A study on independent component analysis over galois fields

Dissertação (mestrado)—Universidade de Brasília, Faculdade de Tecnologia, Departamento de Engenharia Elétrica, 2015.

Speech recognition using ad-hoc microphone arrays

While close talking microphones give the best signal quality and produce the highest accuracy from current Automatic Speech Recognition (ASR) systems, the speech signal enhanced by microphone array has been shown to be an effective alternative in a noisy environment. The use of microphone arrays in contrast to close talking microphones alleviates the feeling of discomfort and distraction to the user. For this reason, microphone arrays are popular and have been used in a wide range of applications such as teleconferencing, hearing aids, speaker tracking, and as the front-end to speech recognition systems. With advances in sensor and sensor network technology, there is considerable potential for applications that employ ad-hoc networks of microphone-equipped devices collaboratively as a virtual microphone array. By allowing such devices to be distributed throughout the users’ environment, the microphone positions are no longer constrained to traditional fixed geometrical arrangements. This flexibility in the means of data acquisition allows different audio scenes to be captured to give a complete picture of the working environment. In such ad-hoc deployment of microphone sensors, however, the lack of information about the location of devices and active speakers poses technical challenges for array signal processing algorithms which must be addressed to allow deployment in real-world applications. While not an ad-hoc sensor network, conditions approaching this have in effect been imposed in recent National Institute of Standards and Technology (NIST) ASR evaluations on distant microphone recordings of meetings. The NIST evaluation data comes from multiple sites, each with different and often loosely specified distant microphone configurations. This research investigates how microphone array methods can be applied for ad-hoc microphone arrays. A particular focus is on devising methods that are robust to unknown microphone placements in order to improve the overall speech quality and recognition performance provided by the beamforming algorithms. In ad-hoc situations, microphone positions and likely source locations are not known and beamforming must be achieved blindly. There are two general approaches that can be employed to blindly estimate the steering vector for beamforming. The first is direct estimation without regard to the microphone and source locations. An alternative approach is instead to first determine the unknown microphone positions through array calibration methods and then to use the traditional geometrical formulation for the steering vector. Following these two major approaches investigated in this thesis, a novel clustered approach which includes clustering the microphones and selecting the clusters based on their proximity to the speaker is proposed. Novel experiments are conducted to demonstrate that the proposed method to automatically select clusters of microphones (ie, a subarray), closely located both to each other and to the desired speech source, may in fact provide a more robust speech enhancement and recognition than the full array could.

Separation of Doppler radar-based respiratory signatures

Respiration detection using microwave Doppler radar has attracted significant interest primarily due to its unobtrusive form of measurement. With less preparation in comparison with attaching physical sensors on the body or wearing special clothing, Doppler radar for respiration detection and monitoring is particularly useful for long-term monitoring applications such as sleep studies (i.e. sleep apnoea, SIDS). However, motion artefacts and interference from multiple sources limit the widespread use and the scope of potential applications of this technique. Utilising the recent advances in independent component analysis (ICA) and multiple antenna configuration schemes, this work investigates the feasibility of decomposing respiratory signatures into each subject from the Doppler-based measurements. Experimental results demonstrated that FastICA is capable of separating two distinct respiratory signatures from two subjects adjacent to each other even in the presence of apnoea. In each test scenario, the separated respiratory patterns correlate closely to the reference respiration strap readings. The effectiveness of FastICA in dealing with the mixed Doppler radar respiration signals confirms its applicability in healthcare applications, especially in long-term home-based monitoring as it usually involves at least two people in the same environment (i.e. two people sleeping next to each other). Further, the use of FastICA to separate involuntary movements such as the arm swing from the respiratory signatures of a single subject was explored in a multiple antenna environment. The separated respiratory signal indeed demonstrated a high correlation with the measurements made by a respiratory strap used currently in clinical settings.

Hand somatosensory sub-cortical sources arehypoactive in migraine interictally: a functionalsource separation analysis

Background: Recent morpho-functional evidences pointed out that abnormalities in the thalamus could play a major role in the expression of migraine neurophysiological and clinical correlates. Whether this phenomenon is primary or secondary to its functional disconnection from the brain stem remains to be determined.Aim: We used a Functional Source Separation algorithmof EEG signal to extract the activity of the different neuronal pools recruited at different latencies along the somatosensory pathway in interictal migraine without aura(MO) patients. Method: Twenty MO patients and 20 healthy volunteers(HV) underwent EEG recording. Four ad-hoc functional constraints, two sub-cortical (FS14 at brain stem andFS16 at thalamic level) and two cortical (FS20 radial andFS22 tangential parietal sources), were used to extract the activity of successive stages of somatosensory information processing in response to the separate left and right median nerve electric stimulation. A band-pass digital filter (450–750 Hz) was applied offline in order to extract high-frequency oscillatory (HFO) activity from the broadband EEG signal. Results: In both stimulated sides, significant reduced subcortical brain stem (FS14) and thalamic (FS16) HFO activations characterized MO patients when compared with HV. No difference emerged in the two cortical HFO activations between two groups. Conclusion: Present results are the first neurophysiological evidence supporting the hypothesis that a functional disconnection of the thalamus from the subcortical monoaminergicsystem may underline the interictal cortical abnormal information processing in migraine. Further studiesare needed to investigate the precise directional connectivity across the entire primary subcortical and cortical somatosensory pathway in interictal MO.

Motion artefact separation in single channel doppler radar respiration measurement

Direct conversion Doppler radar has the capability to remotely monitor human respiratory activity in a non-contact form. However, the motion or movement from the subject will degrade the acquired respiration signal. As the respiration pattern is one of the essential parameters in respiratory medicine intrinsically containing more information about the respiratory function, it is particularly important to suppress or to separate these motion artefacts in order to reconstruct the corresponding patterns. Experiment results show that EMD-ICA algorithm is capable of separating the mixed respiration signal by recovering the useful information of the breathing pattern as well as the motion signatures using only a single channel measurement when using the source separation algorithm. This reduces the complexity and the cost of the sensing system while removing the undesirable artefacts. A high correlation was also observed from the recovered respiration pattern in comparison to the standard respiration strap for both experiments setup (a seated and a supine position).

Immobilized Microalgae for Nutrient Recovery from Source Separated Urine

Shortages in supply of nutrients and freshwater for a growing human population are critical global issues. Traditional centralized sewage treatment can prevent eutrophication and provide sanitation, but is neither efficient nor sustainable in terms of water and resources. Source separation of household wastes, combined with decentralized resource recovery, presents a novel approach to solve these issues. Urine contains within 1 % of household waste water up to 80 % of the nitrogen (N) and 50 % of the phosphorus (P). Since microalgae are efficient at nutrient uptake, growing these organisms in urine might be a promising technology to concomitantly clean urine and produce valuable biomass containing the major plant nutrients. While state-of-the-art suspension systems for algal cultivation have mayor shortcomings in their application, immobilized cultivation on Porous Substrate Photobioreactors (PSBRs) might be a feasible alternative. The aim of this study was to develop a robust process for nutrient recovery from minimally diluted human urine using microalgae on PSBRs. The green alga Desmodesmus abundans strain CCAC 3496 was chosen for its good growth, after screening 96 algal strains derived from urine-specific isolations and culture collections. Treatment of urine, 1:1 diluted with tap water and without addition of nutrients, was performed at a light intensity of 600 μmol photons m-2 s-1 with 2.5 % CO2 and at pH 6.5. A growth rate of 7.2 g dry weight m-² day-1 and removal efficiencies for N and P of 13.1 % and 94.1 %, respectively, were determined. Pre-treatment of urine with activated carbon was found to eliminate possible detrimental effects of pharmaceuticals. These results provide a basis for further development of the technology at pilot-scale. If found to be safe in terms human and environmental health, the biomass produced from three persons could provide the P for annual production of 31 kg wheat grain and 16 kg soybean, covering the caloric demand in food for almost one month of the year for such a household. In combination with other technologies, PSBRs could thus be applied in a decentralized resource recovery system, contributing to locally close the link between sanitation and food production.

Auditory spatial acuity approximates the resolving power of space-specific neurons

The relationship between neuronal acuity and behavioral performance was assessed in the barn owl (Tyto alba), a nocturnal raptor renowned for its ability to localize sounds and for the topographic representation of auditory space found in the midbrain. We measured discrimination of sound-source separation using a newly developed procedure involving the habituation and recovery of the pupillary dilation response. The smallest discriminable change of source location was found to be about two times finer in azimuth than in elevation. Recordings from neurons in its midbrain space map revealed that their spatial tuning, like the spatial discrimination behavior, was also better in azimuth than in elevation by a factor of about two. Because the PDR behavioral assay is mediated by the same circuitry whether discrimination is assessed in azimuth or in elevation, this difference in vertical and horizontal acuity is likely to reflect a true difference in sensory resolution, without additional confounding effects of differences in motor performance in the two dimensions. Our results, therefore, are consistent with the hypothesis that the acuity of the midbrain space map determines auditory spatial discrimination.

The detection of incipient faults in small multi-cylinder diesel engines using multiple acoustic emission sensors

This thesis investigates condition monitoring (CM) of diesel engines using acoustic emission (AE) techniques. The AE signals recorded from a small size diesel engine are mixtures of multiple sources from multiple cylinders. Thus, it is difficult to interpret the information conveyed in the signals for CM purposes. This thesis develops a series of practical signal processing techniques to overcome this problem. Various experimental studies conducted to assess the CM capabilities of AE analysis for diesel engines. A series of modified signal processing techniques were proposed. These techniques showed promising results of capability for CM of multiple cylinders diesel engine using multiple AE sensors.

Modeling High-Dimensional Audio Sequences with Recurrent Neural Networks

Cette thèse étudie des modèles de séquences de haute dimension basés sur des réseaux de neurones récurrents (RNN) et leur application à la musique et à la parole. Bien qu'en principe les RNN puissent représenter les dépendances à long terme et la dynamique temporelle complexe propres aux séquences d'intérêt comme la vidéo, l'audio et la langue naturelle, ceux-ci n'ont pas été utilisés à leur plein potentiel depuis leur introduction par Rumelhart et al. (1986a) en raison de la difficulté de les entraîner efficacement par descente de gradient. Récemment, l'application fructueuse de l'optimisation Hessian-free et d'autres techniques d'entraînement avancées ont entraîné la recrudescence de leur utilisation dans plusieurs systèmes de l'état de l'art. Le travail de cette thèse prend part à ce développement. L'idée centrale consiste à exploiter la flexibilité des RNN pour apprendre une description probabiliste de séquences de symboles, c'est-à-dire une information de haut niveau associée aux signaux observés, qui en retour pourra servir d'à priori pour améliorer la précision de la recherche d'information. Par exemple, en modélisant l'évolution de groupes de notes dans la musique polyphonique, d'accords dans une progression harmonique, de phonèmes dans un énoncé oral ou encore de sources individuelles dans un mélange audio, nous pouvons améliorer significativement les méthodes de transcription polyphonique, de reconnaissance d'accords, de reconnaissance de la parole et de séparation de sources audio respectivement. L'application pratique de nos modèles à ces tâches est détaillée dans les quatre derniers articles présentés dans cette thèse. Dans le premier article, nous remplaçons la couche de sortie d'un RNN par des machines de Boltzmann restreintes conditionnelles pour décrire des distributions de sortie multimodales beaucoup plus riches. Dans le deuxième article, nous évaluons et proposons des méthodes avancées pour entraîner les RNN. Dans les quatre derniers articles, nous examinons différentes façons de combiner nos modèles symboliques à des réseaux profonds et à la factorisation matricielle non-négative, notamment par des produits d'experts, des architectures entrée/sortie et des cadres génératifs généralisant les modèles de Markov cachés. Nous proposons et analysons également des méthodes d'inférence efficaces pour ces modèles, telles la recherche vorace chronologique, la recherche en faisceau à haute dimension, la recherche en faisceau élagué et la descente de gradient. Finalement, nous abordons les questions de l'étiquette biaisée, du maître imposant, du lissage temporel, de la régularisation et du pré-entraînement.

Automatic artefact removal from event-related potentials via clustering

This paper outlines a method for automatic artefact removal from multichannel recordings of event-related potentials (ERPs). The proposed method is based on, firstly, separation of the ERP recordings into independent components using the method of temporal decorrelation source separation (TDSEP). Secondly, the novel lagged auto-mutual information clustering (LAMIC) algorithm is used to cluster the estimated components, together with ocular reference signals, into clusters corresponding to cerebral and non-cerebral activity. Thirdly, the components in the cluster which contains the ocular reference signals are discarded. The remaining components are then recombined to reconstruct the clean ERPs.