Speaker verification using I-vector features

This PhD research has provided novel solutions to three major challenges which have prevented the wide spread deployment of speaker recognition technology: (1) combating enrolment/ verification mismatch, (2) reducing the large amount of development and training data that is required and (3) reducing the duration of speech required to verify a speaker. A range of applications of speaker recognition technology from forensics in criminal investigations to secure access in banking will benefit from the research outcomes.

Higher order spectra based support vector machine for arrhythmia classification

Heart rate variability (HRV) refers to the regulation of the sinoatrial node, the natural pacemaker of the heart by the sympathetic and parasympathetic branches of the autonomic nervous system. HRV analysis is an important tool to observe the heart’s ability to respond to normal regulatory impulses that affect its rhythm. Like many bio-signals, HRV signals are non-linear in nature. Higher order spectral analysis (HOS) is known to be a good tool for the analysis of non-linear systems and provides good noise immunity. A computer-based arrhythmia detection system of cardiac states is very useful in diagnostics and disease management. In this work, we studied the identification of the HRV signals using features derived from HOS. These features were fed to the support vector machine (SVM) for classification. Our proposed system can classify the normal and other four classes of arrhythmia with an average accuracy of more than 85%.

Classification of echolocation calls from 14 species of bat by Support Vector Machines and Ensembles of Neural Networks

Calls from 14 species of bat were classified to genus and species using discriminant function analysis (DFA), support vector machines (SVM) and ensembles of neural networks (ENN). Both SVMs and ENNs outperformed DFA for every species while ENNs (mean identification rate – 97%) consistently outperformed SVMs (mean identification rate – 87%). Correct classification rates produced by the ENNs varied from 91% to 100%; calls from six species were correctly identified with 100% accuracy. Calls from the five species of Myotis, a genus whose species are considered difficult to distinguish acoustically, had correct identification rates that varied from 91 – 100%. Five parameters were most important for classifying calls correctly while seven others contributed little to classification performance.

Reliable Fault Diagnosis for Low-Speed Bearings Using Individually Trained Support Vector Machines With Kernel Discriminative Feature Analysis

This paper proposes a highly reliable fault diagnosis approach for low-speed bearings. The proposed approach first extracts wavelet-based fault features that represent diverse symptoms of multiple low-speed bearing defects. The most useful fault features for diagnosis are then selected by utilizing a genetic algorithm (GA)-based kernel discriminative feature analysis cooperating with one-against-all multicategory support vector machines (OAA MCSVMs). Finally, each support vector machine is individually trained with its own feature vector that includes the most discriminative fault features, offering the highest classification performance. In this study, the effectiveness of the proposed GA-based kernel discriminative feature analysis and the classification ability of individually trained OAA MCSVMs are addressed in terms of average classification accuracy. In addition, the proposedGA- based kernel discriminative feature analysis is compared with four other state-of-the-art feature analysis approaches. Experimental results indicate that the proposed approach is superior to other feature analysis methodologies, yielding an average classification accuracy of 98.06% and 94.49% under rotational speeds of 50 revolutions-per-minute (RPM) and 80 RPM, respectively. Furthermore, the individually trained MCSVMs with their own optimal fault features based on the proposed GA-based kernel discriminative feature analysis outperform the standard OAA MCSVMs, showing an average accuracy of 98.66% and 95.01% for bearings under rotational speeds of 50 RPM and 80 RPM, respectively.

DS-VBF : dual sink vector-based routing protocol for underwater wireless sensor network

Underwater wireless sensor networks (UWSNs) have become the seat of researchers' attention recently due to their proficiency to explore underwater areas and design different applications for marine discovery and oceanic surveillance. One of the main objectives of each deployed underwater network is discovering the optimized path over sensor nodes to transmit the monitored data to onshore station. The process of transmitting data consumes energy of each node, while energy is limited in UWSNs. So energy efficiency is a challenge in underwater wireless sensor network. Dual sinks vector based forwarding (DS-VBF) takes both residual energy and location information into consideration as priority factors to discover an optimized routing path to save energy in underwater networks. The modified routing protocol employs dual sinks on the water surface which improves network lifetime. According to deployment of dual sinks, packet delivery ratio and the average end to end delay are enhanced. Based on our simulation results in comparison with VBF, average end to end delay reduced more than 80%, remaining energy increased 10%, and the increment of packet reception ratio was about 70%.

Prediction of GMA welding characteristic parameter by artificial neural network system

Nowadays, demand for automated Gas metal arc welding (GMAW) is growing and consequently need for intelligent systems is increased to ensure the accuracy of the procedure. To date, welding pool geometry has been the most used factor in quality assessment of intelligent welding systems. But, it has recently been found that Mahalanobis Distance (MD) not only can be used for this purpose but also is more efficient. In the present paper, Artificial Neural Networks (ANN) has been used for prediction of MD parameter. However, advantages and disadvantages of other methods have been discussed. The Levenberg–Marquardt algorithm was found to be the most effective algorithm for GMAW process. It is known that the number of neurons plays an important role in optimal network design. In this work, using trial and error method, it has been found that 30 is the optimal number of neurons. The model has been investigated with different number of layers in Multilayer Perceptron (MLP) architecture and has been shown that for the aim of this work the optimal result is obtained when using MLP with one layer. Robustness of the system has been evaluated by adding noise into the input data and studying the effect of the noise in prediction capability of the network. The experiments for this study were conducted in an automated GMAW setup that was integrated with data acquisition system and prepared in a laboratory for welding of steel plate with 12 mm in thickness. The accuracy of the network was evaluated by Root Mean Squared (RMS) error between the measured and the estimated values. The low error value (about 0.008) reflects the good accuracy of the model. Also the comparison of the predicted results by ANN and the test data set showed very good agreement that reveals the predictive power of the model. Therefore, the ANN model offered in here for GMA welding process can be used effectively for prediction goals.

Influence of the structural characteristic of pyrolysis products on thermal stability of styrene-butadiene rubber composites reinforced by different particle sized kaolinites

A series of rubber composites were prepared by blending styrene-butadiene rubber (SBR) latex and the different particle sized kaolinites. The thermal stabilities of the rubber composites were characterized using thermogravimetry, digital photography, scanning electron microscopy, X-ray diffraction, Fourier transform infrared spectroscopy, and Raman spectroscopy. Kaolinite SBR composites showed much greater thermal stability when compared with that of the pure SBR. With the increase of kaolinite particle size, the pyrolysis products became much looser; the char layer and crystalline carbon content gradually decreased in the pyrolysis residues. The pyrolysis residues of the SBR composites filled with the different particle sized kaolinites showed some remarkable changes in structural characteristics. The increase of kaolinite particle size was not beneficial to form the compact and stable crystalline carbon in the pyrolysis process, and resulted in a negative influence in improving the thermal stability of kaolinite/SBR composites.

GPU accelerated algorithms for computing matrix function vector products with applications to exponential integrators and fractional diffusion

The efficient computation of matrix function vector products has become an important area of research in recent times, driven in particular by two important applications: the numerical solution of fractional partial differential equations and the integration of large systems of ordinary differential equations. In this work we consider a problem that combines these two applications, in the form of a numerical solution algorithm for fractional reaction diffusion equations that after spatial discretisation, is advanced in time using the exponential Euler method. We focus on the efficient implementation of the algorithm on Graphics Processing Units (GPU), as we wish to make use of the increased computational power available with this hardware. We compute the matrix function vector products using the contour integration method in [N. Hale, N. Higham, and L. Trefethen. Computing Aα, log(A), and related matrix functions by contour integrals. SIAM J. Numer. Anal., 46(5):2505–2523, 2008]. Multiple levels of preconditioning are applied to reduce the GPU memory footprint and to further accelerate convergence. We also derive an error bound for the convergence of the contour integral method that allows us to pre-determine the appropriate number of quadrature points. Results are presented that demonstrate the effectiveness of the method for large two-dimensional problems, showing a speedup of more than an order of magnitude compared to a CPU-only implementation.

Assessment of antenna characteristic effects on pedestrian and cyclists travel-time estimation based on Bluetooth and WiFi MAC addresses

Monitoring pedestrian and cyclists movement is an important area of research in transport, crowd safety, urban design and human behaviour assessment areas. Media Access Control (MAC) address data has been recently used as potential information for extracting features from people’s movement. MAC addresses are unique identifiers of WiFi and Bluetooth wireless technologies in smart electronics devices such as mobile phones, laptops and tablets. The unique number of each WiFi and Bluetooth MAC address can be captured and stored by MAC address scanners. MAC addresses data in fact allows for unannounced, non-participatory, and tracking of people. The use of MAC data for tracking people has been focused recently for applying in mass events, shopping centres, airports, train stations etc. In terms of travel time estimation, setting up a scanner with a big value of antenna’s gain is usually recommended for highways and main roads to track vehicle’s movements, whereas big gains can have some drawbacks in case of pedestrian and cyclists. Pedestrian and cyclists mainly move in built distinctions and city pathways where there is significant noises from other fixed WiFi and Bluetooth. Big antenna’s gains will cover wide areas that results in scanning more samples from pedestrians and cyclists’ MAC device. However, anomalies (such fixed devices) may be captured that increase the complexity and processing time of data analysis. On the other hand, small gain antennas will have lesser anomalies in the data but at the cost of lower overall sample size of pedestrian and cyclist’s data. This paper studies the effect of antenna characteristics on MAC address data in terms of travel-time estimation for pedestrians and cyclists. The results of the empirical case study compare the effects of small and big antenna gains in order to suggest optimal set up for increasing the accuracy of pedestrians and cyclists’ travel-time estimation.

Role of enhanced vector transmission of a new West Nile virus strain in an outbreak of equine disease in Australia in 2011

Background In 2011, a variant of West Nile virus Kunjin strain (WNVKUN) caused an unprecedented epidemic of neurological disease in horses in southeast Australia, resulting in almost 1,000 cases and a 9% fatality rate. We investigated whether increased fitness of the virus in the primary vector, Culex annulirostris, and another potential vector, Culex australicus, contributed to the widespread nature of the outbreak. Methods Mosquitoes were exposed to infectious blood meals containing either the virus strain responsible for the outbreak, designated WNVKUN2011, or WNVKUN2009, a strain of low virulence that is typical of historical strains of this virus. WNVKUN infection in mosquito samples was detected using a fixed cell culture enzyme immunoassay and a WNVKUN- specific monoclonal antibody. Probit analysis was used to determine mosquito susceptibility to infection. Infection, dissemination and transmission rates for selected days post-exposure were compared using Fisher’s exact test. Virus titers in bodies and saliva expectorates were compared using t-tests. Results There were few significant differences between the two virus strains in the susceptibility of Cx. annulirostris to infection, the kinetics of virus replication and the ability of this mosquito species to transmit either strain. Both strains were transmitted by Cx. annulirostris for the first time on day 5 post-exposure. The highest transmission rates (proportion of mosquitoes with virus detected in saliva) observed were 68% for WNVKUN2011 on day 12 and 72% for WNVKUN2009 on day 14. On days 12 and 14 post-exposure, significantly more WNVKUN2011 than WNVKUN2009 was expectorated by infected mosquitoes. Infection, dissemination and transmission rates of the two strains were not significantly different in Culex australicus. However, transmission rates and the amount of virus expectorated were significantly lower in Cx. australicus than Cx. annulirostris. Conclusions The higher amount of WNVKUN2011 expectorated by infected mosquitoes may be an indication that this virus strain is transmitted more efficiently by Cx. annulirostris compared to other WNVKUN strains. Combined with other factors, such as a convergence of abundant mosquito and wading bird populations, and mammalian and avian feeding behaviour by Cx. annulirostris, this may have contributed to the scale of the 2011 equine epidemic.

VIP Barcoding: Composition vector-based software for rapid species identification based on DNA barcoding

Species identification based on short sequences of DNA markers, that is, DNA barcoding, has emerged as an integral part of modern taxonomy. However, software for the analysis of large and multilocus barcoding data sets is scarce. The Basic Local Alignment Search Tool (BLAST) is currently the fastest tool capable of handling large databases (e.g. >5000 sequences), but its accuracy is a concern and has been criticized for its local optimization. However, current more accurate software requires sequence alignment or complex calculations, which are time-consuming when dealing with large data sets during data preprocessing or during the search stage. Therefore, it is imperative to develop a practical program for both accurate and scalable species identification for DNA barcoding. In this context, we present VIP Barcoding: a user-friendly software in graphical user interface for rapid DNA barcoding. It adopts a hybrid, two-stage algorithm. First, an alignment-free composition vector (CV) method is utilized to reduce searching space by screening a reference database. The alignment-based K2P distance nearest-neighbour method is then employed to analyse the smaller data set generated in the first stage. In comparison with other software, we demonstrate that VIP Barcoding has (i) higher accuracy than Blastn and several alignment-free methods and (ii) higher scalability than alignment-based distance methods and character-based methods. These results suggest that this platform is able to deal with both large-scale and multilocus barcoding data with accuracy and can contribute to DNA barcoding for modern taxonomy. VIP Barcoding is free and available at http://msl.sls.cuhk.edu.hk/vipbarcoding/.

The significance of controlled conditions in lentiviral vector titration and in the use of multiplicity of infection (MOI) for predicting gene transfer events

Background: Although lentiviral vectors have been widely used for in vitro and in vivo gene therapy researches, there have been few studies systematically examining various conditions that may affect the determination of the number of viable vector particles in a vector preparation and the use of Multiplicity of Infection (MOI) as a parameter for the prediction of gene transfer events. Methods: Lentiviral vectors encoding a marker gene were packaged and supernatants concentrated. The number of viable vector particles was determined by in vitro transduction and fluorescent microscopy and FACs analyses. Various factors that may affect the transduction process, such as vector inoculum volume, target cell number and type, vector decay, variable vector - target cell contact and adsorption periods were studied. MOI between 0-32 was assessed on commonly used cell lines as well as a new cell line. Results: We demonstrated that the resulting values of lentiviral vector titre varied with changes of conditions in the transduction process, including inoculum volume of the vector, the type and number of target cells, vector stability and the length of period of the vector adsorption to target cells. Vector inoculum and the number of target cells determine the frequencies of gene transfer event, although not proportionally. Vector exposure time to target cells also influenced transduction results. Varying these parameters resulted in a greater than 50-fold differences in the vector titre from the same vector stock. Commonly used cell lines in vector titration were less sensitive to lentiviral vector-mediated gene transfer than a new cell line, FRL 19. Within 0-32 of MOI used transducing four different cell lines, the higher the MOI applied, the higher the efficiency of gene transfer obtained. Conclusion: Several variables in the transduction process affected in in vitro vector titration and resulted in vastly different values from the same vector stock, thus complicating the use of MOI for predicting gene transfer events. Commonly used target cell lines underestimated vector titre. However, within a certain range of MOI, it is possible that, if strictly controlled conditions are observed in the vector titration process, including the use of a sensitive cell line, such as FRL 19 for vector titration, lentivector-mediated gene transfer events could be predicted. © 2004 Zhang et al; licensee BioMed Central Ltd.

Role of predictors in downscaling surface temperature to river basin in India for IPCC SRES scenarios using support vector machine

In this paper, downscaling models are developed using a support vector machine (SVM) for obtaining projections of monthly mean maximum and minimum temperatures (T-max and T-min) to river-basin scale. The effectiveness of the model is demonstrated through application to downscale the predictands for the catchment of the Malaprabha reservoir in India, which is considered to be a climatically sensitive region. The probable predictor variables are extracted from (1) the National Centers for Environmental Prediction (NCEP) reanalysis dataset for the period 1978-2000, and (2) the simulations from the third-generation Canadian Coupled Global Climate Model (CGCM3) for emission scenarios A1B, A2, B1 and COMMIT for the period 1978-2100. The predictor variables are classified into three groups, namely A, B and C. Large-scale atmospheric variables Such as air temperature, zonal and meridional wind velocities at 925 nib which are often used for downscaling temperature are considered as predictors in Group A. Surface flux variables such as latent heat (LH), sensible heat, shortwave radiation and longwave radiation fluxes, which control temperature of the Earth's surface are tried as plausible predictors in Group B. Group C comprises of all the predictor variables in both the Groups A and B. The scatter plots and cross-correlations are used for verifying the reliability of the simulation of the predictor variables by the CGCM3 and to Study the predictor-predictand relationships. The impact of trend in predictor variables on downscaled temperature was studied. The predictor, air temperature at 925 mb showed an increasing trend, while the rest of the predictors showed no trend. The performance of the SVM models that are developed, one for each combination of predictor group, predictand, calibration period and location-based stratification (land, land and ocean) of climate variables, was evaluated. In general, the models which use predictor variables pertaining to land surface improved the performance of SVM models for downscaling T-max and T-min

A frame-invariant scheme for the geometrically exact beam using rotation vector parametrization

While frame-invariant solutions for arbitrarily large rotational deformations have been reported through the orthogonal matrix parametrization, derivation of such solutions purely through a rotation vector parametrization, which uses only three parameters and provides a parsimonious storage of rotations, is novel and constitutes the subject of this paper. In particular, we employ interpolations of relative rotations and a new rotation vector update for a strain-objective finite element formulation in the material framework. We show that the update provides either the desired rotation vector or its complement. This rules out an additive interpolation of total rotation vectors at the nodes. Hence, interpolations of relative rotation vectors are used. Through numerical examples, we show that combining the proposed update with interpolations of relative rotations yields frame-invariant and path-independent numerical solutions. Advantages of the present approach vis-a-vis the updated Lagrangian formulation are also analyzed.