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.

Visual target separation determines the extent of generalisation between opposing visuomotor rotations

Here we investigated the influence of angular separation between visual and motor targets on concurrent adaptation to two opposing visuomotor rotations. We inferred the extent of generalisation between opposing visuomotor rotations at individual target locations based on whether interference (negative transfer) was present. Our main finding was that dual adaptation occurred to opposing visuomotor rotations when each was associated with different visual targets but shared a common motor target. Dual adaptation could have been achieved either within a single sensorimotor map (i.e. with different mappings associated with different ranges of visual input), or by forming two different internal models (the selection of which would be based on contextual information provided by target location). In the present case, the pattern of generalisation was dependent on the relative position of the visual targets associated with each rotation. Visual targets nearest the workspace of the opposing visuomotor rotation exhibited the most interference (i.e. generalisation). When the minimum angular separation between visual targets was increased, the extent of interference was reduced. These results suggest that the separation in the range of sensory inputs is the critical requirement to support dual adaptation within a single sensorimotor mapping.

Max separation clustering for feature extraction from Optical Emission Spectroscopy Data

This paper proposes max separation clustering (MSC), a new non-hierarchical clustering method used for feature extraction from optical emission spectroscopy (OES) data for plasma etch process control applications. OES data is high dimensional and inherently highly redundant with the result that it is difficult if not impossible to recognize useful features and key variables by direct visualization. MSC is developed for clustering variables with distinctive patterns and providing effective pattern representation by a small number of representative variables. The relationship between signal-to-noise ratio (SNR) and clustering performance is highlighted, leading to a requirement that low SNR signals be removed before applying MSC. Experimental results on industrial OES data show that MSC with low SNR signal removal produces effective summarization of the dominant patterns in the data.

Construction of Drug–Polymer Thermodynamic Phase Diagrams Using Flory–Huggins Interaction Theory: Identifying the Relevance of Temperature and Drug Weight Fraction to Phase Separation within Solid Dispersions

Amorphous drug-polymer solid dispersions have the potential to enhance the dissolution performance and thus bioavailability of BCS class II drug compounds. The principle drawback of this approach is the limited physical stability of amorphous drug within the dispersion. Accurate determination of the solubility and miscibility of drug in the polymer matrix is the key to the successful design and development of such systems. In this paper, we propose a novel method, based on Flory-Huggins theory, to predict and compare the solubility and miscibility of drug in polymeric systems. The systems chosen for this study are (1) hydroxypropyl methylcellulose acetate succinate HF grade (HPMCAS-HF)-felodipine (FD) and (2) Soluplus (a graft copolymer of polyvinyl caprolactam-polyvinyl acetate-polyethylene glycol)-FD. Samples containing different drug compositions were mixed, ball milled, and then analyzed by differential scanning calorimetry (DSC). The value of the drug-polymer interaction parameter ? was calculated from the crystalline drug melting depression data and extrapolated to lower temperatures. The interaction parameter ? was also calculated at 25 °C for both systems using the van Krevelen solubility parameter method. The rank order of interaction parameters of the two systems obtained at this temperature was comparable. Diagrams of drug-polymer temperature-composition and free energy of mixing (?G mix) were constructed for both systems. The maximum crystalline drug solubility and amorphous drug miscibility may be predicted based on the phase diagrams. Hyper-DSC was used to assess the validity of constructed phase diagrams by annealing solid dispersions at specific drug loadings. Three different samples for each polymer were selected to represent different regions within the phase diagram

Investigation into mercury bound to biothiols:structural identification using ESI-ion-trap MS and introduction of a method for their HPLC separation with simultaneous detection by ICP-MS and ESI-MS

Mercury in plants or animal tissue is supposed to occur in the form of complexes formed with biologically relevant thiols (biothiols), rather than as free cation. We describe a technique for the separation and molecular identification of mercury and methylmercury complexes derived from their reactions with cysteine (Cys) and glutathione (GS): Hg(Cys)(2), Hg(GS)(2), MeHgCys, MeHgGS. Complexes were characterised by electrospray mass spectrometry (MS) equipped with an ion trap and the fragmentation pattern of MeHgCys was explained by using MP2 and B3LYP calculations, showing the importance of mercury-amine interactions in the gas phase. Chromatographic baseline separation was performed within 10 min with formic acid as the mobile phase on a reversed-phase column. Detection was done by online simultaneous coupling of ES-MS and inductively coupled plasma MS. When the mercury complexes were spiked in real samples (plant extracts), no perturbation of the separation and detection conditions was observed, suggesting that this method is capable of detecting mercury biothiol complexes in plants.

Developing an integrated hydrograph separation and lumped modelling approach to quantifying hydrological pathways in Irish river catchments

An appreciation of the quantity of streamflow derived from the main hydrological pathways involved in transporting diffuse contaminants is critical when addressing a wide range of water resource management issues. In order to assess hydrological pathway contributions to streams, it is necessary to provide feasible upper and lower bounds for flows in each pathway. An important first step in this process is to provide reliable estimates of the slower responding groundwater pathways and subsequently the quicker overland and interflow pathways. This paper investigates the effectiveness of a multi-faceted approach applying different hydrograph separation techniques, supplemented by lumped hydrological modelling, for calculating the Baseflow Index (BFI), for the development of an integrated approach to hydrograph separation. A semi-distributed, lumped and deterministic rainfall runoff model known as NAM has been applied to ten catchments (ranging from 5 to 699 km2). While this modelling approach is useful as a validation method, NAM itself is also an important tool for investigation. These separation techniques provide a large variation in BFI, a difference of 0.741 predicted for BFI in a catchment with the less reliable fixed and sliding interval methods and local minima turning point methods included. This variation is reduced to 0.167 with these methods omitted. The Boughton and Eckhardt algorithms, while quite subjective in their use, provide quick and easily implemented approaches for obtaining physically realistic hydrograph separations. It is observed that while the different separation techniques give varying BFI values for each of the catchments, a recharge coefficient approach developed in Ireland, when applied in conjunction with the Master recession Curve Tabulation method, predict estimates in agreement with those obtained using the NAM model, and these estimates are also consistent with the study catchments’ geology. These two separation methods, in conjunction with the NAM model, were selected to form an integrated approach to assessing BFI in catchments.

Improved Detection of Mycobacterium bovis Infection in Bovine Lymph Node Tissue Using Immunomagnetic Separation (IMS)-Based Methods

Immunomagnetic separation (IMS) can selectively isolate and concentrate Mycobacterium bovis cells from lymph node tissue to facilitate subsequent detection by PCR (IMS-PCR) or culture (IMS-MGIT). This study describes application of these novel IMS-based methods to test for M. bovis in a survey of 280 bovine lymph nodes (206 visibly lesioned (VL), 74 non-visibly lesioned (NVL)) collected at slaughter as part of the Northern Ireland bovine TB eradication programme. Their performance was evaluated relative to culture. Overall, 174 (62.1%) lymph node samples tested positive by culture, 162 (57.8%) by IMS-PCR (targeting IS6110), and 196 (70.0%) by IMS-MGIT culture. Twelve (6.9%) of the 174 culture positive lymph node samples were not detected by either of the IMS-based methods. However, an additional 78 M. bovis positive lymph node samples (26 (12.6%) VL and 54 (73.0%) NVL) were detected by the IMS-based methods and not by culture. When low numbers of viable M. bovis are present in lymph nodes (e.g. in NVLs of skin test reactor cattle) decontamination prior to culture may adversely affect viability, leading to false negative culture results. In contrast, IMS specifically captures whole M. bovis cells (live, dead or potentially dormant) which are not subject to any deleterious treatment before detection by PCR or MGIT culture. During this study only 2.7% of NVL lymph nodes tested culture positive, whereas 73% of the same samples tested M. bovis positive by the IMS-based tests. Results clearly demonstrate that not only are the IMS-based methods more rapid but they have greater detection sensitivity than the culture approach currently used for the detection of M. bovis infection in cattle.. Adoption of the IMS-based methods for lymph node testing would have the potential to improve M. bovis detection in clinical samples.

Production and evaluation of the utility of novel phage display-derived peptide ligands to Salmonella spp. for magnetic separation

Aims: The objectives of this study were to produce Salmonella-specific peptide ligands by phage display biopanning and evaluate their use for magnetic separation (MS).
Methods and Results: Four phage display biopanning rounds were performed and the peptides expressed by the two most Salmonella-specific (on the basis of phage binding ELISA results) phage clones, MSal020401 and MSal020417, were chemically synthesized and coupled to MyOne™ tosylactivated Dynabeads®. Peptide capture capability for whole Salmonella cells from non-enriched broth cultures was quantified by MS + plate counts and MS + Greenlight™ detection, and compared to capture capability of anti-Salmonella (antibody-coated) Dynabeads®. MS + Greenlight™ gave a more comprehensive picture of capture capability than MS + plate counts and showed that Peptide MSal020417-coated beads exhibited at least similar, if not better, capture capability to anti-Salmonella Dynabeads® (mean capture values of 36.0 ± 18.2 % and 31.2 ± 20.1 %, respectively, over Salmonella spp. concentration range 3 x 101 - 3 x 106 cfu ml-1) with minimal cross-reactivity (= 1.9 %) to three other foodborne bacteria.
Conclusions: One of the phage display-derived peptide ligands was demonstrated by MS + Greenlight™ to be a viable antibody-alternative for MS of Salmonella spp.
Significance and Impact of Study: This study demonstrates an antibody-free approach to Salmonella detection and opens substantial possibilities for more rapid tests for this bacterium.

Programmable motion and separation of single magnetic particles on patterned magnetic surfaces

Motion of single micrometer-sized magnetic particles on patterned magnetic surfaces is controlled by a rotating magnetic field (see Figure and cover). Patterns of thin-film magnetic elements are tailored to form transport lines. Individual particles are separated by adding junctions to the transport lines. The method can improve existing applications in biotechnology and generate new ones in life sciences.