 11 official languages in South Africa, Sepedi is spoken by at least 2.8 million people. Pre-recorded voices were acquired from a speech and language national repository, namely, the National Centre for Human Language Technology (NCHLT), were we selected the Sepedi NCHLT Speech Corpus. The open-source pyAudioAnalysis python library was used to extract three types of acoustic features of speech namely, time, frequency and cepstral domain features, from the acquired speech data. The effects and compatibility of these acoustic features was investigated. It was observed that combining the three acoustic features of speech had a more significant effect than using individual features as far as speaker recognition accuracy is concerned. The study also investigated the performance of machine learning algorithms on low-resourced languages such as Sepedi. Five machine learning (ML) algorithms implemented on Scikit-learn namely, K-nearest neighbours (KNN), support vector machines (SVM), random forest (RF), logistic regression (LR), and multi-layer perceptrons (MLP) were used to train different classifier models. The GridSearchCV algorithm, also implemented on Scikit-learn, was used to deduce ideal hyper-parameters for each of the five ML algorithms. The classifier models were evaluated on recognition accuracy and the results show that the MLP classifier, with a recognition accuracy of 98%, outperforms KNN, RF, LR and SVM classifiers. A graphical user interface (GUI) is developed and the best performing classifier model, MLP, is deployed on the developed GUI intended to be used for real time speaker identification and verification tasks. Participants were recruited to the GUI performance and acceptable results were obtained

Thesis (M.Sc.) -- University of Limpopo, 2021 / Cognitive Radio (CR) is a promising technology designed to solve many issues, especially spectrum underutilisation and scarcity. The requirement for spectrum effectiveness was essential, and consequently, the possibility of CR arrived along and introduced the unlicensed Secondary Users (SU). SU can operate on the unlicensed and licensed spectrum bands on a condition that they avoid interference with the licensed Primary Users (PU). This approach is called the Dynamic Spectrum Allocation (DSA) and has solved the underutilisation of spectrum using the spectrum holes. The United States of America’s telecommunication regulator Federal Communication Commission (FCC) introduces spectrum bands by unlicensed users looking at the rapid growth of wireless applications and devices; therefore, the Fixed Spectrum Allocation (FSA) become inadequate because of the spectrum crowded issues. Accomplishing this design requirement while meeting the Quality of Service (QoS) of SU is a challenge; thus, the cross-layer design (CLD) was introduced to enhance the efficiency and effectiveness of Cognitive Network (CN). CLD arrangements in Cognitive Radio Network (CRN) are empowering; however, there are yet numerous issues and difficulties that must be addressed, such as resource allocation and others that may negatively impact network performance. Routing in CRN also necessitates the cross-layering approach. Therefore, in this work, designing a protocol that will solve routing issues and channel selection will also maximise spectrum opportunistically. In this study, we propose the Optimised Cognitive Cross-layer Multipath Probabilistic Routing (OCCMPR) protocol, which is the optimised version of Cognitive Cross-layer Multipath Probabilistic Routing (CCMPR). We used MATLAB simulation installed in the Windows 10 operating system (OS) tool to generate comparison results. We compared the OCCMPR protocol with the existing protocols, the Cognitive Ad-hoc On-demand Distance Vector (CAODV) and the CCMPR protocols. / Council for Scientific and Industrial Research (CSIR)

Persistent non-target language co-activation in spoken and visual language comprehension has been found both at the word-level and at the level of a sentence, although in the latter case, sentence bias has been observed to modulate the co-activation which can create lexical competition. In the case of trilingual speakers, both non-target languages may potentially compete with the third language (L3). The current study aimed to investigate how cross-linguistic (or interlingual) competition across three languages is modulated by sentence bias while listening to the L3. Of particular interest was whether top-down sentential information would modulate not only single but also double bottom-up driven cross-linguistic competition.  A picture-word recognition task was given to 44 L1 Russian L2 English late L3 Swedish learners, listening to Swedish sentences online while their reaction times and accuracy were collected. The results revealed shorter processing times and higher accuracy for high- compared to low-constraint sentences and overall lower accuracy (and slower reactions in high-constraint sentences) when an L1 Russian competitor’s translation phonological onset overlapped with a Swedish target word. The findings suggest that when trilinguals were processing their L3 speech, top-down information from the sentential context did not modulate the bottom-up guided L1 phonological competition. However, the effect of an L2 English L3 Swedish cognate competitor was not significant. This pattern of results is in line with BLINCS (Shook & Marian, 2013), which assumes gradual co-activation decay (i.e., a strong cross-linguistic competition effect might be observed in the end-course reaction times) and a direct visual information influence on linguistic processing. It is, however, inconsistent with the BIA+ model (Dijkstra and Van Heuven, 2002), which predicts that a high-constraint sentence context can modulate cross-linguistic competition, particularly, at later processing stages. / Ihållande samaktivering av icke-målspråk i talad och visuell språkförståelse har hittats både på ordnivå och på meningsnivå, även om i det senare fallet har meningsbias observerats för att modulera samaktiveringen som kan skapa lexikal konkurrens. När det gäller trespråkiga talare kan båda icke-målspråken potentiellt konkurrera med det tredje språket (L3). Den aktuella studien syftade till att undersöka hur den tvärlingvistiska (eller interlinguala) konkurrensen mellan tre språk moduleras av meningsförspänning när man lyssnar på L3. Av särskilt intresse var huruvida top-down meningsinformation skulle modulera inte bara enstaka utan också dubbel-bottom-up-guidade tvärlingvistisk interferens.  En bild-ordsigenkänningsuppgift gavs till 44 L1 ryska L2 engelska senlärda L3 svenska talare, som lyssnade på svenska meningar online medan deras reaktionstider och noggrannhet samlades in. Resultaten avslöjade kortare bearbetningstider och högre noggrannhet för meningar med hög jämfört med meningar med låg begränsning och lägre noggrannhet (och långsammare reaktioner i meningar med hög begränsning) totalt när en L1 rysk konkurrents fonologiska översättningsstart överlappade ett svenskt målord. Resultaten tyder på att när trespråkiga bearbetade sitt L3-tal, modulerade top-down information från sententiella sammanhang inte den bottom-up guidade L1 fonologiska konkurrensen. Effekten av en L2 engelsk L3 svensk besläktad konkurrent var dock inte signifikant. Detta resultatmönster är i linje med BLINCS (Shook & Marian, 2013), som förutsätter ett gradvis samaktiveringsförfall (dvs. en stark tvärlingvistisk konkurrenseffekt kan observeras i slutförloppets reaktionstid) och en direkt visuell informationsinflytande på språklig bearbetning. Det är dock oförenligt med BIA+ (Dijkstra och Van Heuven, 2002) som förutsäger att en meningskontext med hög begränsning kan modulera tvärspråklig konkurrens, särskilt i de senare bearbetningsstadierna.

No description available.

Thesis (PhD (Electrical and Electronic Engineering))--University of Stellenbosch, 2010. / ENGLISH ABSTRACT: We propose a new methodology, based on proper scoring rules, for the evaluation of the goodness of pattern recognizers with probabilistic outputs. The recognizers of interest take an input, known to belong to one of a discrete set of classes, and output a calibrated likelihood for each class. This is a generalization of the traditional use of proper scoring rules to evaluate the goodness of probability distributions. A recognizer with outputs in well-calibrated probability distribution form can be applied to make cost-effective Bayes decisions over a range of applications, having di fferent cost functions. A recognizer with likelihood output can additionally be employed for a wide range of prior distributions for the to-be-recognized classes. We use automatic speaker recognition and automatic spoken language recognition as prototypes of this type of pattern recognizer. The traditional evaluation methods in these fields, as represented by the series of NIST Speaker and Language Recognition Evaluations, evaluate hard decisions made by the recognizers. This makes these recognizers cost-and-prior-dependent. The proposed methodology generalizes that of the NIST evaluations, allowing for the evaluation of recognizers which are intended to be usefully applied over a wide range of applications, having variable priors and costs. The proposal includes a family of evaluation criteria, where each member of the family is formed by a proper scoring rule. We emphasize two members of this family: (i) A non-strict scoring rule, directly representing error-rate at a given prior. (ii) The strict logarithmic scoring rule which represents information content, or which equivalently represents summarized error-rate, or expected cost, over a wide range of applications. We further show how to form a family of secondary evaluation criteria, which by contrasting with the primary criteria, form an analysis of the goodness of calibration of the recognizers likelihoods. Finally, we show how to use the logarithmic scoring rule as an objective function for the discriminative training of fusion and calibration of speaker and language recognizers. / AFRIKAANSE OPSOMMING: Ons wys hoe om die onsekerheid in die uittree van outomatiese sprekerherkenning- en taalherkenningstelsels voor te stel, te meet, te kalibreer en te optimeer. Dit maak die bestaande tegnologie akkurater, doeltre ender en meer algemeen toepasbaar.

Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2007. / The diagonal covariance Gaussian Probability Density Function (PDF) has been a very popular choice as the base PDF for Automatic Speech Recognition (ASR) systems. The only choices thus far have been between the spherical, diagonal and full covariance Gaussian PDFs. These classic methods have been used for some time, but no single document could be found that contains a comparative study on these methods in the use of Pattern Recognition (PR). There also is a gap between the complexity and speed of the diagonal and full covariance Gaussian implementations. The performance differences in accuracy, speed and size between these two methods differ drastically. There is a need to find one or more models that cover this area between these two classic methods. The objectives of this thesis are to evaluate three new PDF types that fit into the area between the diagonal and full covariance Gaussian implementations to broaden the choices for ASR, to document a comparative study on the three classic methods and the newly implemented methods (from previous work) and to construct a test system to evaluate these methods on phoneme recognition. The three classic density functions are examined and issues regarding the theory, implementation and usefulness of each are discussed. A visual example of each is given to show the impact of assumptions made by each (if any). The three newly implemented PDFs are the Sparse-, Probabilistic Principal Component Analysis- (PPCA) and Factor Analysis (FA) covariance Gaussian PDFs. The theory, implementation and practical usefulness are shown and discussed. Again visual examples are provided to show the difference in modelling methodologies. The construction of a test system using two speech corpora is shown and includes issues involving signal processing, PR and evaluation of the results. The NTIMIT and AST speech corpora were used in initialisation and training the test system. The usage of the system to evaluate the PDFs discussed in this work is explained. The testing results of the three new methods confirmed that they indeed fill the gap between the diagonal and full covariance Gaussians. In our tests the newly implemented methods produced a relative improvement in error rate over a similar implemented diagonal covariance Gaussian of 0.3–4%, but took 35–78% longer to evaluate. When compared relative to the full covariance Gaussian the error rates were 18–22% worse, but the evaluation times were 61–70% faster. When all the methods were scaled to approximately the same accuracy, all the above methods were 29–143% slower than the diagonal covariance Gaussian (excluding the spherical covariance method).

My doctoral research addressed two questions regarding the influence of orthographic knowledge on speech processing. First, I attempted to identify the locus of the orthographic effects observed in spoken word recognition tasks in which the orthographic consistency and the congruency between the phonological and orthographic representations of the stimuli were manipulated. Several studies provided converging results suggesting that only phonological representations activated at lexical or postlexical processing levels are affected by orthographic knowledge, while those activated at prelexical levels are not. However, the lexical processing level is not the only factor that determines the occurrence and/or the size of the orthographic effects. Regardless of the processing level tapped by the task, the characteristics of the material and the way in which participants perform the tasks also play an important role. Second, I examined the generality of the orthographic effects both in the suprasegmental domain and in the operation of working memory. Overall, the results showed orthographic effects in both situations./La question de l’influence des connaissances orthographiques sur le traitement de la parole a été abordée sous différents angles à travers les études menées dans le cadre de ma thèse de doctorat. Plus précisément, le locus des effets orthographiques a été examiné dans des tâches de reconnaissance de la parole grâce à une manipulation de la consistance orthographique et de la congruence entre les représentations phonologique et orthographique des stimuli. Les résultats obtenus convergent pour indiquer que seules les représentations phonologiques activées dans les situations qui exigent un traitement lexical et/ou post-lexical sont affectées par les représentations orthographiques. Cependant, l’occurrence et/ou la magnitude des effets orthographiques obtenus semblent dépendre également des caractéristiques du matériel et de la manière dont les participants effectuent la tâche. La question de la généralité des effets orthographiques a aussi été abordée : les effets orthographiques ont été démontrés d’une part dans le domaine suprasegmental (sur le ton lexical) et, d’autre part, dans le fonctionnement de la mémoire de travail.

Les systèmes de dialogues oraux Homme-Machine sont des interfaces entre un utilisateur et des services. Ces services sont présents sous plusieurs formes : services bancaires, systèmes de réservations (de billets de train, d’avion), etc. Les systèmes de dialogues intègrent de nombreux modules notamment ceux de reconnaissance de la parole, de compréhension, de gestion du dialogue et de synthèse de la parole. Le module qui concerne la problématique de cette thèse est celui de compréhension de la parole. Le processus de compréhension de la parole est généralement séparé du processus de transcription. Il s’agit, d’abord, de trouver la meilleure hypothèse de reconnaissance puis d’appliquer un processus de compréhension. L’approche proposée dans cette thèse est de conserver l’espace de recherche probabiliste tout au long du processus de compréhension en l’enrichissant à chaque étape. Cette approche a été appliquée lors de la campagne d’évaluation MEDIA. Nous montrons l’intérêt de notre approche par rapport à l’approche classique. En utilisant différentes sorties du module de RAP sous forme de graphe de mots, nous montrons que les performances du décodage conceptuel se dégradent linéairement en fonction du taux d’erreurs sur les mots (WER). Cependant nous montrons qu’une approche intégrée, cherchant conjointement la meilleure séquence de mots et de concepts, donne de meilleurs résultats qu’une approche séquentielle. Dans le souci de valider notre approche, nous menons des expériences sur le corpus MEDIA dans les mêmes conditions d’évaluation que lors de la campagne MEDIA. Il s’agit de produire des interprétations sémantiques à partir des transcriptions sans erreur. Les résultats montrent que les performances atteintes par notre modèle sont au niveau des performances des systèmes ayant participé à la campagne d’évaluation. L’étude détaillée des résultats obtenus lors de la campagne MEDIA nous permet de montrer la corrélation entre, d’une part, le taux d’erreur d’interprétation et, d’autre part, le taux d’erreur mots de la reconnaissance de la parole, la taille du corpus d’apprentissage, ainsi que l’ajout de connaissance a priori aux modèles de compréhension. Une analyse d’erreurs montre l’intérêt de modifier les probabilités des treillis de mots avec des triggers, un modèle cache ou d’utiliser des règles arbitraires obligeant le passage dans une partie du graphe et s’appliquant sur la présence d’éléments déclencheurs (mots ou concepts) en fonction de l’historique. On présente les méthodes à base de d’apprentissage automatique comme nécessairement plus gourmandes en terme de corpus d’apprentissage. En modifiant la taille du corpus d’apprentissage, on peut mesurer le nombre minimal ainsi que le nombre optimal de dialogues nécessaires à l’apprentissage des modèles de langages conceptuels du système de compréhension. Des travaux de recherche menés dans cette thèse visent à déterminer quel est la quantité de corpus nécessaire à l’apprentissage des modèles de langages conceptuels à partir de laquelle les scores d’évaluation sémantiques stagnent. Une corrélation est établie entre la taille de corpus nécessaire pour l’apprentissage et la taille de corpus afin de valider le guide d’annotations. En effet, il semble, dans notre cas de l’évaluation MEDIA, qu’il ait fallu sensiblement le même nombre d’exemple pour, d’une part, valider l’annotation sémantique et, d’autre part, obtenir un modèle stochastique « de qualité » appris sur corpus. De plus, en ajoutant des données a priori à nos modèles stochastiques, nous réduisons de manière significative la taille du corpus d’apprentissage nécessaire pour atteindre les même scores du système entièrement stochastique (près de deux fois moins de corpus à score égal). Cela nous permet de confirmer que l’ajout de règles élémentaires et intuitives (chiffres, nombres, codes postaux, dates) donne des résultats très encourageants. Ce constat a mené à la réalisation d’un système hybride mêlant des modèles à base de corpus et des modèles à base de connaissance. Dans un second temps, nous nous appliquons à adapter notre système de compréhension à une application de dialogue simple : un système de routage d’appel. La problématique de cette tâche est le manque de données d’apprentissage spécifiques au domaine. Nous la résolvons en partie en utilisant divers corpus déjà à notre disposition. Lors de ce processus, nous conservons les données génériques acquises lors de la campagne MEDIA et nous y intégrons les données spécifiques au domaine. Nous montrons l’intérêt d’intégrer une tâche de classification d’appel dans un processus de compréhension de la parole spontanée. Malheureusement, nous disposons de très peu de données d’apprentissage relatives au domaine de la tâche. En utilisant notre approche intégrée de décodage conceptuel, conjointement à un processus de filtrage, nous proposons une approche sous forme de sac de mots et de concepts. Cette approche exploitée par un classifieur permet d’obtenir des taux de classification d’appels encourageants sur le corpus de test, alors que le WER est assez élevé. L’application des méthodes développées lors de la campagne MEDIA nous permet d’améliorer la robustesse du processus de routage d’appels. / Spoken dialogues systems are interfaces between users and services. Simple examples of services for which theses dialogue systems can be used include : banking, booking (hotels, trains, flights), etc. Dialogue systems are composed of a number of modules. The main modules include Automatic Speech Recognition (ASR), Spoken Language Understanding (SLU), Dialogue Management and Speech Generation. In this thesis, we concentrate on the Spoken Language Understanding component of dialogue systems. In the past, it has usual to separate the Spoken Language Understanding process from that of Automatic Speech Recognition. First, the Automatic Speech Recognition process finds the best word hypothesis. Given this hypothesis, we then find the best semantic interpretation. This thesis presents a method for the robust extraction of basic conceptual constituents (or concepts) from an audio message. The conceptual decoding model proposed follows a stochastic paradigm and is directly integrated into the Automatic Speech Recognition process. This approach allows us to keep the probabilistic search space on sequences of words produced by the Automatic Speech Recognition module, and to project it to a probabilistic search space of sequences of concepts. The experiments carried out on the French spoken dialogue corpus MEDIA, available through ELDA, show that the performance reached by our new approach is better than the traditional sequential approach. As a starting point for evaluation, the effect that deterioration of word error rate (WER) has on SLU systems is examined though use of different ASR outputs. The SLU performance appears to decrease lineary as a function of ASR word error rate.We show, however, that the proposed integrated method of searching for both words and concets, gives better results to that of a traditionnanl sequential approach. In order to validate our approach, we conduct experiments on the MEDIA corpus in the same assessment conditions used during the MEDIA campaign. The goal is toproduce error-free semantic interpretations from transcripts. The results show that the performance achieved by our model is as good as the systems involved in the evaluation campaign. Studies made on the MEDIA corpus show the concept error rate is related to the word error rate, the size of the training corpus and a priori knwoledge added to conceptual model languages. Error analyses show the interest of modifying the probabilities of word lattice with triggers, a template cache or by using arbitrary rules requiring passage through a portion of the graph and applying the presence of triggers (words or concepts) based on history. Methods based on machine learning are generally quite demanding in terms of amount of training data required. By changing the size of the training corpus, the minimum and the optimal number of dialogues needed for training conceptual language models can be measured. Research conducted in this thesis aims to determine the size of corpus necessary for training conceptual language models from which the semantic evaluation scores stagnated. A correlation is established between the necessary corpus size for learning and the corpus size necessary to validate the manual annotations. In the case of the MEDIA evaluation campaign, it took roughly the same number of examples, first to validate the semantic annotations and, secondly, to obtain a "quality" corpus-trained stochastic model. The addition of a priori knowledge to our stochastic models reduce significantly the size of the training corpus needed to achieve the same scores as a fully stochastic system (nearly half the size for the same score). It allows us to confirm that the addition of basic intuitive rules (numbers, zip codes, dates) gives very encouraging results. It leeds us to create a hybrid system combining corpus-based and knowledge-based models. The second part of the thesis examines the application of the understanding module to another simple dialogue system task, a callrouting system. A problem with this specific task is a lack of data available for training the requiered language models. We attempt to resolve this issue by supplementing he in-domain data with various other generic corpora already available, and data from the MEDIA campaing. We show the benefits of integrating a call classification task in a SLU process. Unfortunately, we have very little training corpus in the field under consideration. By using our integrated approach to decode concepts, along with an integrated process, we propose a bag of words and concepts approach. This approach used by a classifier achieved encouraging call classification rates on the test corpus, while the WER was relativelyhigh. The methods developed are shown to improve the call routing system process robustness.

Understanding speech in the presence of noise can be difficult, especially when suffering from a hearing loss. This thesis examined behavioural and electrophysiological measures of speech processing with the aim of establishing how they were influenced by hearing loss (internal degradation) and listening condition (external degradation). The hypothesis that more internal and external degradation of a speech signal would result in higher working memory (WM) involvement was investigated in four studies. The behavioural measure of speech recognition consistently decreased with worse hearing, whereas lower WM capacity only resulted in poorer speech recognition when sound were spatially co-located. Electrophysiological data (EEG) recorded during speech processing, revealed that worse hearing was associated with an increase in inhibitory alpha activity (~10 Hz). This indicates that listeners with worse hearing experienced a higher degree of WM involvement during the listening task. When increasing the level of background noise, listeners with poorer hearing exhibited a breakdown in alpha activity, suggesting that these listeners reached a ceiling at which no more WM resources could be released through neural inhibition. Worse hearing was also associated with a reduced ability to selectively attend to one of two simultaneous talkers, brought on by a reduced neural inhibition of the to-be-ignored speech. Increasing the level of background noise reduced the ability to neurally track the to-be-attended speech. That internal and external degradation affected the tracking of ignored and attended speech, respectively, indicates that the two speech streams were neurally processed as independent objects. This thesis demonstrates for the first time that hearing loss causes changes in the induced neural activity during speech processing. In the last paper of the thesis, it is tentatively suggested that neural activity can be utilized from electrodes positioned in the ear canal (EarEEG) for adapting hearing-aid processing to suite the individual listeners and situation. / Att förstå tal i brus kan vara svårt, speciellt när man lider av en hörselnedsättning. Denna avhandling undersöker beteende- och elektrofysiologiska data med föremålet att bestämma hur de påverkas av hörselskada (intern försämring) och lyssningssituation (extern försämring). Hypotesen att båda intern och extern försämring av talsignalen resulterar i mer aktivering av arbetsminnet under bearbetning av talsignaler har undersökts i fyra studier. Beteendedata visade att talförståelse försämrades med större hörselnedsättning, medan lägre arbetsminneskapacitet endast resulterade i sämre talförståelse när ljudkällorna inte var rumsligt sammanfallande. Elektrofysiologiska mätningar (EEG) gjorda under bearbetning av tal, visade at sämre hörsel associerades med högre inhibitorisk alfa-aktivitet (~10 Hz). Detta indikerar att personer med sämre hörsel upplevde en högre involvering av arbetsminnet under lyssningsuppgiften. Då nivån av bakgrundsljud höjdes, visade personer med sämre hörsel ett sammanbrott av alfaaktiviteten, vilket tyder på att de nådde ett tak där ytterligare arbetsminnes-resurser inte kunde frigöras genom neural inhibition. Sämre hörsel var också förknippat med en reducerad förmåga till at fokusera uppmärksamheten på en av två samtidiga talare, förorsakat av en reducerad förmåga till neuralt att undertrycka den störande talsignalen. En ökning av nivån av bakgrundsljud minskade förmågan att inkoda den relevante talsignalen. Att intern och extern försämring påverkade respektive inkodning av störande och relevant tal, indikerar att de två tal-strömma är neuralt behandlas som oavhängiga objekt. Denna avhandling demonstrerar för första gången att hörselskada förorsakar ändringar i den inducerade neurale aktiviteten under bearbetningen av talsignaler. I avhandlingens sista artikel förslås det preliminärt att neural aktivitet kan upptas från elektroder placerade i hörselgången som kan användas till att kontrollera hörapparat signalbehandling. / <p>Funded by the Oticon Foundation. Project number: 11-2757.</p>