CLOSE—A Data-Driven Approach to Speech Separation

This paper studies single-channel speech separation, assuming unknown, arbitrary temporal dynamics for the speech signals to be separated. A data-driven approach is described, which matches each mixed speech segment against a composite training segment to separate the underlying clean speech segments. To advance the separation accuracy, the new approach seeks and separates the longest mixed speech segments with matching composite training segments. Lengthening the mixed speech segments to match reduces the uncertainty of the constituent training segments, and hence the error of separation. For convenience, we call the new approach Composition of Longest Segments, or CLOSE. The CLOSE method includes a data-driven approach to model long-range temporal dynamics of speech signals, and a statistical approach to identify the longest mixed speech segments with matching composite training segments. Experiments are conducted on the Wall Street Journal database, for separating mixtures of two simultaneous large-vocabulary speech utterances spoken by two different speakers. The results are evaluated using various objective and subjective measures, including the challenge of large-vocabulary continuous speech recognition. It is shown that the new separation approach leads to significant improvement in all these measures.

Rapid Fluctuations in the Lower Solar Atmosphere

The Rapid Oscillations in the Solar Atmosphere instrument reveals solar atmospheric fluctuations at high frequencies. Spectra of variations of the G-band intensity (IG ) and Ca II K-line intensity (IK ) show correlated fluctuations above white noise to frequencies beyond 300 mHz and 50 mHz, respectively. The noise-corrected G-band spectrum for f = 28-326 mHz shows a power law with exponent -1.21 ± 0.02, consistent with the presence of turbulent motions. G-band spectral power in the 25-100 mHz ("UHF") range is concentrated at the locations of magnetic bright points in the intergranular lanes and is highly intermittent in time. The intermittence of the UHF G-band fluctuations, shown by a positive kurtosis ?, also suggests turbulence. Combining values of IG , IK , UHF power, and ? reveals two distinct states of the solar atmosphere. State 1, including almost all the data, is characterized by low IG , IK , and UHF power and ? ˜ 6. State 2, including only a very small fraction of the data, is characterized by high IG , IK , and UHF power and ? ˜ 3. Superposed epoch analysis shows that the UHF power peaks simultaneously with spatio-temporal IG maxima in either state. For State 1, IK shows 3.5 minute chromospheric oscillations with maxima occurring 21 s after IG maxima implying a 150-210 km effective height difference. However, for State 2 the IK and IG maxima are simultaneous; in this highly magnetized environment sites of G-band and K-line emission may be spatially close together.

Development of a novel experimental model to investigate radiobiological implications of respiratory motion in advanced radiotherapy

Respiratory motion introduces complex spatio-temporal variations in the dosimetry of radiotherapy. There is a paucity of literature investigating the radiobiological consequences of intrafraction motion and concerns regarding the impact of movement when applied to cancer cell lines in vitro exist. We have addressed this by developing a novel model which accurately replicates respiratory motion under experimental conditions to allow clinically relevant irradiation of cell lines. A bespoke phantom and motor driven moving platform was adapted to accommodate flasks containing medium and cells in order to replicate respiratory motion using varying frequencies and amplitude settings. To study this effect on cell survival in vitro, dose response curves were determined for human lung cancer cell lines H1299 and H460 exposed to a uniform 6 MV radiation field under moving or stationary conditions. Cell survival curves showed no significant difference between irradiation at different dose points for these cell lines in the presence or absence of motion. These data indicate that motion of unshielded cells in vitro does not affect cell survival in the presence of uniform irradiation. This model provides a novel research platform to investigate the radiobiological consequences of respiratory motion in radiotherapy.

A Corpus-Based Approach to Speech Enhancement from Nonstationary Noise

Temporal dynamics and speaker characteristics are two important features of speech that distinguish speech from noise. In this paper, we propose a method to maximally extract these two features of speech for speech enhancement. We demonstrate that this can reduce the requirement for prior information about the noise, which can be difficult to estimate for fast-varying noise. Given noisy speech, the new approach estimates clean speech by recognizing long segments of the clean speech as whole units. In the recognition, clean speech sentences, taken from a speech corpus, are used as examples. Matching segments are identified between the noisy sentence and the corpus sentences. The estimate is formed by using the longest matching segments found in the corpus sentences. Longer speech segments as whole units contain more distinct dynamics and richer speaker characteristics, and can be identified more accurately from noise than shorter speech segments. Therefore, estimation based on the longest recognized segments increases the noise immunity and hence the estimation accuracy. The new approach consists of a statistical model to represent up to sentence-long temporal dynamics in the corpus speech, and an algorithm to identify the longest matching segments between the noisy sentence and the corpus sentences. The algorithm is made more robust to noise uncertainty by introducing missing-feature based noise compensation into the corpus sentences. Experiments have been conducted on the TIMIT database for speech enhancement from various types of nonstationary noise including song, music, and crosstalk speech. The new approach has shown improved performance over conventional enhancement algorithms in both objective and subjective evaluations.

Decoding semantics across fMRI sessions with different stimulus modalities:a practical MVPA study

Both embodied and symbolic accounts of conceptual organization would predict partial sharing and partial differentiation between the neural activations seen for concepts activated via different stimulus modalities. But cross-participant and cross-session variability in BOLD activity patterns makes analyses of such patterns with MVPA methods challenging. Here, we examine the effect of cross-modal and individual variation on the machine learning analysis of fMRI data recorded during a word property generation task. We present the same set of living and non-living concepts (land-mammals, or work tools) to a cohort of Japanese participants in two sessions: the first using auditory presentation of spoken words; the second using visual presentation of words written in Japanese characters. Classification accuracies confirmed that these semantic categories could be detected in single trials, with within-session predictive accuracies of 80-90%. However cross-session prediction (learning from auditory-task data to classify data from the written-word-task, or vice versa) suffered from a performance penalty, achieving 65-75% (still individually significant at p « 0.05). We carried out several follow-on analyses to investigate the reason for this shortfall, concluding that distributional differences in neither time nor space alone could account for it. Rather, combined spatio-temporal patterns of activity need to be identified for successful cross-session learning, and this suggests that feature selection strategies could be modified to take advantage of this.

Synthesis of Walking Sounds for Alleviating Gait Disturbances in Parkinson’s Disease

Managing gait disturbances in people with Parkinson’s disease is a pressing challenge, as symptoms can contribute to injury and morbidity through an increased risk of falls. While drug-based interventions have limited efficacy in alleviating gait impairments, certain non-pharmacological methods, such as cueing, can also induce transient improvements to gait. The approach adopted here is to use computationally-generated sounds to help guide and improve walking actions. The first method described uses recordings of force data taken from the steps of a healthy adult which in turn were used to synthesize realistic gravel-footstep sounds that represented different spatio-temporal parameters of gait, such as step duration and step length. The second method described involves a novel method of sonifying, in real time, the swing phase of gait using real-time motion-capture data to control a sound synthesis engine. Both approaches explore how simple but rich auditory representations of action based events can be used by people with Parkinson’s to guide and improve the quality of their walking, reducing the risk of falls and injury. Studies with Parkinson’s disease patients are reported which show positive results for both techniques in reducing step length variability. Potential future directions for how these sound approaches can be used to manage gait disturbances in Parkinson’s are also discussed.

Investigating the influence of respiratory motion on the radiation induced bystander effect in modulated radiotherapy

Respiratory motion introduces complex spatio-temporal variations in the dosimetry of radiotherapy and may contribute towards uncertainties in radiotherapy planning. This study investigates the potential radiobiological implications occurring due to tumour motion in areas of geometric miss in lung cancer radiotherapy. A bespoke phantom and motor-driven platform to replicate respiratory motion and study the consequences on tumour cell survival in vitro was constructed. Human non-small-cell lung cancer cell lines H460 and H1299 were irradiated in modulated radiotherapy configurations in the presence and absence of respiratory motion. Clonogenic survival was calculated for irradiated and shielded regions. Direction of motion, replication of dosimetry by multi-leaf collimator (MLC) manipulation and oscillating lead shielding were investigated to confirm differences in cell survival. Respiratory motion was shown to significantly increase survival for out-of-field regions for H460/H1299 cell lines when compared with static irradiation (p <0.001). Significantly higher survival was found in the in-field region for the H460 cell line (p <0.030). Oscillating lead shielding also produced these significant differences. Respiratory motion and oscillatory delivery of radiation dose to human tumour cells has a significant impact on in- and out-of-field survival in the presence of non-uniform irradiation in this in vitro set-up. This may have important radiobiological consequences for modulated radiotherapy in lung cancer. 

Longitudinal proton probing of ultrafast and high-contrast laser-solid interactions

We have performed an experiment aimed at measuring self-generated magnetic fields produced in solids by high electron currents following high-intensity and high contrast short-pulse laser irradiation. This was done using longitudinal high resolution proton deflectometry. The experiment was performed at the Titan-JLF laser facility with a high-power short-pulse beam (700 fs, ~ 110 J) split into two beams irradiating two solid targets. One beam is used for the generation of protons and the other beam for the generation of the ultra-high currents of electrons and of the associated magnetic fields. This capability allows us to study the spatio-temporal evolution of the magnetic fields and its dependence on the laser intensity and target material. © Owned by the authors, published by EDP Sciences, 2013.

Holocene “turn-on” and evolution of the Southern Great Barrier Reef: Revisiting reef cores from the Capricorn Bunker Group

Extensive drilling of the Great Barrier Reef (GBR) in the 70s and 80s illuminated the main factors controlling reef growth during the Holocene. However, questions remain about: (1) the precise nature and timing of reef "turnon" or initiation, (2) whether consistent spatio-temporal patterns occur in the bio-sedimentologic response of the reef to Holocene sea-level rise then stability, and (3) how these factors are expressed in the context of the different evolutionary states (juvenile-mature-senile reefs). Combining 21 new C14-AMS and 146 existing recalibrated radiocarbon and U/Th ages, we investigated the detailed spatial and temporal variations in sedimentary facies and coralgal assemblages in fifteen cores across four reefs (Wreck, Fairfax, One Tree and Fitzroy) from the Southern GBR. Our newly defined facies and assemblages record distinct chronostratigraphic patterns in the cores, displaying both lateral zonation across the different reefs and shallowing upwards sequences, characterised by a transition from deep (Porites/faviids) to shallow (Acropora/Isopora) coral types. The revised reef accretion curves show a significant lag period, ranging from 0.7-2 ka, between flooding of the antecedent Pleistocene substrate and Holocene reef turn-on. This lag period and dominance of more environmentally tolerant early colonizers (e.g., domal Porites and faviids), suggests initial conditions that were unfavourable for coral growth. We contend that higher input of fine siliciclastic material from regional terrigenous sources, exposure to hydrodynamic forces and colonisation in deeper waters are the main factors influencing initially reduced growth and development. All four reefs record a time lag and we argue that the size and shape of the antecedent platform is most important in determining the duration between flooding and recolonisation of the Holocene reef. Finally, our study of Capricorn Bunker Group Holocene reefs suggests that the size and shape of the antecedent substrate has a greater impact on reef evolution and final evolutionary state (mature vs. senile), than substrate depth alone. 

Towards a general framework for social media research using big data

The increasing adoption of cloud computing, social networking, mobile and big data technologies provide challenges and opportunities for both research and practice. Researchers face a deluge of data generated by social network platforms which is further exacerbated by the co-mingling of social network platforms and the emerging Internet of Everything. While the topicality of big data and social media increases, there is a lack of conceptual tools in the literature to help researchers approach, structure and codify knowledge from social media big data in diverse subject matter domains, many of whom are from nontechnical disciplines. Researchers do not have a general-purpose scaffold to make sense of the data and the complex web of relationships between entities, social networks, social platforms and other third party databases, systems and objects. This is further complicated when spatio-temporal data is introduced. Based on practical experience of working with social media datasets and existing literature, we propose a general research framework for social media research using big data. Such a framework assists researchers in placing their contributions in an overall context, focusing their research efforts and building the body of knowledge in a given discipline area using social media data in a consistent and coherent manner.

Evidential event inference in transport video surveillance

This paper presents a new framework for multi-subject event inference in surveillance video, where measurements produced by low-level vision analytics usually are noisy, incomplete or incorrect. Our goal is to infer the composite events undertaken by each subject from noise observations. To achieve this, we consider the temporal characteristics of event relations and propose a method to correctly associate the detected events with individual subjects. The Dempster–Shafer (DS) theory of belief functions is used to infer events of interest from the results of our vision analytics and to measure conflicts occurring during the event association. Our system is evaluated against a number of videos that present passenger behaviours on a public transport platform namely buses at different levels of complexity. The experimental results demonstrate that by reasoning with spatio-temporal correlations, the proposed method achieves a satisfying performance when associating atomic events and recognising composite events involving multiple subjects in dynamic environments.

Mining Money: the generative capacities of scientific and legal knowledge practices in the making of resource economies

In 1997 a scandal associated with Bre-X, a junior mining firm, and its prospecting activities in Indonesia, exposed to public scrutiny the ways in which mineral exploration firms acquire, assess and report on scientific claims about the natural environment. At stake here was not just how investors understood the provisional nature of scientific knowledge, but also evidence of fraud. Contemporaneous mining scandals not only included the salting of cores, but also unreliable proprietary sample preparation and assay methods, mis-representations of visual field estimates as drilling results and ‘overly optimistic’ geological reports. This paper reports on initiatives taken in the wake of these scandals and prompted by the Mining Standards Task Force (TSE/OSC 1999). For regulators, mandated to increase investor confidence in Canada’s leading role within the global mining industry, efforts focused first and foremost upon identifying and removing sources of error and wilfulness within the production and circulation of scientific knowledge claims. A common goal cross-cutting these initiatives was ‘a faithful representation of nature’ (Daston and Galison 2010), however, as the paper argues, this was manifest in an assemblage of practices governed by distinct and rival regulative visions of science and the making of markets in claims about ‘nature’. These ‘practices of fidelity’, it is argued, can be consequential in shaping the spatial and temporal dynamics of the marketization of nature.

Vacuum birefringence in strong inhomogeneous electromagnetic fields

Birefringence is one of the fascinating properties of the vacuum of quantum electrodynamics (QED) in strong electromagnetic fields. The scattering of linearly polarized incident probe photons into a perpendicularly polarized mode provides a distinct signature of the optical activity of the quantum vacuum and thus offers an excellent opportunity for a precision test of nonlinear QED. Precision tests require accurate predictions and thus a theoretical framework that is capable of taking the detailed experimental geometry into account. We derive analytical solutions for vacuum birefringence which include the spatio-temporal field structure of a strong optical pump laser field and an x-ray probe. We show that the angular distribution of the scattered photons depends strongly on the interaction geometry and find that scattering of the perpendicularly polarized scattered photons out of the cone of the incident probe x-ray beam is the key to making the phenomenon experimentally accessible with the current generation of FEL/high-field laser facilities.

Validating the use of intrinsic markers in body feathers to identify inter‑individual differences in non‑breeding areas of northern fulmars

Many wildlife studies use chemical analyses to explore spatio-temporal variation in diet, migratory patterns and contaminant exposure. Intrinsic markers are particularly valuable for studying non-breeding marine predators, when direct methods of investigation are rarely feasible. However, any inferences regarding foraging ecology are dependent upon the time scale over which tissues such as feathers are formed. In this study, we validate the use of body feathers for studying non-breeding foraging patterns in a pelagic seabird, the northern fulmar. Analysis of carcasses of successfully breeding adult fulmars indicated that body feathers moulted between September and March, whereas analyses of carcasses and activity patterns suggested that wing feather and tail feather moult occurred during more restricted periods (September to October and September to January, respectively). By randomly sampling relevant body feathers, average values for individual birds were shown to be consistent. We also integrated chemical analyses of body feather with geolocation tracking data to demonstrate that analyses of δ13C and δ15N values successfully assigned 88 % of birds to one of two broad wintering regions used by breeding adult fulmars from a Scottish study colony. These data provide strong support for the use of body feathers as a tool for exploring non-breeding foraging patterns and diet in wide-ranging, pelagic seabirds.

Microbiological quality of bivalves: relevance to food safety

Os moluscos bivalves constituem um recurso haliêutico de elevada importância na economia (inter)nacional pelas suas características organolépticas, valor nutritivo e relevância na gastronomia tradicional. Não obstante, representam um produto alimentar de elevado risco para a saúde pública. A contaminação microbiológica (autóctone e antropogénica), sendo crónica nos bancos de bivalves das zonas estuarino-lagunares, constitui uma das principais preocupações associadas à segurança alimentar. Aquando da filtração inerente aos processos de respiração e alimentação, os bivalves bioacumulam passivamente microrganismos incluindo os patogénicos. A sua colocação no mercado impõe pois, prévia salubrização para níveis microbiológicos compatíveis com a legislação em vigor, salvaguardando a saúde pública. Apesar da monitorização das áreas de apanha e produção, das medidas de prevenção e da depuração, a ocorrência de surtos associados ao consumo de bivalves tem aumentado. Tal deve-se à insuficiente monitorização da contaminação microbiológica dos bivalves, contribuindo para uma gestão ineficaz do produto e consequente sub-valorização. O presente trabalho pretendeu caracterizar o estado de desenvolvimento do sector de exploração de bivalves em Portugal do ponto de vista da segurança alimentar, e analisar os aspectos cruciais da monitorização e da depuração do produto apresentando alternativas abrangentes e aplicáveis ao sector. Assim, desenvolveu-se uma metodologia de base molecular passível de adaptação à monitorização dos bivalves das zonas conquícolas, como alternativa ao método de referência vigente do Número Mais Provável que é baseado apenas na quantificação de Escherichia coli. O mexilhão (Mytilus edulis) da Ria de Aveiro, bivalve de interesse comercial a nível (inter)nacional serviu de modelo para a comparação de protocolos de extração de DNA. Esta metodologia foi desenvolvida de modo a que os métodos de extração de DNA sejam passíveis de aplicação a outras matrizes biológicas ou ambientais. Para além da detecção e quantificação directa de bactérias patogénicas, esta metodologia poderá ser aplicada à monitorização da transferência vertical microbiana nos bancos de bivalves bem como à caracterização da dinâmica espacio-temporal das populações microbianas no ambiente e à monitorização dos processos de depuração. Foi ainda abordado o potencial da aplicação de bacteriófagos ou de enzimas líticas para a optimização dos processos de purificação. O trabalho realizado e as perspectivas futuras propostas pretendem contribuir para a dinamização e requalificação do sector de exploração de bivalves através da melhoria do nível de segurança alimentar dos moluscos bivalves comercializados para alimentação humana, valorizando este recurso.