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.

Three-dimensional flow-through protein platform

We have developed a new protein microarray (Immuno-Flow Protein Platform, IFPP) that utilizes a porous nitrocellulose (NC) membrane with printed spots of capture probes. The sample is pumped actively through the NC membrane, to enhance binding efficiency and introduce stringency. Compared to protein microarrays assayed with the conventional incubation-shaking method the rate of binding is enhanced on the IFPP by at least a factor of 10, so that the total assay time can be reduced drastically without compromising sensitivity. Similarly, the sensitivity can be improved. We demonstrate the detection of 1 pM of C-reactive protein (CRP) in 70 mu L of plasma within a total assay time of 7 min. The small sample and reagent volumes, combined with the speed of the assay, make our IFPP also well-suited for a point-of-care/near-patient setting. The potential clinical application of the IFPP is demonstrated by validating CRP detection both in human plasma and serum samples against standard clinical laboratory methods.

Large-scale analysis of soccer matches using spatiotemporal tracking data

Although the collection of player and ball tracking data is fast becoming the norm in professional sports, large-scale mining of such spatiotemporal data has yet to surface. In this paper, given an entire season's worth of player and ball tracking data from a professional soccer league (approx 400,000,000 data points), we present a method which can conduct both individual player and team analysis. Due to the dynamic, continuous and multi-player nature of team sports like soccer, a major issue is aligning player positions over time. We present a "role-based" representation that dynamically updates each player's relative role at each frame and demonstrate how this captures the short-term context to enable both individual player and team analysis. We discover role directly from data by utilizing a minimum entropy data partitioning method and show how this can be used to accurately detect and visualize formations, as well as analyze individual player behavior.

Identifying team style in soccer using formations learned from spatiotemporal tracking data

To the trained-eye, experts can often identify a team based on their unique style of play due to their movement, passing and interactions. In this paper, we present a method which can accurately determine the identity of a team from spatiotemporal player tracking data. We do this by utilizing a formation descriptor which is found by minimizing the entropy of role-specific occupancy maps. We show how our approach is significantly better at identifying different teams compared to standard measures (i.e., shots, passes etc.). We demonstrate the utility of our approach using an entire season of Prozone player tracking data from a top-tier professional soccer league.

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%.

Fred’s Flow (Canada) and Murphy Well (Australia) : thick komatiitic lava flows with contrasting compositions, emplacement mechanisms and water contents

Two Archaean komatiitic flows, Fred’s Flow in Canada and the Murphy Well Flow in Australia, have similar thicknesses (120 and 160 m) but very different compositions and internal structures. Their contrasting differentiation profiles are keys to determine the cooling and crystallization mechanisms that operated during the eruption of Archaean ultramafic lavas. Fred’s Flow is the type example of a thick komatiitic basalt flow. It is strongly differentiated and consists of a succession of layers with contrasting textures and compositions. The layering is readily explained by the accumulation of olivine and pyroxene in a lower cumulate layer and by evolution of the liquid composition during downward growth of spinifex-textured rocks within the upper crust. The magmas that erupted to form Fred’s Flow had variable compositions, ranging from 12 to 20 wt% MgO, and phenocryst contents from 0 to 20 vol%. The flow was emplaced by two pulses. A first ~20-m-thick pulse was followed by another more voluminous but less magnesian pulse that inflated the flow to its present 120 m thickness. Following the second pulse, the flow crystallized in a closed system and differentiated into cumulates containing 30–38 wt% MgO and a residual gabbroic layer with only 6 wt% MgO. The Murphy Well Flow, in contrast, has a remarkably uniform composition throughout. It comprises a 20-m-thick upper layer of fine-grained dendritic olivine and 2–5 vol% amygdales, a 110–120 m intermediate layer of olivine porphyry and a 20–30 m basal layer of olivine orthocumulate. Throughout the flow, MgO contents vary little, from only 30 to 33 wt%, except for the slightly more magnesian basal layer (38–40 wt%). The uniform composition of the flow and dendritic olivine habits in the upper 20 m point to rapid cooling of a highly magnesian liquid with a composition like that of the bulk of the flow. Under equilibrium conditions, this liquid should have crystallized olivine with the composition Fo94.9, but the most magnesian composition measured by electron microprobe in samples from the flow is Fo92.9. To explain these features, we propose that the parental liquid contained around 32 wt% MgO and 3 wt% H2O. This liquid degassed during the eruption, creating a supercooled liquid that solidified quickly and crystallized olivine with non-equilibrium textures and compositions.

A hybrid maximum power point tracking for partially shaded photovoltaic systems in the tropics

Partial shading and rapidly changing irradiance conditions significantly impact on the performance of photovoltaic (PV) systems. These impacts are particularly severe in tropical regions where the climatic conditions result in very large and rapid changes in irradiance. In this paper, a hybrid maximum power point (MPP) tracking (MPPT) technique for PV systems operating under partially shaded conditions witapid irradiance change is proposed. It combines a conventional MPPT and an artificial neural network (ANN)-based MPPT. A low cost method is proposed to predict the global MPP region when expensive irradiance sensors are not available or are not justifiable for cost reasons. It samples the operating point on the stairs of I–V curve and uses a combination of the measured current value at each stair to predict the global MPP region. The conventional MPPT is then used to search within the classified region to get the global MPP. The effectiveness of the proposed MPPT is demonstrated using both simulations and an experimental setup. Experimental comparisons with four existing MPPTs are performed. The results show that the proposed MPPT produces more energy than the other techniques and can effectively track the global MPP with a fast tracking speed under various shading patterns.

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.

Modeling kinect sensor noise for improved 3D reconstruction and tracking

We contribute an empirically derived noise model for the Kinect sensor. We systematically measure both lateral and axial noise distributions, as a function of both distance and angle of the Kinect to an observed surface. The derived noise model can be used to filter Kinect depth maps for a variety of applications. Our second contribution applies our derived noise model to the KinectFusion system to extend filtering, volumetric fusion, and pose estimation within the pipeline. Qualitative results show our method allows reconstruction of finer details and the ability to reconstruct smaller objects and thinner surfaces. Quantitative results also show our method improves pose estimation accuracy. © 2012 IEEE.

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.

Hydrochemical evolution and groundwater flow processes in the Galilee and Eromanga basins, Great Artesian Basin, Australia: A multivariate statistical approach

The Galilee and Eromanga basins are sub-basins of the Great Artesian Basin (GAB). In this study, a multivariate statistical approach (hierarchical cluster analysis, principal component analysis and factor analysis) is carried out to identify hydrochemical patterns and assess the processes that control hydrochemical evolution within key aquifers of the GAB in these basins. The results of the hydrochemical assessment are integrated into a 3D geological model (previously developed) to support the analysis of spatial patterns of hydrochemistry, and to identify the hydrochemical and hydrological processes that control hydrochemical variability. In this area of the GAB, the hydrochemical evolution of groundwater is dominated by evapotranspiration near the recharge area resulting in a dominance of the Na–Cl water types. This is shown conceptually using two selected cross-sections which represent discrete groundwater flow paths from the recharge areas to the deeper parts of the basins. With increasing distance from the recharge area, a shift towards a dominance of carbonate (e.g. Na–HCO3 water type) has been observed. The assessment of hydrochemical changes along groundwater flow paths highlights how aquifers are separated in some areas, and how mixing between groundwater from different aquifers occurs elsewhere controlled by geological structures, including between GAB aquifers and coal bearing strata of the Galilee Basin. The results of this study suggest that distinct hydrochemical differences can be observed within the previously defined Early Cretaceous–Jurassic aquifer sequence of the GAB. A revision of the two previously recognised hydrochemical sequences is being proposed, resulting in three hydrochemical sequences based on systematic differences in hydrochemistry, salinity and dominant hydrochemical processes. The integrated approach presented in this study which combines different complementary multivariate statistical techniques with a detailed assessment of the geological framework of these sedimentary basins, can be adopted in other complex multi-aquifer systems to assess hydrochemical evolution and its geological controls.

Predicting flow and transport in highly heterogeneous alluvial aquifers

Successful prediction of groundwater flow and solute transport through highly heterogeneous aquifers has remained elusive due to the limitations of methods to characterize hydraulic conductivity (K) and generate realistic stochastic fields from such data. As a result, many studies have suggested that the classical advective-dispersive equation (ADE) cannot reproduce such transport behavior. Here we demonstrate that when high-resolution K data are used with a fractal stochastic method that produces K fields with adequate connectivity, the classical ADE can accurately predict solute transport at the macrodispersion experiment site in Mississippi. This development provides great promise to accurately predict contaminant plume migration, design more effective remediation schemes, and reduce environmental risks. Key Points Non-Gaussian transport behavior at the MADE site is unraveledADE can reproduce tracer transport in heterogeneous aquifers with no calibrationNew fractal method generates heterogeneous K fields with adequate connectivity