Sectra Communications AB has developed a terminal for encrypted communication called the Tiger platform. During voice communication delays have sometimes been experienced resulting in conversational complications. A solution to this problem, as was proposed by Sectra, would be to introduce voice activity detection, which means a separation of speech parts and non-speech parts of the input signal, to the Tiger platform. By only transferring the speech parts to the receiver, the bandwidth needed should be dramatically decreased. A lower bandwidth needed implies that the delays slowly should disappear. The problem is then to come up with a method that manages to distinguish the speech parts from the input signal. Fortunately a lot of theory on the subject has been done and numerous voice activity methods exist today. In this thesis the theory of voice activity detection has been studied. A review of voice activity detectors that exist on the market today followed by an evaluation of some of these was performed in order to select a suitable candidate for the Tiger platform. This evaluation would later become the foundation for the selection of a voice activity detector for implementation. Finally, the implementation of the chosen voice activity detector, including a comfort noise generator, was done on the platform. This implementation was based on the special requirements of the platform. Tests of the implementation in office environments show that possible delays are steadily being reduced during periods of speech inactivity, while the active speech quality is preserved.

La recherche des dernières décennies nous a offert une compréhension détaillée des processus par lesquels les aires visuelles du cerveau reconstituent les signaux physiques de l’environnement pour en générer des représentations. Cependant, la proposition selon laquelle la perception serait également le produit d’inférences et attentes, qui nous permettraient d’interpréter plus exactement les informations entrantes à l’aide d’expériences passées, est récurrente dans l’histoire des neurosciences cognitives. Le predictive coding (PC), qui est actuellement un modèle influent de la perception, propose qu’un des rôles principaux du cerveau est de prédire les informations entrantes. L’apprentissage visuel serait ainsi orienté en fonction d’informations n’ayant pas été correctement prédites ou d’erreurs de prédiction. Le PC est associé depuis quelques années par le phénomène de suppression neuronale (SN), où la réduction graduelle de l’activité cérébrale associée au traitement répété d’un stimulus, représenterait la réduction des erreurs de prédiction. Cette thèse propose premièrement que bien que la SN puisse être le reflet d’un processus assimilable au PC, celle-ci ne le représente possiblement qu’en partie. Une mesure additionnelle reflétant la correction ou l’ajustement des prédictions déclenché par l’erreur de prédiction serait alors nécessaire. Dans un premier temps, une revue critique des principaux courants de la recherche sur l’apprentissage est présentée sous la forme d’un chapitre de livre du domaine plus large du développement des capacités d’apprentissage. Celle-ci permet de préciser les aspects fondamentaux de l’habituation, la SN et la capacité à associer des éléments en mémoire, ainsi que l’importance de caractériser ces phénomènes aussi pleinement que possible par l’utilisation de nouvelles mesures, ce qui motive les études expérimentales présentées subséquemment. Par la suite, une première étude visant à identifier une mesure complémentaire à celle de la SN reflétant un processus d’ajustement de prédictions est présentée. Cette mesure, nommée entropie multi-échelles (EME), offre une estimation de la quantité d’information d’un signal électroencéphalographique (EEG) et de la capacité de traitement des réseaux neuronaux sous-jacents. La première hypothèse de cette étude était donc que la SN serait accompagnée d’une augmentation de l’EME au-dessus de la région occipito-temporale lors d’un apprentissage de visages. Puisque les phénomènes reflétés par la SN et l’EME s’appuieraient sur la contribution de régions distantes dont le cortex préfrontal dorsolatéral, la deuxième hypothèse était que ces mesures seraient altérées par une modulation exogène de l’activité de cette région préfrontale par stimulation électrique transcrânienne à courant direct (SETCD). Les résultats ont montré que le signal EEG présentait à la fois une SN et une augmentation de l’EME avec l’apprentissage. De plus, la modulation préfrontale par SETCD a entraîné des variations de l’EME de la région occipito-temporale, sans toutefois avoir un impact sur la mesure de SN. La première étude suggère ainsi que la SN et l’EME reflètent des mécanismes cérébraux impliqués dans l’apprentissage visuel et compatibles au modèle de PC. Dans la deuxième étude, l’hypothèse d’une association entre les mesures de SN et d’EME a été reprise, cette fois dans le contexte d’un apprentissage visuel relationnel, étant donné le potentiel que représente les connaissances d’associations passées entre items pour la génération de prédictions. Dans ce contexte, des effets de SN et d’augmentation d’EME ont été obtenus à nouveau et étaient associées à la réussite de l’encodage d’associations de visages-paysages. Un deuxième aspect de cette étude visait à investiguer la présence d’effets semblables chez de jeunes enfants sains, étant donné plusieurs études suggérant que le PC et la mémoire relationnelle soient fonctionnels dans la première année de vie. Cependant, étant donné l’absence d’effets dans ce groupe, les résultats de la deuxième étude suggèrent que la présence du PC tôt dans le développement s’appuie possiblement sur d’autres ressources que la mémoire relationnelle. Les études de cette thèse sont une première démonstration du potentiel que représentent les mesures de SN et d’EME dans la compréhension des mécanismes qui sous-tendent la perception et l’apprentissage visuel. / Research over the last decades has offered detailed knowledge of the processes by which visual areas use physical signals from the environment to represent it accurately. However, the proposition that perception also relies on inferences and predictions based on past experience to allow more efficiency in the interpretation of incoming signals has been recurrent throughout the history of cognitive neuroscience. In recent years, the predictive coding (PC) model, which proposes that the brain acts as a predictor of incoming information, has been influential in this field. Learning is therefore driven by prediction error and encoding is essentially restricted to unpredicted inputs, thus allowing adjustments to predictions. PC has been associated with repetition suppression (RS), whereby the gradual reduction in brain responses associated with the repeated processing of a stimulus is thought to represent prediction error reduction. This thesis proposes that although RS may be attributable to a PC process, it may not represent it fully. To do so would necessitate the use of an additional measure reflecting prediction adjustments carried out as a consequence of prediction error. A critical review of the principal currents in the cerebral mechanisms underlying learning is presented first. This review underlines the fundamental aspects of habituation, RS and the ability to associate elements to one another in memory and the importance of characterizing these phenomena fully using new measures of learning, which motivates the experimental studies presented next. Then, a study aimed at identifying a measure complementary to RS and reflecting a prediction adjustment process is presented. This measure, named multiscale entropy (MSE), offers an estimation of the information content of an electroencephalogram (EEG), and of the underlying neural networks. The first study’s main hypothesis was that RS would be accompanied with an increase in MSE over occipito-temporal areas during learning of faces. As the processes reflected by these measures would rely on distal contributions including the dorsolateral prefrontal cortex (DLPFC), the second hypothesis was that exogenous modulation of this region using transcranial direct current stimulation (tDCS) would alter RS and MSE effects found over occipito-temporal cortex. As predicted by hypotheses, EEG signal showed both RS and MSE increase from the first presentation of a face to the second over occipito-temporal sites. Additionally, prefrontal tDCS modulated brain signal complexity over right occipito-temporal cortex during learning, but did not influence RS over the same region. The first study therefore suggests that RS and MSE reflect mechanisms involved in learning of visual stimuli that appear compatible with the PC account of perception and learning. In the second study the hypothesis of an association between RS and MSE increase was investigated again, this time in the context of a visual relational memory task, given the high potential past associations of items represent for prediction generation. In this context, RS and MSE increase effects were replicated in study trials leading to correct associations of face-landscape pairings. The second study also investigated the presence of similar effects in a sample of young healthy children, given that recent studies have found evidence of both PC mechanisms and relational memory ability emerging in the first year of life. However, given the lack of effects in this sample of participants, we suggest that while PC mechanisms may emerge early, relational memory may contribute later in the course of development. Together, the studies presented in this thesis represent the first demonstration of the potential the combined use of measures of RS and signal complexity represent in further understanding the cerebral underpinnings of visual perception and learning.

A substantial number of works have aimed at modeling the receptive field properties of the primary visual cortex (V1). Their evaluation criterion is usually the similarity of the model response properties to the recorded responses from biological organisms. However, as several algorithms were able to demonstrate some degree of similarity to biological data based on the existing criteria, we focus on the robustness against loss of information in the form of occlusions as an additional constraint for better understanding the algorithmic level of early vision in the brain. We try to investigate the influence of competition mechanisms on the robustness. Therefore, we compared four methods employing different competition mechanisms, namely, independent component analysis, non-negative matrix factorization with sparseness constraint, predictive coding/biased competition, and a Hebbian neural network with lateral inhibitory connections. Each of those methods is known to be capable of developing receptive fields comparable to those of V1 simple-cells. Since measuring the robustness of methods having simple-cell like receptive fields against occlusion is difficult, we measure the robustness using the classification accuracy on the MNIST hand written digit dataset. For this we trained all methods on the training set of the MNIST hand written digits dataset and tested them on a MNIST test set with different levels of occlusions. We observe that methods which employ competitive mechanisms have higher robustness against loss of information. Also the kind of the competition mechanisms plays an important role in robustness. Global feedback inhibition as employed in predictive coding/biased competition has an advantage compared to local lateral inhibition learned by an anti-Hebb rule.

Cette thèse est centralisée sur cette question : comment est-ce que le système visuel peut coder efficacement la position des objets en mouvement, en dépit des diverses sources d'incertitude ? Cette étude déploie une hypothèse sur la connaissance a priori de la cohérence temporelle du mouvement (Burgi et al 2000; Yuille and Grzywacz 1989). Nous avons ici étendu le cadre de modélisation précédemment proposé pour expliquer le problème de l'ouverture (Perrinet and Masson, 2012). C'est un cadre d'estimation de mouvement Bayésien mis en oeuvre par un filtrage particulaire, que l'on appelle la prévision basé sur le mouvement (MBP). Sur cette base, nous avons introduit une théorie du codage de position basée sur le mouvement, et étudié comment les mécanismes neuronaux codant la position instantanée de l'objet en mouvement pourraient être affectés par le signal de mouvement le long d'une trajectoire. Les résultats de cette thèse suggèrent que le codage de la position basé sur le mouvement peut constituer un calcul neuronal générique parmi toutes les étapes du système visuel. Cela peut en partie compenser les effets cumulatifs des délais neuronaux dans le codage de la position. En outre, il peut expliquer des changements de position basés sur le mouvement, comme par example, l'Effect de Saut de Flash. Comme un cas particulier, nous avons introduit le modèle de MBP diagonal et avons reproduit la réponse anticipée de populations de neurones dans l'aire cortical V1. Nos résultats indiquent qu'un codage en position efficace et robuste peut être fortement dépendant de l'intégration le long de la trajectoire. / Coding the position of moving objects is an essential ability of the visual system in fulfilling precise and robust tracking tasks. This thesis is focalized upon this question: How does the visual system efficiently encode the position of moving objects, despite various sources of uncertainty? This study deploys the hypothesis that the visual systems uses prior knowledge on the temporal coherency of motion (Burgi et al 2000; Yuille and Grzywacz 1989). We implemented this prior by extending the modeling framework previously proposed to explain the aperture problem (Perrinet and Masson, 2012), so-called motion-based prediction (MBP). This model is a Bayesian motion estimation framework implemented by particle filtering. Based on that, we have introduced a theory on motion-based position coding, to investigate how neural mechanisms encoding the instantaneous position of moving objects might be affected by motion. Results of this thesis suggest that motion-based position coding might be a generic neural computation among all stages of the visual system. This mechanism might partially compensate the accumulative and restrictive effects of neural delays in position coding. Also it may account for motion-based position shifts as the flash lag effect. As a specific case, results of diagonal MBP model reproduced the anticipatory response of neural populations in the primary visual cortex of macaque monkey. Our results imply that an efficient and robust position coding might be highly dependent on trajectory integration and that it constitutes a key neural signature to study the more general problem of predictive coding in sensory areas.

<p>This dissertation considers packet-switched wireless networks for transmission of variable-rate layered hybrid video streams. Target applications are video streaming and broadcasting services. The work can be divided into two main parts.</p><p>In the first part, a novel quality-scalable scheme based on coefficient refinement and encoder quality constraints is developed as a possible extension to the video coding standard H.264. After a technical introduction to the coding tools of H.264 with the main focus on error resilience features, various quality scalability schemes in previous research are reviewed. Based on this discussion, an encoder decoder framework is designed for an arbitrary number of quality layers, hereby also enabling region-of-interest coding. After that, the performance of the new system is exhaustively tested, showing that the bit rate increase typically encountered with scalable hybrid coding schemes is, for certain coding parameters, only small to moderate. The double- and triple-layer constellations of the framework are shown to perform superior to other systems.</p><p>The second part considers layered code streams as generated by the scheme of the first part. Various error propagation issues in hybrid streams are discussed, which leads to the definition of a decoder quality constraint and a segmentation of the code stream to transmit. A packetization scheme based on successive source rate consumption is drafted, followed by the formulation of the channel code rate optimization problem for an optimum assignment of available codes to the channel packets. Proper MSE-based error metrics are derived, incorporating the properties of the source signal, a terminate-on-error decoding strategy, error concealment, inter-packet dependencies, and the channel conditions. The Viterbi algorithm is presented as a low-complexity solution to the optimization problem, showing a great adaptivity of the joint source channel coding scheme to the channel conditions. An almost constant image qualiity is achieved, also in mismatch situations, while the overall channel code rate decreases only as little as necessary as the channel quality deteriorates. It is further shown that the variance of code distributions is only small, and that the codes are assigned irregularly to all channel packets.</p><p>A double-layer constellation of the framework clearly outperforms other schemes with a substantial margin. </p><p>Keywords — Digital lossy video compression, visual communication, variable bit rate (VBR), SNR scalability, layered image processing, quality layer, hybrid code stream, predictive coding, progressive bit stream, joint source channel coding, fidelity constraint, channel error robustness, resilience, concealment, packet-switched, mobile and wireless ATM, noisy transmission, packet loss, binary symmetric channel, streaming, broadcasting, satellite and radio links, H.264, MPEG-4 AVC, Viterbi, trellis, unequal error protection</p>

This dissertation considers packet-switched wireless networks for transmission of variable-rate layered hybrid video streams. Target applications are video streaming and broadcasting services. The work can be divided into two main parts. In the first part, a novel quality-scalable scheme based on coefficient refinement and encoder quality constraints is developed as a possible extension to the video coding standard H.264. After a technical introduction to the coding tools of H.264 with the main focus on error resilience features, various quality scalability schemes in previous research are reviewed. Based on this discussion, an encoder decoder framework is designed for an arbitrary number of quality layers, hereby also enabling region-of-interest coding. After that, the performance of the new system is exhaustively tested, showing that the bit rate increase typically encountered with scalable hybrid coding schemes is, for certain coding parameters, only small to moderate. The double- and triple-layer constellations of the framework are shown to perform superior to other systems. The second part considers layered code streams as generated by the scheme of the first part. Various error propagation issues in hybrid streams are discussed, which leads to the definition of a decoder quality constraint and a segmentation of the code stream to transmit. A packetization scheme based on successive source rate consumption is drafted, followed by the formulation of the channel code rate optimization problem for an optimum assignment of available codes to the channel packets. Proper MSE-based error metrics are derived, incorporating the properties of the source signal, a terminate-on-error decoding strategy, error concealment, inter-packet dependencies, and the channel conditions. The Viterbi algorithm is presented as a low-complexity solution to the optimization problem, showing a great adaptivity of the joint source channel coding scheme to the channel conditions. An almost constant image qualiity is achieved, also in mismatch situations, while the overall channel code rate decreases only as little as necessary as the channel quality deteriorates. It is further shown that the variance of code distributions is only small, and that the codes are assigned irregularly to all channel packets. A double-layer constellation of the framework clearly outperforms other schemes with a substantial margin. Keywords — Digital lossy video compression, visual communication, variable bit rate (VBR), SNR scalability, layered image processing, quality layer, hybrid code stream, predictive coding, progressive bit stream, joint source channel coding, fidelity constraint, channel error robustness, resilience, concealment, packet-switched, mobile and wireless ATM, noisy transmission, packet loss, binary symmetric channel, streaming, broadcasting, satellite and radio links, H.264, MPEG-4 AVC, Viterbi, trellis, unequal error protection

Predictive coding theorizes the capacity of neural structures to form predictions about forthcoming sensory events based on previous sensory input. This concept increasingly gains attention within experimental psychology and cognitive neuroscience. In auditory research, predictive coding has become a useful model that elegantly explains different aspects of auditory sensory processing and auditory perception. Many of these aspects are backed up by experimental evidence. However, certain fundamental features of predictive auditory processing have not been addressed so far by experimental investigations, like correlates of neural predictions that show up before the onset of an expected event. Four experiments were designed to investigate the proposed mechanism under more realistic conditions as compared to previous studies by manipulating different aspects of predictive (un)certainty, thereby examining the ecological validity of predictive processing in audition. Moreover, predictive certainty was manipulated gradually across five conditions from unpredictable to fully predictable in linearly increasing steps which drastically decreases the risk of discovering incidental findings. The results obtained from the conducted experiments partly confirm the results from previous studies by demonstrating effects of predictive certainty on ERPs in response to omissions of potentially predictable stimuli. Furthermore, results partly suggest that the auditory system actively engages in stimulus predictions in a literal sense as evidenced by gradual modulations of pre-stimulus ERPs associated with different degrees of predictive certainty. However, the current results remain inconsistent because the observed effects were relatively small and could not consistently be replicated in all follow-up experiments. The observed effects could be regained after accumulating the data across all experiments in order to increase statistical power. However, certain questions remain unanswered regarding a valid interpretation of the results in terms of predictive coding. Based on the current state of results, recommendations for future investigations are provided at the end of the current thesis in order to improve certain methodological aspects of investigating predictive coding in audition, including considerations on the design of experiments, possible suitable measures to investigate predictive coding in audition, recommendations for data acquisition and data analysis as well as recommendations for publication of results.

Predictive coding theorizes the capacity of neural structures to form predictions about forthcoming sensory events based on previous sensory input. This concept increasingly gains attention within experimental psychology and cognitive neuroscience. In auditory research, predictive coding has become a useful model that elegantly explains different aspects of auditory sensory processing and auditory perception. Many of these aspects are backed up by experimental evidence. However, certain fundamental features of predictive auditory processing have not been addressed so far by experimental investigations, like correlates of neural predictions that show up before the onset of an expected event. Four experiments were designed to investigate the proposed mechanism under more realistic conditions as compared to previous studies by manipulating different aspects of predictive (un)certainty, thereby examining the ecological validity of predictive processing in audition. Moreover, predictive certainty was manipulated gradually across five conditions from unpredictable to fully predictable in linearly increasing steps which drastically decreases the risk of discovering incidental findings. The results obtained from the conducted experiments partly confirm the results from previous studies by demonstrating effects of predictive certainty on ERPs in response to omissions of potentially predictable stimuli. Furthermore, results partly suggest that the auditory system actively engages in stimulus predictions in a literal sense as evidenced by gradual modulations of pre-stimulus ERPs associated with different degrees of predictive certainty. However, the current results remain inconsistent because the observed effects were relatively small and could not consistently be replicated in all follow-up experiments. The observed effects could be regained after accumulating the data across all experiments in order to increase statistical power. However, certain questions remain unanswered regarding a valid interpretation of the results in terms of predictive coding. Based on the current state of results, recommendations for future investigations are provided at the end of the current thesis in order to improve certain methodological aspects of investigating predictive coding in audition, including considerations on the design of experiments, possible suitable measures to investigate predictive coding in audition, recommendations for data acquisition and data analysis as well as recommendations for publication of results.:1. Introduction ... 5 1.1 An introduction to predictive coding theory ... 9 1.2 Predictive coding in audition ... 11 1.3 Electrophysiological correlates of predictive auditory processing ... 14 1.4 Limitations of previous research and aims of the thesis ... 21 2. Traditional correlates of auditory prediction ... 24 2.1 Experiment 1: Reliability of auditory predictions ... 25 2.2 Experiment 2: Accuracy of auditory predictions ... 39 3. Pre-stimulus correlates of auditory prediction ...47 3.1 Pre-stimulus effects in Experiment 1 and 2 ... 48 3.2 Experiment 3: Temporal dynamics of auditory prediction ... 56 3.3 Experiment 4: The influence of omissions on stimulus processing ... 64 4 Results across experiments ... 74 4.1 Methods ... 76 4.2 Results ... 80 4.3 Discussion ... 82 5. General Discussion ... 87 5.1 Implications for current research ... 89 5.2 Recommendations for future investigations ... 93 5.3 Future prospects ... 101 5.4 Conclusion ... 104 References ...106 Appendix ... 116

A substantial number of works have aimed at modeling the receptive field properties of the primary visual cortex (V1). Their evaluation criterion is usually the similarity of the model response properties to the recorded responses from biological organisms. However, as several algorithms were able to demonstrate some degree of similarity to biological data based on the existing criteria, we focus on the robustness against loss of information in the form of occlusions as an additional constraint for better understanding the algorithmic level of early vision in the brain. We try to investigate the influence of competition mechanisms on the robustness. Therefore, we compared four methods employing different competition mechanisms, namely, independent component analysis, non-negative matrix factorization with sparseness constraint, predictive coding/biased competition, and a Hebbian neural network with lateral inhibitory connections. Each of those methods is known to be capable of developing receptive fields comparable to those of V1 simple-cells. Since measuring the robustness of methods having simple-cell like receptive fields against occlusion is difficult, we measure the robustness using the classification accuracy on the MNIST hand written digit dataset. For this we trained all methods on the training set of the MNIST hand written digits dataset and tested them on a MNIST test set with different levels of occlusions. We observe that methods which employ competitive mechanisms have higher robustness against loss of information. Also the kind of the competition mechanisms plays an important role in robustness. Global feedback inhibition as employed in predictive coding/biased competition has an advantage compared to local lateral inhibition learned by an anti-Hebb rule.