Speech Enhancement from Additive Noise and Channel Distortion - a Corpus-Based Approach

This paper presents a new approach to single-channel speech enhancement involving both noise and channel distortion (i.e., convolutional noise). The approach is based on finding longest matching segments (LMS) from a corpus of clean, wideband speech. The approach adds three novel developments to our previous LMS research. First, we address the problem of channel distortion as well as additive noise. Second, we present an improved method for modeling noise. Third, we present an iterative algorithm for improved speech estimates. In experiments using speech recognition as a test with the Aurora 4 database, the use of our enhancement approach as a preprocessor for feature extraction significantly improved the performance of a baseline recognition system. In another comparison against conventional enhancement algorithms, both the PESQ and the segmental SNR ratings of the LMS algorithm were superior to the other methods for noisy speech enhancement. Index Terms: corpus-based speech model, longest matching segment, speech enhancement, speech recognition

Enhancement of ion generation in femtosecond ultraintense laser-foil interactions by defocusing

A simple method to enhance ion generation with femtosecond ultraintense lasers is demonstrated experimentally by defocusing laser beams on target surface. When the laser is optimally defocused, we find that the population of medium and low energy protons from ultra-thin foils is increased significantly while the proton cutoff energy is almost unchanged. In this way, the total proton yield can be enhanced by more than 1 order, even though the peak laser intensity drops. The depression of the amplified spontaneous emission (ASE) effect and the population increase of moderate-energy electrons are believed to be the main reasons for the effective enhancement. © 2012 American Institute of Physics.

An Investigation of Compressor Map Width Enhancement and the Inducer Flow Field Using Various Configurations of Shroud Bleed Slot

This paper describes an investigation of various shroud bleed slot configurations of a centrifugal compressor using CFD with a manual multi-block structured grid generation method. The compressor under investigation is used in a turbocharger application for a heavy duty diesel engine of approximately 400hp. The baseline numerical model has been developed and validated against experimental performance measurements. The influence of the bleed slot flow field on a range of operating conditions between surge and choke has been analysed in detail. The impact of the returning bleed flow on the incidence at the impeller blade leading edge due to its mixing with the main through-flow has also been studied. From the baseline geometry, a number of modifications to the bleed slot width have been proposed, and a detailed comparison of the flow characteristics performed. The impact of slot variations on the inlet incidence angle has been investigated, highlighting the improvement in surge and choked flow capability. Along with this, the influence of the bleed slot on stabilizing the blade passage flow by the suction of the tip and over-tip vortex flow by the slot has been considered near surge.

Synapsin- and actin-dependent frequency enhancement in mouse hippocampal mossy fiber synapses

The synapsin proteins have different roles in excitatory and inhibitory synaptic terminals. We demonstrate a differential role between types of excitatory terminals. Structural and functional aspects of the hippocampal mossy fiber (MF) synapses were studied in wild-type (WT) mice and in synapsin double-knockout mice (DKO). A severe reduction in the number of synaptic vesicles situated more than 100 nm away from the presynaptic membrane active zone was found in the synapsin DKO animals. The ultrastructural level gave concomitant reduction in F-actin immunoreactivity observed at the periactive endocytic zone of the MF terminals. Frequency facilitation was normal in synapsin DKO mice at low firing rates (approximately 0.1 Hz) but was impaired at firing rates within the physiological range (approximately 2 Hz). Synapses made by associational/commissural fibers showed comparatively small frequency facilitation at the same frequencies. Synapsin-dependent facilitation in MF synapses of WT mice was attenuated by blocking F-actin polymerization with cytochalasin B in hippocampal slices. Synapsin III, selectively seen in MF synapses, is enriched specifically in the area adjacent to the synaptic cleft. This may underlie the ability of synapsin III to promote synaptic depression, contributing to the reduced frequency facilitation observed in the absence of synapsins I and II.

Amphiphysin is a component of clathrin coats formed during synaptic vesicle recycling at the lamprey giant synapse

Amphiphysin is a protein enriched at mammalian synapses thought to function as a clathrin accessory factor in synaptic vesicle endocytosis. Here we examine the involvement of amphiphysin in synaptic vesicle recycling at the giant synapse in the lamprey. We show that amphiphysin resides in the synaptic vesicle cluster at rest and relocates to sites of endocytosis during synaptic activity. It accumulates at coated pits where its SH3 domain, but not its central clathrin/AP-2-binding (CLAP) region, is accessible for antibody binding. Microinjection of antibodies specifically directed against the CLAP region inhibited recycling of synaptic vesicles and caused accumulation of clathrin-coated intermediates with distorted morphology, including flat patches of coated presynaptic membrane. Our data provide evidence for an activity-dependent redistribution of amphiphysin in intact nerve terminals and show that amphiphysin is a component of presynaptic clathrin-coated intermediates formed during synaptic vesicle recycling.

Understanding Graduate Recruitment, Development and Retention for the Enhancement of Talent Management: Sharpening “the edge” of Graduate Talent

Graduates are deemed to be a key source of talent within many organisations and thus recruiting, developing and retaining them is viewed as a logical talent management (TM) strategy. However, there has been little attention paid to university graduates as part of an organisation’s TM strategy. Such a specific focus addresses the need for further research into the segmentation of talent pools and the specific challenges different talent pools are likely to create. This research, which utilised a qualitative data collection strategy, examined the experiences and practices of six large UK organisations in relation to graduate TM. Drawing from Gallardo-Gallardo, Dries and González-Cruz’s (2013. What is the meaning of ‘talent’ in the world of work? Human Resource Management Review, 23, 290–300.) framework for the conceptualisation of talent, the findings from this research indicate and explain why graduate employers are frequently compelled to use the object approach (talent as characteristics of people) due to the unique characteristics that recent graduates possess, even though other studies have found that a subject approach (talent as people and what they do) is preferred by most employers. Ultimately, employers conceptualise graduate talent by what they describe as ‘the edge’ which needs to be ‘sharpened’ to fully realise the potential that graduates offer.

Wide Matching - An Approach to Improving Noise Robustness for Speech Enhancement

It is shown that under certain conditions it is possible to obtain a good speech estimate from noise without requiring noise estimation. We study an implementation of the theory, namely wide matching, for speech enhancement. The new approach performs sentence-wide joint speech segment estimation subject to maximum recognizability to gain noise robustness. Experiments have been conducted to evaluate the new approach with variable noises and SNRs from -5 dB to noise free. It is shown that the new approach, without any estimation of the noise, significantly outperformed conventional methods in the low SNR conditions while retaining comparable performance in the high SNR conditions. It is further suggested that the wide matching and deep learning approaches can be combined towards a highly robust and accurate speech estimator.

Enhancement of Wear and Corrosion Resistance of Beta Titanium Alloy by Laser Gas Alloying with Nitrogen

The relatively high elastic modulus coupled with the presence of toxic vanadium (V) in Ti6Al4 V alloy has long been a concern in orthopaedic applications. To solve the problem, a variety of non-toxic and low modulus beta-titanium (beta-Ti) alloys have been developed. Among the beta-Ti alloy family, the quaternary Ti-Nb-Zr-Ta (TNZT) alloys have received the highest attention as a promising replacement for Ti6Al4 V due to their lower elastic modulus and outstanding long term stability against corrosion in biological environments. However, the inferior wear resistance of TNZT is still a problem that must be resolved before commercialising in the orthopaedic market. In this work, a newly-developed laser surface treatment technique was employed to improve the surface properties of Ti-35.3Nb-7.3Zr-5.7Ta alloy. The surface structure and composition of the laser-treated TNZT surface were examined by grazing incidence x-ray diffraction (GI-XRD) and x-ray photoelectron spectroscopy (XPS). The wear and corrosion resistance were evaluated by pin-on-plate sliding test and anodic polarisation test in Hanks’ solution. The experimental results were compared with the untreated (or base) TNZT material. The research findings showed that the laser surface treatment technique reported in this work can effectively improve the wear and corrosion resistance of TNZT.

Gamma-ray Spectra and Enhancement Factors for Positron Annihilation with Core Electrons

Many-body theory is developed to calculate the γ spectra for positron annihilation in noble-gas atoms. Inclusion of electron-positron correlation effects and core annihilation gives spectra in excellent agreement with experiment [K. Iwata et al., Phys. Rev. Lett. 79, 39 (1997)]. The calculated correlation enhancement factors γnl for individual electron orbitals nl are found to scale with the ionization energy I_nl (in eV), as γnl=1+ √A/Inl+(B/I_nl)β, where A≈40  eV, B≈24  eV, and β≈2.3.

Generating Data Augmented Spectroscopic Data For Performance Enhancement

The application of chemometrics in food science has revolutionized the field by allowing the creation of models able to automate a broad range of applications such as food authenticity and food fraud detection. In order to create effective and general models able to address the complexity of real life problems, a vast amount of varied training samples are required. Training dataset has to cover all possible types of sample and instrument variability. However, acquiring a varied amount of samples is a time consuming and costly process, in which collecting samples representative of the real world variation is not always possible, specially in some application fields. To address this problem, a novel framework for the application of data augmentation techniques to spectroscopic data has been designed and implemented. This is a carefully designed pipeline of four complementary and independent blocks which can be finely tuned depending on the desired variance for enhancing model's robustness: a) blending spectra, b) changing baseline, c) shifting along x axis, and d) adding random noise.
This novel data augmentation solution has been tested in order to obtain highly efficient generalised classification model based on spectroscopic data. Fourier transform mid-infrared (FT-IR) spectroscopic data of eleven pure vegetable oils (106 admixtures) for the rapid identification of vegetable oil species in mixtures of oils have been used as a case study to demonstrate the influence of this pioneering approach in chemometrics, obtaining a 10% improvement in classification which is crucial in some applications of food adulteration.

Behavioral modeling optimization and enhancement for high-speed analog mixed-signal I/O interfaces

A integridade do sinal em sistemas digitais interligados de alta velocidade, e avaliada através da simulação de modelos físicos (de nível de transístor) é custosa de ponto vista computacional (por exemplo, em tempo de execução de CPU e armazenamento de memória), e exige a disponibilização de detalhes físicos da estrutura interna do dispositivo. Esse cenário aumenta o interesse pela alternativa de modelação comportamental que descreve as características de operação do equipamento a partir da observação dos sinais eléctrico de entrada/saída (E/S). Os interfaces de E/S em chips de memória, que mais contribuem em carga computacional, desempenham funções complexas e incluem, por isso, um elevado número de pinos. Particularmente, os buffers de saída são obrigados a distorcer os sinais devido à sua dinâmica e não linearidade. Portanto, constituem o ponto crítico nos de circuitos integrados (CI) para a garantia da transmissão confiável em comunicações digitais de alta velocidade. Neste trabalho de doutoramento, os efeitos dinâmicos não-lineares anteriormente negligenciados do buffer de saída são estudados e modulados de forma eficiente para reduzir a complexidade da modelação do tipo caixa-negra paramétrica, melhorando assim o modelo standard IBIS. Isto é conseguido seguindo a abordagem semi-física que combina as características de formulação do modelo caixa-negra, a análise dos sinais eléctricos observados na E/S e propriedades na estrutura física do buffer em condições de operação práticas. Esta abordagem leva a um processo de construção do modelo comportamental fisicamente inspirado que supera os problemas das abordagens anteriores, optimizando os recursos utilizados em diferentes etapas de geração do modelo (ou seja, caracterização, formulação, extracção e implementação) para simular o comportamento dinâmico não-linear do buffer. Em consequência, contributo mais significativo desta tese é o desenvolvimento de um novo modelo comportamental analógico de duas portas adequado à simulação em overclocking que reveste de um particular interesse nas mais recentes usos de interfaces de E/S para memória de elevadas taxas de transmissão. A eficácia e a precisão dos modelos comportamentais desenvolvidos e implementados são qualitativa e quantitativamente avaliados comparando os resultados numéricos de extracção das suas funções e de simulação transitória com o correspondente modelo de referência do estado-da-arte, IBIS.